EXTREME Overclocking Forums
Home | Reviews | Forums | Downloads | $ EXTREME Deals $ | RealTime Pricing | Free Magazines | Gear | Folding Stats Newsletter | Contact Us


Go Back   EXTREME Overclocking Forums > General CPU Cooling > Water Cooling - (Plain & Chilled)
Register Forum Rules FAQ Today's Posts

Welcome Guest Visitor! Please Register, It's Free and Fun To Participate!
The EXTREME Overclocking Forums are a place for people to learn how to overclock and tweak their PC's components like the CPU, memory (RAM), or video card in order to gain the maximum performance out of their system. There are lots of discussions about new processors, graphics cards, cooling products, power supplies, cases, and so much more!

You are currently viewing our boards as a "guest" which gives you limited access to view most discussions. You need to register before you can post: click the register link to proceed. Before you register, please read the forum rules. By joining our free community you will have access to post topics, communicate privately with other members (PM), respond to polls, upload your own pictures, and access many other special features. Registration is fast, simple, and absolutely free! To start viewing messages, select the forum that you want to visit from the selection below.

After you have registered and read the forum rules, you can check out the FAQ for more information on using the forum. We hope you enjoy your stay here!

Note To Spammers: We do not allow unsolicited advertising! Spam is usually reported & deleted within minutes of it being posted, so don't waste your time (or ours)!


Please Register to Post a Reply
 
Thread Tools
Old 05-14-2011, 02:18 PM   #1
musicfan
ambient water-cooling
musicfan's Avatar
Senior Member
 
Posts: 2,745
Last Seen: Today
Age: 59
From: Oregon, USA
iTrader: 0 / 0%
Post $$$ 2014 Water-Cooling Parts-Buying Guide $$$

The Water-Cooling Parts-Buying Guide updated for 2014

Beginning Water-Cooling Guide


All-in-One Buying Guide

1. Introduction.
This EOCF Parts-Buying Guide sticky updated for 2014 is to help choose newer parts that were unavailable for ricecrispi's EOCF Parts-Buying Guide for 2009.
This Guide is not the Beginning Water-Cooling Guide. Start with the Beginning Water-Cooling Guide if you are new to custom water-cooling. The Beginning Guide will tell you when to come to this Guide to choose parts.

In this Guide, one can either read from top to bottom like a long post or go
to the section relevant to the part. This Guide is for ambient custom/kit watercooling with no discussion of sub-ambient cooling or All-in-Ones. There is a separate All-in-One Buying Guide.

This Guide is structured so the Tables compare parts such as blocks or pumps.
Below the table are Table Notes to explain the table. Below table notes are Sources that link to testing. Finally, Results are discussed. To lessen bias, I calculate an average or arithmetic mean of result numbers along with Standard Deviation (SD). Statistics are imperfect but allow grouping of similar-performing parts. For more about statistics, please see explanations and links in the final section.

To save time, most sections contain Suggestions, which are only my opinions. These are to keep beginners out of trouble - never to hinder the adventuresome.

Test results are color-coded by relative performance testing.
Green ink shows excellent results (usually to 2 SD). It also show most desirable dimensions such as slim radiators, which are most likely to fit.
Blue ink shows good results, better than average by at least 1 SD.
Black ink showed average performance. Average performance among among good performers might mean a part "passes the test", so black ink is often good.
Red ink cautions beginners to STOP, THINK, and ASK. With results in the red zone, anticipate unique performance. Ask questions on a thread if you are uncertain why something is flagged in red.

Carefully read testing sources in the links. When ready, do searches, read reviews, and start a thread to ask other water-coolers about the part, especially if there have been any surprises since the product was released.

2. CPU Water-Blocks. Suggestion: Look at the table immediately below. Buy any Green-colored #1 Thermal testing, especially if overclocking. Buy a Green #1 Flow-rating if possible. The Blue-colored #2 Flow-testing is more restrictive but can work well with a good pump or small loop. Avoid buying red or read the testing sources to see why it is in red ink. The most consistent mounting temperatures in Stren's 2012 testing were with the Koolance 380, MIPS, EK Supremacy, and Apogee HD - so those should be the products with the best designed mounting hardware. In June 2013, Intel released the Haswell CPU for socket LGA 1150. Fortunately, mounting holes are identical to 1155/1156 motherboards.



Table Notes:
1. Blocks
were tested on Intel Core i7 CPU's. A unique model, the Swiftech Apogee HD, was added in 2011 after independent testing by Jay Hall here confirmed proprietary Swiftech testing. The Apogee HD has one inlet and 3 outlets. It can work in single CPU systems but is designed for parallel flow into multiple blocks with Swiftech custom rads.
2. Prices are Intel October 2012. Blocks are low to high $US to avoid bias. Similar models often come in different materials at different prices. Blocks with an asterisk (*) are no longer available but remain for comparisons.
3. In November 2008, Martinm210 published an Intel Q6600 CPU Block Shootout. The Danger Den MC-TDX at $60 (May 2011) graded between the Apogee GT and Fuzion V1 so is likely a 2 thermal. It is a 1 flow. The Danger Den CPU MC-TDX water-block is not in the table only because no independent testing data on the Intel core i7 was found. Martinm210 has good data on other older blocks tested on an overclocked Q6600 at the same Shootout link.
4. In the above table, the best performers are green, > 1x SD from mean is blue, and > 2x SD from mean is red. Therefore, 1 = best, 2 = fair, 3 = least.

5. Vendors:
www.swiftech.com; www.sidewindercomputers.com; www.performance-pcs.com; www.koolance.com; www.jab-tech.com; www.dangerden.com; www.aquatuning.us; www.evga.com; www.frozencpu.com; www.dtwaterblocks.com/
6. In 2012 Danger Den closed, but many DD blocks remain in circulation.
7. In 2013 Detroit Thermal (DT) Waterblocks closed. MIPS closed also.

Testing Sources:
Bundymania [English]: http://www.techpowerup.com/forums/sh...d.php?t=153781
Skinnee Labs: http://skinneelabs.com/cpu-blocks/
Martin's Lab: http://martinsliquidlab.org/
Jay Hall's testing of the Apogee HD:
http://www.xtremesystems.org/forums/...her-Top-Blocks
Cartago's testing of the Apogee HD: http://www.xtremesystems.org/forums/...ago%B4s-Review
st0ned's testing of the EK Supremacy and DT Sniper: http://www.xtremesystems.org/forums/showthread.php?281687-EK-Supremacy-Review
rge's testing of the DT Sniper and EK Supreme HF: http://www.overclockers.com/forums/showthread.php?t=710046
stren's testing of CPU blocks in 2012: http://www.xtremesystems.org/forums/...-Block-Roundup or http://www.xtremerigs.net/reviews/wa...block-roundup/and in 2014: http://www.xtremerigs.net/2014/02/07...block-testing/
moonmanovich's testing of CPU blocks in 2014: http://www.overclock.net/t/1505481/s...-round-up-2014

Results:
Bundymania published the largest group, N = 22 on a 1366 platform. The mean CPU-water delta T was 42.1°C +/- 1.7°C [1SD]. The thermal results are so close that the worst above are > 1 SD but < 2 SD from the mean so are recorded in blue. Mean flow was 155 L/h +/- 8.6 L/h. Flow > 1 SD worse than mean are blue and those > 2 SD or the worst are in red. Skinnee Labs' and Martin's Lab data were reviewed including roundup #3 which uses pressure drop instead of flow rate to measure restriction. The Apogee XT rev.2 is 30% less restrictive than rev.1 but pressure drop is still > 1 SD from the mean so is recorded in blue. The Aqua Computer Cuplex Kryos is slightly more restrictive than the Apogee XT rev.2 so is also recorded in blue. In September 2011, Martin tested the XSPC Raystorm. Compared to other blocks he tested it was the lowest restriction and tied the Koolance 370 for best thermal performance so is listed as a 1/1 performance block along with other top blocks. In November 2011, Jay Hall tested the Apogee HD with uniform mounting pressure and found thermals similar to the Raystorm and slightly superior to the Koolance 370 and EK Supreme HF. Flow was better than the Apogee XT rev.2 but still showed "moderate restriction".

In January 2012 Martinm210 tested the Danger Den M-6 on an Intel 2600k and the link to his lab is above. Cartago's testing confirmed earlier results and is reflected in the thermal and flow table. In March 2012, Martinm210 tested the Detroit Thermal 5Noz. Temps were excellent and restriction very low. Martin also tested the value model EK LTX and performance was good although slightly less than the EK Supreme HF metal top block. In June 2012 rge posted thermal testing showing the EK Supreme HF about 1.5°C better cpu core-water delta than the DT Sniper. In June 2012 st0ned posted flow and thermal testing comparing the EK Supremacy, DT Sniper, and Phobya UC-1. The EK Supremacy and DT Sniper demonstrated best flow. The EK Supremacy core-water delta was 2°C better than the DT Sniper or Phobya UC-1.

In July of 2012 stren posted a testing for 15 CPU blocks on a socket 2011 Sandy Bridge E system. He got a CPU core - water delta of 41.9°C +/- 1.2°C [1SD]. Stren introduced some new blocks and his data exposed some aging blocks and perhaps differences between socket 2011 and 1155 or 1366. Some changes were made in the table above after his data was reviewed. For example, the Rasa showed its age and the Alphacool has a threading problem at this time (see photos from rge). Stren also made an interesting table showing the standard deviation of 5 temperature differences after mounting a block as a proxy for mounting consistency. The MIPS, EK Supremacy, and Apogee HD were best while the Danger Den, Heatkiller, and Rasa were the only blocks one SD worse than mean on mount variation. Mounting temperature variation is one more performance aspect to consider. In October, he tested the Koolance CPU-380, which finished in the top tier.


In February 2014, stren posted testing of the same blocks he tested above on the Sandy Bridge E but this time on the Haswell 4770k, a smaller CPU die, to see if there were major differences between the blocks (linked above). He found no significant differences stating the decision whether to de-lid the 4770k was more important than choice of which premium water-block to buy.

In August 2014, moonmanovich posted testing of elven blocks from eight manufacturers. The distance between best and worst performers was about 7°C on the Haswell LGA 1150 and 5°C Sandy Bridge-E LGA 2001. The best performer on each system was the EK Supremacy. He also tested the Swiftech Apogee XL and got relatively poor results - worse than the Swiftech's older Apogee HD. This placed the new Swiftech block and the budget EK LTX near the bottom. He felt there was a mounting issue with the new Swiftech XL - maybe they can fix it.

Discussion: There are several models that are best performers in both thermal and flow/restriction performance with the entry level around $50. The EK LT and Swiftech Apogee XTL are kit blocks and not as advanced as premium models. The XTL can be upgraded to an XT rev.2 for $30 but still has below average flow. The Apogee HD with one inlet and 3 outlets offers many options for parallel flow. At this time, there is no reason to purchase a poor performing CPU block. Top CPU waterblocks will perform within a few degrees of each other so also consider mounting ease and TIM choice. Buyers can choose based on cost, materials, appearance, mounting, fitting compatibility, CPU brand, restrictiveness of other blocks inside the loop, serial or parallel loop designs, or total pumping power available.

3. Graphics Card Water-Blocks. Suggestion: Buy Green, Blue, or Black GPU, VRM and Flow. Discuss Red on a thread. Google for new reviews. Koolance's Connecting Video Water Blocks page discusses nozzle orientation, multiple VGA blocks, and serial versus parallel flow design for multiple VGA blocks. Block manufacturers, like EK, make FC bridges or SLI/Crossfire connectors of serial or parallel flow design to fit their blocks. Increasingly, more manufacturers make SLI connectors, so check websites before buying. Many like the bridges but either tubing or Koolance's VID connectors can also work to connect multiple VGA card blocks.



2013 nVidia GTX Titan/780 full cover testing by stren - see below. He has tables, video, and a discussion thread.

2014 AMD R9 290X full cover testing by stren - see below. Final results are pending but AC is looking best as of March 2014.

Table Notes
:
1. Skinnee Labs tested nVidia gtx480 blocks only. But one tester with an AMD/ATI 6950 card thermally compared his XSPC Razor block favorably to his EK.
2. Prices: nVidia GTX 480 are April 2011, GTX 580 are April 2012, and AMD HD7970 are February 2013. Blocks are listed low to high but prices don't include backplate or fittings.
3. EVGA Hydro Copper prices include the VGA card. The design for the GTX 580 is full-cover while the GTX 480 was GPU universal hybrid-air. Swiftech is OEM for EVGA hydro blocks.
4. In the Skinnee Labs testing work thread, one poster defends EVGA/Swiftech saying it has a warranty others lack. If warranty is important; please read it before removing any air cooling to apply any water block.
5. Flow testing is more accurate by Skinnee Labs and stren since they did pressure drop testing. Bundymania only read flow value from a meter those results provide only a relative value specific to that system.

In 2010, Skinnee Labs compared eight full-cover blocks on an nVidia gtx480
: http://skinneelabs.com/gtx480-full-cover-block-roundup
[SL site down summer 2013 but the workthread at XS still exists - see Table Notes 4 for XS link]. GPU thermal data show stock and overclock results correlate. Bitspower or BP is the least effective at cooling the GPU. Consider BP if one needs the ports coming straight up since it is the only block like that. VRM data are important but new so there is controversy among testers about importance of cooling VGA VRM compared to GPU. The EVGA Hydro Copper gtx480 is least effective at cooling VRM. Consider EVGA Hydro Copper if you need a warranty. Some in the work thread consider VRM temps as important as GPU temps. BP was > 2 SD from mean GPU temp. EVGA was > 2 SD from mean VRM temp. They got the red color #3 because they failed to cool as well as the #1 group in green that all clustered around mean temperature. The flow data shows > 1 SD from the mean in both the highest and lowest flows so there is a trend. This is why we need to keep the big picture in mind. In other words, our loops are far more than just one block on a test stand. For busy readers, seeing 1 best, 2 average, & 3 worst flow is important because it could change block choices when seeking SLI or X-fire with a weak pump inside a restrictive loop.

In 2012, Bundymania compared ten full-cover blocks on an nVidia gtx580 here in English. The results are similar to tests done previously on the gtx480 by Skinnee Labs except VRM temperatures were not tested on the gtx580. No pressure-drop testing was done this time so only flow results are available. Bundymania's top thermal performers within 1 SD of the mean temperature were EVGA/Swiftech, Koolance, Watercool Heatkiller, AquaComputer, EK, Alphacool, Liquid Extasy, and XSPC. Danger Den barely scored 1 SD higher than the mean so got a 2 in thermal performance on the gtx580 but was close to the XSPC. The Bitspower was 2 SD higher than the mean, so got the undesirable 3 again. Danger Den had best flow again and the Bitspower and EVGA/Swiftech blocks were 1 SD lower in flow from the mean (or more restrictive) so were rated 2. If they had tested 2 SD lower, they would have scored an undesirable 3.

Also in 2012, Bundymania compared nine full-cover blocks (six in June and three in September) on an AMD hd7970 full cover roundup. Thermally, Watercool and Aqua Computer models performed slightly better than average. The recently closed Danger Den manufacturer released the only model that gave significantly poorer than average GPU thermal results.
Flow is harder to interpret. Flow values show the Danger Den far (2 SD) above average with the remaining in the average to low-average range. Flow should not be a problem for most pumps with a single VGA card but no pressure drop was done. All flow values are relatively low when converted to GPM. VRM temperature recording was not done on the HD7970 testing. Those who overclock the HD7970 may want to choose from the middle (EK) to the top (Heatkiller, AC, Koolance, Swiftech) and avoid those on the bottom half (Alphacool, Liquid Extasy, XSPC Razor, and Danger Den) until we learn more. Some feel that reading VRM temp values is not accurate enough and an IR thermometer is needed. Others feel VGA VRM cooling on ambient systems does not limit OC. For more on VRM cooling for the HD7970 read here and here. For a discussion with links, see: http://forums.extremeoverclocking.co...d.php?t=368800.

In 2013, stren compared seven full-cover blocks on an nVidia Titan in his GTX 780/Titan roundup (website). He also made a video review of the roundup at YouTube. The XS discussion thread is here. He tested full-covers from Aquacomputer, EK, EVGA, Koolance, Swiftech, Watercool, and XSPC on an overclocked nVidia Titan. Bitspower, Alphacool, and Phobya failed to provide samples. Stren also did pressure-drop testing instead of simple flow readings. GPU core temps were good with all but XSPC was best including at low flow rates. VRM and VRAM was best with EK. He gave his awards to the XSPC and EK and said the Watercool and Koolance were close. The EVGA hydro uses a Swiftech Komodo block. There are many helpful pictures and graphs at stren's site. I recommend a thorough reading of his testing from his website before any VGA block purchase. There is also a good discussion, especially of VRM and VRAM temps at XS. The video is an hour long but is fantastic for learning about VGA full-cover waterblocks.

In January 2014, stren released testing water-block testing results for the AMD R9 290X at his website: http://www.xtremerigs.net/2014/01/01...mance-summary/ but the tests are ongoing. Currently AC has improved any problem with restriction and with the AC backplate is looking like the top block - beating EK and Koolance for VRM and VRAM cooling. As of March 2014, there is no data yet for XSPC Razer or WC Heatkiller or Swiftech Komodo.

Discussion:
Although newer VGA block reviews were posted on the Internet, Skinnee Labs (SL) was the only tester to place thermal probes on the VRM until stren did this in 2013. SL also measured pressure drops instead of only reading flow meters and stren did this too. VGA card testing is difficult. SL and sten each ruined the VGA card during final testing. And, models change rapidly making large roundups unusual. The cards tested above are premium single-GPU VGA cards and will run hotter than other single-GPU models in that card series. Checking TDP helps compare, but recall that TDP is without overclock. It is hard to know maximum heat production so reviewers sometimes use both synthetic testing like Furmark (generates the most heat) and game engine benchmarks (best to determine gaming stability) to analyze maximum power draw. For comparison, the TDP of the gtx 480 is 250 Watts, gtx 580 is 244 Watts, and the reference HD 7970 is 230 Watts. The tdp for the gtx 680 went down to 195 Watts. But in 2013, the Titan and gtx 780 both have TDP ratings of 250 W. The PCI slots are rated to 75 Watts, 6-pin VGA to 75 Watts and 8-pin to 150 Watts. In theory, one can deliver 300 Watts to a VGA card without exceeding specs. Some cards used power-throttling presumably to remain within specs. Water-cooling hot VGA cards is popular not only to control heat, especially from multiple VGA cards, but also to eliminate the small noisy fan often part of stock air-cooling. It is wise to check manufacturer sites for water-block compatibility and even better to write them. It is easier to find full-cover blocks for a reference VGA card, which is usually the prototype for the lineup and what block manufacturers use as a template. If your card is not reference, sometimes EK will make a block to fit (check their site) or you can use a universal (discussed below) to cool the GPU.

Universal GPU Blocks: Skinnee Labs tested universals on the thermally demanding nVidia gtx480 here: http://skinneelabs.com/universalgpu-nv480/8/. As water-coolers know, full-cover VGA blocks are sometimes hard to find making universal GPU blocks necessary. If you use a universal instead of a full-cover, apply heatsinks to the Voltage Regulator Modules or VRM and make sure there is good air-flow across the heatsinks. If overclocking, most recommend heatsinks on VGA memory chips too for higher OC. Here is a Watercooling Guide by rubix_1011. If you scroll near the bottom, lunaticwoda shows photos of VRM locations and describes what to cool. To add cooling to the universal GPU, one can use a Swiftech one-piece heatsink or individual heatsinks like these or these or these. Heatsinks can be shaped with a rotary-tool like a Dremel. Most brand name heatsinks come with double-sided thermal adhesive tape but it can be wise to have extra.

Materials: There is no performance advantage to Nickel plating Copper blocks. Nickel plating can flake off and whether EK's new plating method has lessened this risk is unclear. I buy copper only but that choice is one the buyer must make. Do you really want thin Nickel plate because it looks shinier even though no one sees it? If you need more information, read this thread: http://forums.extremeoverclocking.com/showthread.php?t=353071 and perform some searches. Read your block warranty too if that is important to you.

4. Full Motherboard Water-Blocks by Bitspower, Danger Den, EK, and Koolance were tested on an Asus Rampage III Extreme by Skinnee Labs and published in January 2011: http://skinneelabs.com/asus-r3e_blocks/. The motherboard does not create as much heat as a CPU or a VGA card so can usually be air-cooled. For example in Skinnee Labs' testing, stock motherboard temperature was never greater than 53.5°C even with an Intel Core i7 950 overclocked beyond 4 GHz and air-cooled gtx480. In spite of that; some parts of the motherboard, particularly mosfets around the CPU can be hard to cool. Spot cooling with a case fan usually works well. Replacing manufacturer's TIM with better TIM can improve cooling. Skinnee offers reasons why he uses a full motherboard block. Motherboard blocks are usually the most restrictive block. In this testing, EK is most restrictive while Danger Den is least but overall flow rates were marginal with a Jingway 1200 pump. Koolance and Bitspower gave lowest overall temps but one should research carefully before purchasing a full motherboard block. If you are interested and find a compatible board and block, discuss it on the forum.

5. Pumps.
Suggestion: Many consider Swiftech's PWM (Pulse Width Modulation) pumps ideal DC 12V water-cooling pumps. Green is best performance. Red performance pumps can work for small or low-restriction loops.



Table Notes:
1. The Laing D5 strong pump needs greater than 12 V to perform better than the Laing D5 vario at 12V. Converters are available from Koolance. The D5 frame/top has native 1/2 inch barbs. The PMP-500 also benefits from Koolance voltage control to lessen noise. This > before the base price is to factor in the extra price of a controller.
2. Older Laing DDC's and modifications remain available. The DDC has native 3/8 inch barbs but performs better with after-market tops, which also conveniently allow choice of G 1/4 inch fittings.
3. Prices are 2013 from FrozenCPU, Performance PC's, Sidewindercomputers, Jab-tech, or manufacturer site.
4. PWM is quieter when set up properly through BIOS.
5. There are other Laing and XSPC models but these are the more popular 12V DC. Performance PC's has a large inventory of both DC & AC water-pumps.
5. Removing a manufacturer's stock top for any reason can void warranty so read carefully.
6. In November 2012 Swiftech introduced the mcp655-PWM pump body that is sold exclusively through FrozenCPU. You need to choose a top for the mcp655 but a D5 frame/top adds only $10. The 35x
has Swiftech's top with the pump. Swiftech now makes PWM models in both D5 and DDC sizes.
7. For P/Q curves on the non-PWM DDC pump family including: Laing 3.1, 3.2, 3.15, and 3.25 models, link to https://www.dangerden.com/store/file...s/ddc/DDC3.pdf

Sources:
Skinnee Labs: http://skinneelabs.com/pumps/
Martin's Lab: http://martinsliquidlab.org/category/pumps/
Martin's Archive: http://martinsliquidlab.i4memory.com/DDC31PumpTopTesting.html

Bundymania [English]: http://www.overclock.net/water-cooli...reservoir.html

Helpful Link: Convert Units of Pressure Online Calculator

Discussion: The testing of the Laing D5/Swiftech 655/Koolance 450 at Skinnee Labs is very helpful. The D5 Vario has a manual dial up to setting 5. With pump flow at 1.5 GPM, the Laing Basic or B model has no vario dial so the head pressure is at setting 4 permanently. The OEM Jingway DP-600P approximates the Laing D5 vario head pressure on setting 3 while the XSPC 750 approximates the Laing D5 pressure on setting 2. This is why less powerful pumps work fine for a small, low restriction loop but may not provide adequate flow for a large loop with restrictive components. Components may be tested at 1.5 GPM but that flow rate can be noisy with minimal temperature improvement over 1.0 GPM. Many consider 1.0 GPM ideal so the table now has this flow rate listed for comparisons. Under 1.0 GPM, some encounter issues with bleeding air from the loop, including removing bubbles from bay reservoirs or from radiators with the ports at the bottom or ports pointed down
. Usually bleeding air is helped by gently rocking the case to move bubbles along and being patient as long as the pump is filling adequately. Cooling temperatures are affected relatively little by flow until reaching less than 0.7 GPM.

The new PWM control pumps by Swiftech and XSPC offer automatic pump-speed control from the BIOS. Adjusting the PWM controls can allow quieter pumping.

Here is Martin's New Pump Planning Guide. Martin's old Liquid Lab Pump and Radiator Optimizer Spreadsheet is no longer updated but is still useful. Martin explains:
Quote:
This spreadsheet was created to help other water coolers with estimating their flow rate. I’ll be providing updates to the estimator here. It is an educational Excel file I wanted to share with you to assist with pump and radiator planning.
6. Radiators. To help with understanding the tables including abbreviations, please read all table notes found below the tables. To read the testing, see the sources linked below the notes. The tables compare models and list dimensions. If you need a slim radiator to fit or do push/pull, look for green numbers in the size column. Radiator testing is very labor intensive and as you see from the tables, different testers test differently.

It's important when choosing size, to note that one 120.3 size radiator thermally performs nearly three times better than the same model 120.1. On the other hand, one 120.3 size radiator has nearly identical restriction as the same model 120.1.

The most frequent question about "how much radiator do I need?" is introduced in the Beginning Guide sticky. One should at least know the heat load of the components one wishes to water-cool. An XSPC RX 120.1 can cool around 300 Watts with a good air-water Delta T around 10°C if using a quality 38 mm fan at 2800 rpm. But that is loud and why understanding one's noise tolerance is important. A 25 mm thick 1000 rpm fan on the same 120.1 rad will likely be quiet but cools about 150 Watts with a similar Delta T. That is why using a "one size fits all" formula for every build can lead to surprises and understanding basics like heat load discussed in the Beginning sticky is important. Moreover, Intel CPU's tolerate higher load temperatures than AMD CPU desktop processors and extra cooling is necessary if seeking lower core temps. Source for XSPC RX 120.1 and RX 120.3 testing is Skinnee Labs.

Radiator Suggestion: Avoid red-flagged radiators because there are better alternatives. Radiators seldom contribute much restriction. But if the pump is weak or loop is large, note that radiators marked red in the drop/psi/flow column can be as restrictive as a CPU block. Fitting a radiator can be a challenge. My favorites include the Alphacool ST 30 because it is slim, performs at all speeds, and has guards to avoid puncturing the radiator water-tubes with screws. The slim XSPC RS offers best value. If you can fit thick rads, the Alphacool UT 60 performs best. Cooling is most efficient when one matches radiator and fan. Some choose fans prior to radiators because there are increasing fan choices including PWM, motor type, color choices, or lights. If you intend to buy new fans, skip ahead to Section 7, choose fans and return. After you know your fan rpm, it's easier to choose a radiator. As the suggestions states, simply avoid radiators that tested in the red. There is nothing wrong with a radiator that tests "average" without any glaring weakness. Then one buys needs enough radiator surface area at your maximum tolerated fan speed to cool your heat load with a reasonable air-water delta T such as 10°C.







Table Notes:
1. Prices are May 2012 $US low to high at sidewindercomputers, jab-tech, frozencpu, PerformancePCs, or Koolance. I could not find the Coolgate Acrylic for sale in the US.
2. Only 120.3 radiators with G 1/4" ports and standard fan spacing are listed. The table excludes reservoir/radiator combination or stacker models.
3. In 2008, Martin used his V1 test bench. Skinnee used a similar bench until December 2010. HESmelaugh and Bundymania used computers. In 2012, Martin used his V2 test bench as he explains at his site.
4. The Swiftech MCR 320 was tested each time and allows for limited cross-study comparison. Martin's 2008 MCR was not identified as a QP but has same size & fins per inch as a QP.
5. For thermal and flow/restriction testing, Black is average or mean. Blue is better (1 SD better than mean). Green is best (2 SD better). Red is worst (at least 1 SD worse than mean).

6. Radiator thickness in green is < 36 mm to indicate a desirable slim thickness radiator. Slim may fit or allow push/pull when the thick radiators won't. Blue is for medium thickness - not slim but not thick.
7. Radiator thermal resistance or C/W values were measured at 1.5 gallons/minute. Pressure Drop is listed in Pounds per Square Inch or PSI. dT = delta T in °Celsius. mm = millimeters. 1k = 1000.
8. Only some results can fit in the above tables including common fan speeds. For more details, read the sources.
9. Dimensions are approximate since manufacturers can change models. Aquatuning has a page of different-sized radiator templates for download.
10.The radiator comparison spreadsheet by BobbyBubblehead has more details including fins per inch (fpi) and other comparisons.

Sources:
1. http://martinsliquidlab.org/ [archive]

2. http://skinneelabs.com/radiators/
3. HESmelaugh provides no C/W or Pressure Drop. The dT is water-air for 290 Watts. Flow is Liters/hour. Testing in German and English
4. Bundymania 2011 provides no C/W or Pressure Drop. The dT is water-air and the flow is listed at Liters/hour on a test system. Translated to English.
5. Martinm210 in 2012:
http://martinsliquidlab.org/2012/04/...otout-summary/
6. Bundymania in 2012 is like 2011 with new gear. There are no C/W values or Pressure Drop. The dT is water-air on his test system: http://www.xtremesystems.org/forums/...with-22-Rads-!

Results & Discussion:
Martin's 2008 tests show the XSPC RS thermal testing inferior to the Swiftech MCR but not so in 2012 when testing methods changed. In the tables above, raw results were compared only within each test group but not across different test groups. Mean average and standard deviations were calculated on the numbers, after which the radiators were grouped. Mean average is in black ink. Results at least 1 SD worse are in red. The only results 2 SD worse than the mean were the Swiftech QP at 1800 and 2200 rpm in Martin's 2012 testing. Results 1 SD better than mean are blue. The HW Lab Black Ice Extreme in Skinnee's testing at high-flow 1.5 gpm was the only radiator ever performing 2 SD better so is in green. But it is in the one column only - the highest speed fans - with contrasting red performance at low rpm. All slim size radiators - 36 mm or less - have the height or thickness in green. I don't understand why there are marked flow differences between HESmelaugh's 2009 and Bundymania's 2011 tests but they seem to confirm relative ranking.

In 2012, Martin conducted 120.3 radiator testing on sixteen radiators. His 2012 test method on his newer V2 bench changed from his 2008 testing on his V1 bench and from Skinnee's 2009 testing, which was based on the V1. In 2012, Martin also used different fans than Skinnee did and added insulation. Martin explains here:
Quote:
Yeah...I think the retest with higher speed and higher pressure fans will favor the [HW Black Ice] GTX. At some point that double thickness high density will prevail. I just don't think it shows as well with these titans and with this fixed pumping power method. The titans are more average in cfm/rpm where Skinnee was testing with GT15s which makes nearly a 400 rpm change to the rpm scale at the right. In addition this fixed pumping power method more simulates actual gains and losses from restriction. The GTX is higher in restriction than average so it is testing at a slightly lower flow rate than the lower restriction radiators. Skinnee was testing with fixed flow rate which ignores restriction which is how I did it back on my V1 bench, but that gives a slightly unfair advantage to the more restrictive radiators.
Martin also explains here:
Quote:
Yeah, I'm still not quite sure what to think of the actual wattage numbers. They generally seem low, but all the rads are being tested the same, so relative performance should be ok. I guess insulating the lines and adding the enclosure to minimize heat losses makes a bigger difference than I thought. I always knew open air rad testing was a problem if there is any changes in air flow in the area of testing. I am probably also getting a touch of restriction from the test bench ports but I did try one test with and without the top cover and got results that were within a few percent so I don't think there is enough restriction to cause any major differences. I think most of it is in the line insulation work and testing all of the points with heat loads that near a 10°C delta.
And finally, we need to re-think how much of the heat load variable we are actually water-cooling versus how much escapes to air. Martin said here:
Quote:
We have very little to no data on actual heat dump numbers of blocks on processors or vga cards etc. I suspect the actual loss through the pcb and other areas is fairly high such that a 200w vga card may only dump 100-150 watts into the water cooling loop. We also always over stress parts in synthetic tests to the point where you may only see 50-80% of that in actual programs. There are more unknowns than understanding on how much heat we are actually dealing with.
It is complex to compare radiator results from the V1 bench to V2. One direct comparison (RS360 to RS360) requires about a doubling in heat dissipation for Martin's V2 values from 2012 to match his V1 results from 2008. This is probably from insulated versus open-air testing. Along with differences in fans and flow rates, we can understand why there are differences between between testing from Martin's V1 in 2008, Skinnee Labs' modified V1 from 2009-2011, and Martin's V2 in 2012. The results are valid and complementary but underscore how different methods change raw results.

In September 2012 Bundymania tested 22 radiators on his system and posted them in a roundup roundup linked above. He tests flow rate and delta t values at 600, 800, 1200, & 1500 rpm. He tested fewer new rads but adds helpful information at low rpm. No new table was made.

In December 2012 Martin published testing on the XSPC EX 120.3 at his site. Cooling was similar to the RS360 in the middle range between 1200 & 1800 rpm. Above 1800 rpm the RS cooled slightly better but below 1200 the EX dissipated slightly more heat than the RS model.

120.4 radiators: Bundymania's 120.4 testing is here in English.

140.3 radiators: A 140.3 radiator has about the same surface area as a 120.4 radiator so performance can be compared. For example, reading testing on a Koolance CU1020V 120.3 rad can give insight on the same model of 140.3. In Bundymania's testing of 140mm radiators, the HW Black Ice SR-1 140.3 equaled the Themochill PA 140.3. Both were better than Magicool which equaled the Feser Monsta Lite. Skinnee Labs tested a 140.3 Aqua Computer AirPlex Revolution that has 140 mm mounting on one side yet 120 mm on the other to offer flexible options for mounting or fans.

180.x radiators: Magicool makes a 180.1, 180.2, and 180.3 with specs here. Silverstone makes a 180 mm fan. The surface area of a 180.1 is between a 120.2 and 120.3 but there's no independent testing.

200.x radiators:
A 200.1 has similar surface area to a 140.2. A 200.2 is reported to fit the top of a Coolermaster HAF X. Phobya makes both a 200.1 and 200.2 but there is no independent testing.

Monster Radiators: In 2011, Bundymania compared the MO-RA3 and Phobya 1080/1260. The Phobya 1260 holds nine 140 mm fans or four 200 mm fans. 200 mm fans with higher rpm are now available. In 2012, Bundymania compared the Alphacool NexXxos Monsta 120.3 in a group of 22 radiators. In 2013, Bundymania compared the Phobya Xtreme Nova 1080, Phobya G-Changer 1080 60mm, Koolance 1080 20-FPI Copper, and Koolance 1080 18-FPI Aluminum.


7. Fans and Controllers.
Fans must generate enough air pressure to ventilate the radiator. A typical 3-pin connector single-speed fan has a 12 volt wire, ground, and tach wire coming out of the fan motor. It is designed to run at a single speed as measured in revolutions per minute (rpm). The 4-pin Pulse Width Modulation or PWM fan has a fourth wire that allows a 4-pin motherboard fan header, such as the CPU fan header, to control a PWM fan. Do not get the smaller 4-pin PWM connector mixed up with the larger Molex 4-slot connector. The fourth wire in the PWM fan allows the motherboard, when set properly, to automatically control the range of rpms designed into the fan, using motherboard information such as CPU core temperature. The fan electronics works similarly to the PWM pumps discussed above. The most common size fans are 120 mm where we have the most experience and choice. 140 mm fans come next and although the choices are not as great at 120 mm, they are improving. For other fan sizes, it's best to ask in a thread because there may be little testing.

Bluezero5 wrote a helpful fan tutorial where he explains how fans work and why good case fans don't necessarily make good radiator fans. Martinm210 performed extensive radiator fan testing. The Scythe Gentle Typhoon AP15 at 1850 rpm dominates the middle-speed 120 mm 3-pin radiator fan market because of good air flow relative to noise. Sometimes Gentle Typhoons are hard to find or don't fit a color scheme. The Scythe Kama Flow Two 1900 rpm, Noiseblocker 1800 rpm, or Aerocool Shark 1500 rpm are reliable middle-speed 3-pin alternatives. Budget 3-pin radiator fans include the Yate Loon brand. They make a high-speed 2200 rpm Yate Loon D12SH-12, medium-speed 1650 rpm, or low-speed 1350 rpm. Yate Loons have been used on water-cooling radiators for years and although loud at high rpm, they all run quietly at 5 volts or under 1000 rpm.

4-pin PWM fan choices are greater now. Premium PWM fans include the 1000 - 2000 rpm Noiseblocker PWM fan or the 300 - 1500 rpm Noctua NF-F12 PWM fan tested here. In 2012, Swiftech released the Helix120mm PWM fan designed for radiators. With a price tag of $12, good static pressure and attractive PWM range of 800 - 1800 rpm, the Swiftech Helix is becoming a popular 120 mm PWM radiator fan. Others seeking a silent, budget 120 mm PWM, mention the Arctic F12, rated at 300 - 1350 rpm and only costing around $10.

Although Scythe does not directly sell PWM Gentle Typhoons fans, the high-speed AP-29, AP-30, and AP-31 Gentle Typhoons have a fourth connector for a PWM wire. The medium speed AP-15 does not. Those with soldering skills can add PWM as explained by ehume here or here. Or one can purchase those same fans already modded. Be aware that PWM fans with powerful fan motors can still sound harsh even at 1000 rpm.

In 2008, Martinm210 tested shrouds to find an additional 15% cooling while push/pull fans added 20% more cooling (improved air-water delta t). He retested in 2012 with Gentle Typhoon AP15 1850 rpm fans and shrouds showed minimal benefit. Otherwise, push was slightly better than pull even at lower rpm and push/pull remained helpful. Push/pull advocates recommend identical fans but Martinm210 says identical push/pull fans may be unnecessary. There are no standards for how manufacturers derive fan specifications other than rpm and current draw. This means be skeptical when comparing specifications, especially when using that fan on a radiator.

In addition to Martinm210's extensive fan reviews at his website linked above, below are thorough reviews from Hardware.Fr posted in 2012 and 2013. The French can be converted to your language using Google translate:

Comparisons of 63 single-speed 120 mm fans: http://www.hardware.fr/articles/874-...-a-vs-cfm.html
Comparisons of 40 PWM 120 mm fans: http://www.hardware.fr/articles/867-...pitulatif.html
Comparisons of 23 single-speed 140 mm fans: http://www.hardware.fr/articles/886-...-a-vs-cfm.html

In addition to the above links, Google any fan you intend to purchase for recent reviews because fans enter and exit the market regularly. You can usually get an opinion on the forum too because some are passionate about fans.

Helpful Links: Decibels, Loudness, and Perception and How Much is a Decibel? Perception vs. Reality: What Our Ears Hear

Fan control
is the management of the rotational speed or rpm of an electrical fan, usually to create less noise. One can choose from many types to control a single fan or many, fit PSU power cables, fit PCI cover slots, or inside drive bays. A single-speed 3 pin fan is speed-controlled by design. Some simply use enough radiator to allow slower 3-pin fans to run directly from the PSU at 12 volts with no additional control. Nearly all single-speed fans under 1000 rpm are quiet and under 800 rpm, most fans approach silence. My 1200 rpm fans are not silent but are quiet and ventilate my radiators fine, even under load, so require no additional controller.

Options besides slow single-speed 3-pin fans include 4-pin PWM fans that cover the desired rpm range. The 4-pin fans can be plugged into CPU_Fan or other PWM 4-pin motherboard fan headers using a 4-pin splitter cable or splitter box. This allows controlling rpm of the fans using BIOS or software such as Speedfan while drawing power directly from the PSU. The on-off duty cycles of PWM electronics maintain voltage but generate little heat.

On the other hand, traditional manual fan controllers that use resistance to lower voltage can get hot at high resistance/lower voltage. When comparing between "analog" or "resistance" models, read reviews and look at photos for larger coils, more capacitors, and larger heatsinks so that air-cooling the controller is more effective. Some newer fan controllers that don't have heatsinks use PWM-type electronics to under-volt 3-pin fans. Although PWM type controllers don't get hot, there are reports of clicking or whining fan noises at slowed speeds (YouTube demo here). In this post, Martinm210 explainsfan controllers far better than I can.

Fan start-up amperage to overcome inertia may be important because some fans like Gentle Typhoons draw four to five times the current in amps to start when compared to the current required to maintain full speed. If manufacturer specs are available, one can do the math to stay under specs. Moreover, start-up power is transient and premium controllers allow some temporary overshoot. If you are going to daisy chain or harness many fans, do yourself a favor and purchase a better (more robust) product with a higher Watts rating. If you are not interested in warranty, you can use lessen risk of overload by estimating 5 Watts start-up power for a typical 25 mm 1500 rpm case fan. That means placing no more than six per 30 Watt channel. If you are stacking many fans on a channel, check the PCB for heat with an infrared thermometer or tape one of your controller's temperature probes nearby. If you touch a hot controller, you risk burns or static discharge. Under HEAT TEST Martinm210 explains heatsinks above 65°C are too hot. Testing found manufacturer's specs unreliable. Place a fan inside the case to blow directly onto the inside back of the heatsinks and PCB to prolong the life of your controller. On the AC Aquaero 5 XT, there is even a water-cooling option.

Overloading a controller or fan header risks premature failure. Controllers rated less than twenty Watts per channel are seldom designed for more than one fan per channel. Motherboard fan headers are seldom rated for more than one amp or twelve Watts or one fan per header. If you wish to keep the warranty, write the manufacturer before you buy and ask about fan brand/model compatibility and the total number of fans per channel they warranty. For example, Lamptron replied that 10 Gentle Typhoons can be run off of one 30 Watt channel in their FC5V2 model. On the other hand, Scythe responded to my email to run only one fan per twelve Watts channel in their Kaze Server. That doesn't mean you can't use a low Watt controller if you need a feature, but you may need to buy more than one for your build. The most robust, like the AC Aquaero 5 XT, have built-in overload protection.

An EOCF member searched for controllers that allow bright display lights to be turned off. Some state lights on Lamptron models are less intense than Sunbeam. Some Lamptron models have a jumper to dim LED's and this review above on the Fractal Design Adjust 108 describes a jumper on the PCB to turn off LED's. An EOCF member writes if only one side of an LED lead is cut, it will permanently disable the LED but the fan controller will still work. As always, mod at one's own risk. I have no link to a picture tutorial describing lead cutting. Note that it likely voids warranty and can make a controller less functional if it uses LED brightness to indicate fan rpm.

Suggestion: Aqua Computer Aquaero or high-Watt Lamptron models are best for stacking multiple fans on one channel as they are harder to overload than Budget models with similar specs. In summer 2013, Lamptron introduced the CW611 Water Cooling fan controller to control pumps and fans. In the table below, the Sleep feature in the Lights column means fans can remain on while the display lights turn off. The AC Computer Aquaero 6 was released in O4 2013 with new software but AC will still make the 5 in an LT.



Table Notes:
US$
: March 2012 from FrozenCPU, Performance PC, Newegg or manufacturer sites.
Ch = Channels; W = Watts
Control: dials have no display; digital describes the display
Lights: Sleep means it turns off the lighted display but not the controller. On-off shuts off the controller. On-dim means there is some dimming available. The Lamptron FC-9 has a jumper to turn off the outer columns of lights.
Size: The Aquaero 5 LT can fit on the blank side of a 5.25" bay with a $12 adapter. The NZXT Sentry LXE uses a PCI card with desktop display. The desktop display uses a rechargeable battery.
Warranty: All models in the table have one year warranties.

Sources:
AC Aquaero 6: Aqua Computer Rep (Shoggy) on XS: http://www.xtremesystems.org/forums/...fans-and-pumps
AC Aquaero 5 XT: www.legitreviews.com/article/1813/1/ and Aqua Computer Rep thread on XS: www.xtremesystems.org/forums/showthread.php?265912-NEW-aquaero-5-series
Lamptron website: http://www.lamptron.com/
Scythe Kaze Server: http://www.scythe-usa.com/product/ac...01_detail.html and http://www.overclockers.com/scythe-k...roller-review/
NZXT Sentry 2: http://www.legitreviews.com/article/1249/1/
NZXT LXE: http://www.overclockers.com/nzxt-sentry-lxe
Aerocool website and linked reviews: http://www.aerocool.us/peripheral/f6xt.htm
Sunbeam PL-RS-3 "Rheosmart 3" with PWM: http://martinsliquidlab.org/2011/04/...m-rheosmart-3/

8. Water-Cooling Kits. Suggestion: The kits in the table below perform well when used as intended. Manufacturers juggle kit contents to compete so understanding basic water-cooling is essential. To only cool a CPU, the XSPC RS240 kit represents great value.
..but so does the new Swiftech H220 (now 220X) hybrid All-in-One/custom kit discussed below the table. Performance ratings of Best, Average, and Least are relative. To turn a kit into a custom loop that includes a VGA card or two, make sure you understand the Beginning Guide. There are now a variety of kits so feel free to discuss options on the forum.



Table Notes: Prices are $US low to high from FrozenCPU January 2013. These include a budget and top model from 3 manufacturers but there are more. Alphacool also introduced kits with video here. Some vendors assemble custom kits too. Other kits are not very good. In 2013, Martinm210 reviewed the Larkooler iSkyWater 300 DIY 240, which is markedly inferior to the other models listed and I cannot recommend.

All-in-One Low-Cost Liquid-Cooling units have their own sticky but the Swiftech H220 deserves special mentions in both the All-in-One Guide and this Kit Section. In January 2013, Swiftech released the H220 pre-filled CPU All-in-One/Kit hybrid. It is $140 for the 120.2 CPU kit with refillable reservoir. The pump (Swiftech is the OEM and states via email the maximum head pressure is similar to Swiftech's 35b) is adequate to allow loop expansion to include a VGA card(s) and radiators by using clamps and 3/8" ID & 5/8" tubing. This post explains and shows SLI and X-fire setups Swiftech assembled at the 2013 CES in Las Vegas, Nevada. Here is a youtube video from the 2013 CES with Gabe Rouchon, CEO of Swiftech. Here is a review of the H220 from OCF. Some test results of the H220 are linked from the All-in-One sticky. Thermally, the H220 cools nearly identically to the RS 240 kit so competes favorably. Linked is an excellent review of the H220 by Martinm210. Later in 2013 the H220 was withdrawn from US sales to avoid a lawsuit form Asetek. In 2014, we have the H220X with testing pending. See the All-in-One sticky (or go directly to the Swiftech website) for more detail.

9. Fittings. Check out this excellent video introduction to fittings. The bottom of a fitting is nearly always size G 1/4 inch British Standard Pipe Parallel (BSPP). Parallel threads depend on the O-ring to seal, making it different than a common tapered plumbing thread or British Standard Pipe Tapered (BSPT). The part that connects to tubing is the barb or compression fitting but also includes specialized fittings such as angles, cubes, fill-ports, and T's. I use a mix of barbs and compressions and each has advantages and disadvantages. 7/16" ID tubing clamped onto a 1/2" Danger Den Fatboy barb is very secure and barbs are less expensive than compressions. Some prefer the look of compression fittings but compressions need an exact tubing size fit for both ID & OD. There are even pros and cons to different brands of compressions. XSPC and Koolance have a larger lip on the barb than Bitspower or Feser compressions. This larger lip grabs better to increase security - particularly on large, thick wall tubing. This larger lip may not be necessary and may decrease convenience during tubing change. BP compressions are large while XSPC are smaller. This can be important on a CPU block where two fittings are close. Swiftech Lok-Seal compressions arrived in 2012. Monsoon compressions are popular, add color, and are reviewed by Martinm210 here. Monsoon's new rotaries, angled, and light port LED fittings are reviewed by ronsanut in this video.

Skinnee Labs tested Quick-Disconnect or QDC Type Fittings. They are convenient but have been large and expensive. Until recently, Koolance dominated. In 2013, Swiftech and Bitspower began making lower profile QDC's with G 1/4" ports to allow choice of fittings and Koolance released new QDC. Recently, stren reviewed and tested the new QDC.

10. Water-Tubing. An internal diameter or ID of 3/8 inch to 1/2" gives best flow. A 4 foot length of 1/2" ID adds only 0.16 PSI at 1.5 gpm so offers low restriction (also see this thread). Outer diameter or OD won't affect flow but affects wall performance such as kinking or bend radius. Tubing performance includes durometer rating as explained by Virtual Rain. Extra-stiff tubing has been reported to push against fittings loosening VGA cards in PCI-E slots. As you read in Virtual Rain's review, MasterKleer wall is stiffer than Tygon 3603. For me, 7/16" ID & 5/8" OD bends like Tygon 3603 7/16" ID & 11/16" OD. Primochill 7/16" ID & 5/8" OD also bends well and resists kinking. In the table below, Green color-code is most commonly used. Masterkleer works well for me but Black color-code is too stiff for some. A hair dryer will temporarily soften tubing. Some shape permanent bends alternating hot water and ice water.



Helpful Link: http://www.mcmaster.com/#about-plast...tubing/=g8fjze page 118...

Clouding of clear water tubing strongly suggests plasticizer leaching. If ignored, it can lead to plasticizer staining of acrylic reservoirs. In my experience, clear Masterkleer on barbs is inexpensive and does not cloud or shed plasticizer as fast as Tygon 3603, which can leach and migrate plasticizer as early as 9 to 11 months. Clear Masterkleer leached with migration after about 16 months. Recently, Masterkleer began selling colored tubing. Duralene is another inexpensive clear tubing and Martinm210 likes it. Others prefer Primochill Primoflex Pro LRT or Feser colored tubing. A thread with many photos reported plasticizer leaching in Primochill Primoflex Pro LRT, but my experience was fine with Primochill. There is a new Primochill called Advanced, which is supposed to be less likely to cloud/leach because it removes the phthalate plasticizer DEHP. Some water-coolers use Tygon R3400/Black Tygon or plasticizer-free Tygon 2001 or B-44-3 to lessen plasticizer leaching. If using flexible tubing with plasticizer, it is uncertain how to avoid leaching or whether leaching is bad luck from a bad batch of tubing. Mayhem's Dye manufacturer suggests that heat or perhaps Copper Sulfate biocide may accelerate plasticizer release. Perhaps that extra bit of delta T that keeps the water cooler will prevent clouding? If you see cloudy tubing, I suggest you drain coolant and replace the tubing before leaching occurs. I do not recommend dyes but if you use dyes, make sure your tubing is compatible. If you use a liquid biocide, follow directions.

Clamps: Water-coolers who use 1/2 inch barbs sometimes use 7/16 inch ID tubing for a tight fit. The downside is it often requires putting tubing ends in hot water to fit and may require cutting to remove. Some water-coolers like the look so much they avoid clamps. But in my opinion, barbs require clamps to lessen risk of leaks. Some use screw-clamps, others like the plastic Herbie Clips, some like spring clamps, and others use zip-ties. All prevent tubing coming off a barb if used properly. Those with 1/2 inch barbs may use 1/2 inch ID tubing of course. Those choosing 3/8 inch barbs will want 3/8 inch ID tubing. Those who use compression fittings must use the exact fit of the tubing - both ID and OD - to get best, leak-free fit.

Durability: After tubing is crushed by a clamp, the crushed part should not be re-used but can be trimmed off. I discard all tubing after a year of cooling duty no matter how it looks since I can't see microscopic stresses in what is basically flexible PVC pipe. It's your call. The system owner decides cost versus benefit as to whether to put electronics at risk to salvage tubing.

11. Coolant. Suggestion: Read your block manufacturer's warranty. E-mail a manufacturer to get them on record if unsure.
For example, any Koolance warranty, is void if using coolant other than Koolance brand. Moreover, Koolance never covers corrosion. EK discusses coolant and their Nickel blocks on their site too. Manufacturer's proprietary coolant usually contains a biocide and anti-corrosives. If you use it, you can skip this section and go to 12. But if you seek information about less expensive alternatives to proprietary coolant, then keep reading.

* Performance: Skinnee Labs said about their test:
We have the classic distilled water and a simple biocide, vibrant colors, earth friendly, anti-corrosive, dye ready and Nano-particle coolants represented. The best performer is Water. The worst performer was Fluid XP Nanofluid.
This testing should end claims for performance superiority of proprietary coolant over water. Here is also a nice summary by Martinm210.

* Biocide: If using water to cool, algae can foul an entire loop in weeks. To prevent this, most use a biocide such as PT Nuke or silver. You only need one biocide, which works well when used as directed. Using several together inside the loop is untested and could, in theory, produce worse results than using one correctly. Petra's Tech or PT Nuke PHN (benzalkonium) - the clear solution - is for those who need a pH Neutral biocide (PHN) such as those with a distilled water-only loop. PT Nuke Cu (copper sulfate) - the blue solution - is acidic and more likely to be an effective biocide even with additives such as anti-freeze or dyes. Here is Petra's explanation.

Silver is effective at treating and preventing infection. One can purchase a silver kill coil or make one after obtaining a silver strip, possibly from a jeweler.
One regular-size silver coil works well for a medium-size loop. Place silver coils in tubing immediately before the CPU or GPU block or a 90° angled fitting and do not allow the silver coil to migrate to the pump. Or, use .999 silver plated fittings from Bitspower. One Bitspower .999 silver plated tubing fitting has a greater curved surface area than the surface area of one I & H silver coil.

Math Alert paragraph: The I & H silver coil is 75 mm long x 3.25 mm wide x .5 mm thick. The total surface area = [length x width x 2] for front and back + [(2 x length) + (2 x width)] for a perimeter that is only .5 mm thick. I got 488 mm² + 78 mm² = 566 mm² and use as a minimum size that we know works. My Bitspower 1/2" silver barbs are 24 mm long with an ID of 10 mm. This gives a curved surface area or internal wet area of [length x diameter x pi] or 754
mm². My 3/8" barbs are 24 mm long with an ID of 8 mm at the tubing end and 10 mm at the G 1/4" end. Even using 8 mm the entire length gives us 603 mm² internal wet area or slightly more than one silver coil. Since compression fittings are 28 mm long in the schematic at FrozenCPU.com, those should work as well as barbs.

The math suggests one .999 silver fitting per medium size loop should work as well as a silver coil and it has for me for over 2 years. One special note is a surface area of one G 1/4" stop fitting is only around 80 mm² (10 mm ID for the G 1/4" size). One may need six or seven stop-type fittings to equal a silver coil but they can be put into extra ports in radiators or reservoirs. Some place a .999 silver ingot into the reservoir but Sterling is only .925 silver may not work as well. In 2012 Monsoon released a .999 silver plug they recommend be changed or added to every 6 to 12 months. Unlike a coil, fittings and plugs cannot migrate.


You do not need more than one biocide at a time. Some post that Copper ions from a Copper block also provide adequate antimicrobial control even without PT Nuke or silver but I have not tried Copper alone. If you use silver fittings or coil, inspect and clean the surface at tear-down. Replace fittings if chipped or worn. No matter the choice, use distilled or the cleanest water available and carefully rinse gear before installing to lessen risk of contamination.

Corrosion is the gradual destruction of material, usually metal, by chemical reaction with its environment. The short story for water-coolers is avoid mixing metals by keeping wetted Aluminum out of the loop. Also avoid Nickel blocks since the thin, cheap Nickel-plating used in water-cooling fails. If you use Copper blocks then most feel the only necessary water additive is a biocide. Some manufacturers still make radiators with Aluminum water-tubes but Aluminum fins don't carry coolant so are fine. A recurring theme in this section is read your warranty carefully before you buy, especially if you mix metals or use Nickel blocks.

The longer story about corrosion begins with understanding
metals compatibility or similarity. This chemistry textbook section on electrochemistry helped me greatly. A glossary of corrosion can help. Electrochemical similarity or dissimilarity of materials are usually described by either a Galvanic compatibility table that lists metal alloys, is tested in seawater, and is often from an engineering source; or described by a Standard Electrode Potential (SEP) table that lists water and elements but not metal alloys, is tested using a standard reference electrode like the Standard Hydrogen Electrode (SHE), and is often from a chemistry source. A further refinement at the molecular level is Electronegativity or the ability to attract electrons.

Galvanic corrosion explained here and here occurs when dissimilar metals are in electrical contact in the presence of an electrolyte to create a Galvanic cell. One chemical engineer states on PowerPoint slide 52/54 that Galvanic corrosion can occur without electrical contact. One poster stated the PC case can be an electrical contact to allow a Galvanic cell. He says electrically isolating the Aluminum radiator of his All-in-One prevented corrosion after he modded it to run on water.

In a watercooling loop, electron transfer via oxidation-reduction reactions can also account for corrosion of dissimilar metals without creating a Galvanic cell. During oxidation, electrons leave the more active metal (oxidation is loss of electrons). Electrons migrate toward the less active - or more Noble metal - and also to water. The metal losing electrons (oxidizing) can lose mass until there is corrosion.

Be aware that regardless of the opinion or experience in the water-cooling community, Koolance does not service a warranty unless Koolance brand coolant is used. That means you have to r
ead any warranty before you buy. Nickel loses electrons to Silver since Nickel is more active than Silver (Silver is more Noble). But whether this is significant in a typical water-cooling loop is a matter for debate. My advice is avoid Nickel blocks and only purchase Copper blocks.

Copper will oxidize by losing electrons to water and form copper oxides including the black oxide seen in my CPU block photo at Section 16 - Maintenance - of the Beginning Guide sticky. Copper oxides coat copper and slow corrosion. In spite of the appearance of the block, the coating seems thin with no thermal performance loss during load testing prior to cleaning. You can leave it alone if you choose.

If you must mix wetted Aluminum and Copper, add a corrosion inhibitor like Feser Base Corrosion Blocker or use proprietary coolant containing inhibitor. If you must buy Nickel, please be aware that some including Martinm210 assume the Nickel-plating will fail so recommend a corrosion blocker when using Nickel to delay potential corrosion. Carefully read the Material Safety Data Sheet or MSDS for Feser Base Corrosion Blocker to learn about toxicity and need for careful handling. Corrosion inhibitors coat parts to lessen electrochemical contact and slow corrosion.
Also consider nylon or rubber washers to electrically isolate an Aluminum water-tube in a radiator from the PC case so there is no Galvanic cell.

Using proprietary coolant raises price considerably, which leads some to use a mixture of water and anti-freeze. In ambient watercooling, this is not because of anti-freezing properties; but because automotive antifreeze contains corrosion inhibitors. For more about using anti-freeze, please see this EOCF coolant discussion.

Here is a discussion about sacrificial anodes in water-cooling similar to using a zinc strip inside a steel water-tank.
Monsoon created a G 1/4 inch Plug with Tin as a sacrificial anode to lessen oxidation but there are no independent tests. Monsoon does not claim the anode is a substitute for corrosion inhibitor additives if one is mixing metals.

Helpful Sources on Corrosion:
http://en.wikipedia.org/wiki/Corrosion
http://www.chem1.com/acad/webtext/elchem/index.html
http://corrosion-doctors.org/Contents.htm
http://www.engineersedge.com/galvanic_capatability.htm
http://en.wikipedia.org/wiki/Standar...28data_page%29
http://www.engineersedge.com/corrosi..._corrosion.htm
http://en.wikipedia.org/wiki/Galvanic_corrosion
http://en.wikipedia.org/wiki/Redox
http://www.wisegeek.com/what-is-oxidation.htm

* Dye:
Using dye in coolant creates risk that it can break down to clog small channels inside blocks or pumps. If you use dye, perform maintenance twice as often as usual. A safer alternative to dye is colored tubing, fittings, and lighting technology. If you insist on dye, please choose the best products, so study the Mayhems Website and Mayhems users club thread.

* Colored Lighting:
Beauty is in the eye of the beholder. Linked is ronsanut's video review of Monsoon's light port fittings with LED's, a video review of UV Tubing, and a fun review from Tom's Hardware on lighting mods.


12. Reservoirs, T-Lines, Plugs, Fill-ports, Fill-lines, and Drains
.
A T-line is an inexpensive way to fill a loop with coolant and bleed air. Advantages are that it uses little room because of no reservoir and it sometimes can function as a drain. The disadvantage is no reservoir, which helps in filling the pump and bleeding air. To inexpensively create a T-line, put a nylon T at the highest point of the loop in place of a reservoir before the pump inlet. Attach a fill-line at the 90° part of the T to fill the loop and cap it with a plug or fill-port. One can use a brass cube and fittings or a valve with stopper but this may be a heavy solution for a simple T-type fill-line. Heavy brass can make water lines sag and kink if they are not supported but zip-ties can provide support. Nylon fittings are lighter and less expensive but more apt to crack and leak than brass.

High
performance reservoir/pump tops combined with one or two pumps work together as a performance package. The disadvantage is that some bring a noisy pump closer to a user's ears. For one or two PMP 450 (Laing D5) is the $130 Koolance 452x2 dual 5.25 inch reservoir, reviewed here. For one or two PMP 400 (Laing DDC) Koolance offers the $120 Koolance 402x2 dual 5.25 inch reservoir, reviewed here; or the compact $99 Koolance 401x2 single 5.25 inch reservoir, reviewed here. Others such as XSPC and AlphaCool are competing. Monsoon has upgraded their reservoir that first appeared under the Danger Den name. In 2013, perhaps the most advanced reservoir is the Swiftech Maelstrom that can house either one or two mcp35x PWM pumps. These reservoirs make it easy to create either single (one) or dual (two) loop systems. For the majority of water-coolers, a single loop is sufficient and less expensive. So why would anyone want two loops or two pumps?

The most simple loop is a T-line > pump > water-block > radiator...and around and around. Two loops or a dual loop system needs 2 fill-lines or reservoirs, 2 pumps, 2 or more blocks, and 2 or more radiators.
The primary reason for 2 separate loops is to force the GPU loop into running a higher water-air delta T than the CPU loop; perhaps 10°C to 20°C for the GPU loop versus 5°C to 10°C for the CPU loop. This can allow an internal 120.3 to cool an SLI or X-fire GPU loop while another internal 120.2 or 120.3 cools the CPU loop both at satisfactory temperatures. This requires careful planning but is getting easier as the industry trends toward cooler CPU processors. For example, second generation core i7 Sandy Bridge 2600k stock load Thermal Design Power or TDP is 95W compared to the first-generation core i7 920 at 130W. In Dual Loop versus Single by Gabriel Rouchon, the chief thermal architect at Swiftech, the data and explanation are detailed. You don't need an expensive reservoir system to run two or dual loops. The XSPC acrylic dual 5.25 inch reservoir for two Laing DDC OEM pumps is budget-friendly.

Even with a single loop, some get a second pump for pump redundancy. If you run two pumps in series - inside one loop - one pump provides back-up in case the other fails. Less expensive single pump dual-bay 5.25 inch bay reservoirs include this XSPC Acrylic or this XSPC Nylon. Each house one Laing D5. In my experience, bay reservoirs can be difficult to fill and bleed but one EOCF member got creative. Bay reservoir pump/top combinations are most attractive when case space is limited.

The best value and most functional reservoir in my opinion is the Swiftech MCRES Micro Rev.2 for $23. It has 4 ports and is a small rectangle with several mounting options to fit into tight places. With 4 ports, the top port can have a fill-line with sealing plug or even fill-port drilled into the top of the case. The bottom port can be a straight path to easily fill the pump. The middle ports include an intake from the loop designed to help bleed air and the bonus 4th port can be for a temperature probe or simply plugged. Reservoirs come in many shapes and sizes to satisfy appearance and performance. Some reservoirs that are made to fit certain model pumps such as Swiftech's reservoir for the 35x pump rev.2 still require a slight mod to prevent vortex. For the 35x or especially 35x2 some would choose a Phobya Balancer or EK-Multioption resx2 Advanced with Enzotech rotary joint thread type G 1/4 inch connector or the Bitspower solution described in this EOCF post. Others reservoirs are made to mix and match.

Some water-coolers avoid drains because it is easy to clamp tubing, cut it, and drain the coolant into a pitcher or basin. Others prefer a more orderly drain to change coolant on a regular basis or attempt to preserve tubing longer by using a drain. To make a drain, go to the lowest point in your loop immediately prior to the pump inlet. Either use a T-line with plug as a drain-line similar to the T-line for filling; or use a cube. I prefer the Bitspower BP-TIIR-C G1/4 inch Shining Silver TII-Rotary Adapter cube or T-Type with the thread on top. A G 1/4 inch connector like this or the Enzotech rotary will attach the cube to a G1/4 inch pump inlet. Fittings on the cube can guide the main line of the loop from reservoir into pump inlet. A fitting is used to attach the drain-line at the 90° angle to the cube. Make the drain-line as long as you wish. Some like it long enough to drain outside the case or even use the T-drain as a fill-line in place of a reservoir. Close the the T-line with a clamped sealing plug or fill-port. This can be attached through the case if desired but down low - almost bottom - to drain. Heavy brass in the bottom of the case is rarely a problem. To lessen corrosion risk, avoid Aluminum fill-ports - even anodized models (black Swiftech or colored Danger Den for example). Get Brass or Delrin fill-ports instead.

13. PC Cases for Water-Cooling. Some water-coolers do not wish to drill, saw, or Dremel a new case. Others consider case-modification part of custom water-cooling. To start, if you have have a 120 mm exhaust fan, you can hang a 120.x radiator externally using an attachment by EK, Swiftech, Koolance, or XSPC. I have a 120.4 XSPC RX on my steel Koolance attachment so it can hold a bit of weight. You can lie a radiator on its side with this attachment from XSPC. Here is a parts depot of mesh, radiator guards, and attachments.
mnpctech.com is a reputable vendor for purchasing modding supplies. A frequent question is what case holds a 120.3 radiator internally with no modding? When placing gear internally, one has to consider the sizes of motherboard, drives, and radiator. DVD drives can go external with USB ports. That frees up 5.25 inch drive bays for a top-mounted radiator that pushes forward or allows a dual bay reservoir/pump or a fan controller or monitoring device. Everything you put inside drive bays frees up space inside for radiators. Visit water-cooling forum picture galleries for ideas.

One inexpensive large tower is a Coolermaster HAF 932. One can get a 120.3 up top and 120.4 hanging off the back exhaust. This photo has a 120.2 behind the lower drive cages where I have my pumps. That means you could put 120.9 worth of rad around a case that is often < $150.
The Corsair 800D looks nice, is < $300, and fits a 120.3 up top without any modding. It also fits a 120.2 in the bottom after cutting out a floor section like Mr. Armageddon from Corsair forums shows us and sysigy from EOCF improved by removing less rivets. So what else is there? To help water-coolers, Juggalo23451 made a good video discussing towers by Cooler Master, Silverstone, and Lian Li. Here is a 2012 discussion from EOCF that mentions Coolermaster Cosmos II & HAF 942 (HAF X), NZXT Switch 810, Fulmo GT, Bitfenix Shinobi XL, and Silverstone TJ07. Another EOCF thread shows two 120.3's in a Xigmatek Elysium. 2013 saw the introduction Phanteks Enthoo Primo for < $250 (video review here) plus larger cases by Corsair and NZXT. Extra-large cases fit several 120.3, 120.4, or 140.3 radiators. These include Mountain Mods Pinnacle, Monticle, or Ascension especially if adding a Pedestal; XSPC H1/H2 Cases; Coolermaster Stacker; Little Devil; and Case Labs. Some house water-cooling gear outside a case using this by Phobya. This radiator stand by XSPC is attractive too. Please check websites and contact manufacturers for details.

14. Data Analysis Explanation.

1. Introduction:
Data was analyzed and colored coding used. Black is usually the mean score so is within the SD. Red is either 1 or 2 SD worse than the mean so I flagged it red. Red should be approached cautiously by beginners. The blue is > 1 SD better than the mean. I called this a trend since it was > 1 SD but < 2 SD from the mean and colored it blue.
1 SD from the mean has a confidence of 68% to 95%. Best performer group is colored green and is > 2 SD better than the mean with a confidence > 95%. Sometimes it is easier to understand to label groups as 1, 2, and 3.

2. CPU Blocks:

Why this conclusion table? Because not everyone reads entire tests. This is particularly true with beginners reading large amounts of information. Skinnee Labs' final conclusion table uses a bar graph weighted 50% GPU core, 25% Voltage Regulator Modules or VRM, and 25% flow. In his testing thread Skinnee informs us:The VRM cooling is secondary to cooling the core yes, but I want the VRM cooling to be as active as the core. Skinnee Labs' final conclusion table mixes 75% thermal with 25% flow performance, which can be confusing if using it to make decisions about restrictive loops.

Results:
Bundymania published the largest group so his results were entered as a population. Thermal results of N = 22, were mean CPU-water delta T = 42.1°C, median 42.5°C; SD 1.7°C. The results were so close that the differences between thermal ranking are only 1 SD from the mean with a confidence of 68% to 95%. The 1 SD difference was recorded in blue. Mean flow was 155 L/h, median 153.5, and SD 8.6 L/h. Flow ranking in red are > 2 SD so reaches > 95% confidence. Skinnee Labs' and Martin's Lab data were reviewed including roundup #3 which uses pressure drop instead of flow rate to measure restriction. The Apogee XT rev.2 is 30% less restrictive than rev.1 but pressure drop is still > 1 SD from the mean.

3. VGA Blocks Test Results:


Thermal column statistical analysis explanation: On page 5 of the link to Skinnee Labs' testing, the raw thermal data for GPU stock is: 44.86°C, 45.96, 45.99, 46.25, 46.31, 46.42. 47.02, 47.91, 47.99, 52.66.

N or Population Number = 10, Mean Temp = 47.14°C, Median Temp = 46.34°C, Standard Deviation or SD = 2.15°C. Population SD = 2.04°C. We can use population SD when analyzing the entire population of data such as the entire group of test results.

The Mean (average value or arithmetic mean) and Median (another average value where half the data is above and half below) are close together in a Normal or bell-shaped curve. One SD contains > 68% and 2 SD > 95% (95.3%) of the data surrounding the average or mean. 95% confidence tells us that retesting is 95% likely to show a similar relationship and 5% unlikely. SD describes whether data are similar or not similar.



Continuing with Skinnee Labs' test results:
In the first set of results, GPU stock numbers; 2 SD is +/- 4.08°C also written plus or minus 4.08 degrees Celsius with confidence greater than 95%. With GPU stock numbers, the only outlier or data point > 2 SD is Bitspower at 52.66°C. This tells us the BP has higher stock GPU thermals with > 95% confidence compared to the other models in the population tested. Or there is a greater than 95% chance that the BP will be 2 SD higher than the other models if retested and less than 5% chance that it will not.

GPU Overclock: 49.16°C, 49.57, 49.97, 50.04, 50.05, 50.23, 51.00, 52.17, 52.36, 58.02. N = 10, Mean 51.26°C, Median 50.14°C, SD 2.6°C and Population SD 2.47°C; 2x Pop SD = +/- 4.94°C > 95% confidence. Bitspower at 58.02°C is > 2 SD from the mean of 51.26°C, so BP over-clocked unsurprisingly runs hotter like stock loads.

Voltage Regulator Modules are modules that regulate voltage and get hot. Some consider cooling VGA VRM important to avoid errors. Cooling VRM and RAM persuade many water-coolers to buy a full-cover instead of GPU-only VGA block.

Voltage Regulator Module or VRM stock = 46.29°C, 48.16, 51.61., 52.39, 53.40, 54.02, 54.83, 55.17, 57.24, 66.44°C. N = 10, Mean 53.96°C, Median 53.71°C, SD 5.47°C, Pop SD 5.19°C; 2x Pop SD = +/- 10.38°C > 95% confidence. EVGA Hydro Copper gtx480 at 66.44°C is > 2 SD from the mean of 53.96°C. It is an outlier for VRM temps at stock, but the rest are within 2 SD of the mean.
VRM OC = 55.24°C, 57.86, 63.52, 64.28, 67.12, 68.12, 68.22, 72.95, 74.46, 89.48. N = 10, Mean 68.18°C, Median 67.72°C, SD 9.63°C, Pop SD 9.1°C; 2x Pop SD = +/- 19.26°C > 95% confidence. EVGA Hydro Copper at 89.48°C is > 2 SD from the mean of 68.18°C so over-clocked it unsurprisingly remains hotter just like stock loads.

Flow column statistical analysis explanation:
The gallon per minute or gpm results are 8 numbers since TIM doesn't affect flow. They are 1.26 gpm, 1.29, 1.62, 1.63, 1.68, 1.71, 1.74, 2.05. N = 8, Mean 1.62 gpm, Median 1.66 gpm, SD .253 gpm, Pop SD .237 gpm, 2x Pop SD = +/- .474 gpm from the mean. Pressure Drop in Pounds per Square Inch or PSI at 1.5 gpm yielded 8 numbers: 5.49 PSI, 2.39, 0.60, 1.80, 5.57, 2.20, 2.46, 2.03. N = 8, Mean 2.82 PSI, Median 2.23 PSI, SD 1.77 PSI and Pop SD 1.66 PSI. 2x Pop SD = +/- 3.32 PSI from the mean.

Neither gpm nor PSI drop yielded differences at > 95% confidence or 2 SD but they do show a trend at 1 SD. Skinnee Labs captures this trend well, showing three distinct groups of flow curves in their flow section results but unfortunately these results are merged with thermal performance in their final conclusion table. The highest and lowest values are > 1 SD from the mean. That relation is not as a high of statistical power as > 2 SD so the confidence is between 68% and 95% instead of > 95%. Rather than calling it statistically significant, one can call it a trend. The Danger Den trends toward least restriction/best flow being 1 SD better than anything else and is colored green. Most are average flow and are blue. There is a trend toward most restriction/worst flow in the Aqua Computer and EVGA/Swiftech gtx480 that may be important to a buyer so is colored red.

Even though differences in raw data between a PSI drop of 5.49 versus 0.60 doesn't show 2 SD on a test bench, > 1 SD from the mean can be important in a real loop particularly with multiple VGA cards or a restrictive loop.

Helpful Links:
1. Standard (Statistical) Error
2. Elementary Concepts in Statistics
3. Online Statistics Calculator
4. Normal Distribution
5. Margin of Error
6. Standard Deviation
7. Arithmetic Mean

Please Note: If you have questions or wish to suggest edits or have new information, please start a thread so we all learn. If you find common typos or grammar errors, please send a PM to me to fix it. Thank you.

Last edited by musicfan : 08-13-2014 at 06:09 PM. Reason: grammar, links, edits, additions, new testing, updates
United States  Offline
    Register to Reply to This Post
Old 05-14-2011, 02:27 PM   #2
Doglips
Mad Warranty Voider
Doglips's Avatar
Senior Member
 
Posts: 2,032
Last Seen: 08-10-2014
From: Oregon
iTrader: 12 / 100%
Great Info Musicfan!!!
United States  Offline
    Register to Reply to This Post
Old 05-14-2011, 04:02 PM   #3
gottistar
Overclocker
gottistar's Avatar
Senior Member
 
Posts: 260
Last Seen: 06-24-2014
From: land of oz
iTrader: 0 / 0%
this has got to be "post of the year"
Australia  Offline
    Register to Reply to This Post
Old 05-14-2011, 09:32 PM   #4
Llunker
Absolute Zero !!
Llunker's Avatar
Regular Member
 
Posts: 368
Last Seen: 03-23-2012
Age: 49
From: Abilene, Tx
iTrader: 0 / 0%
WOW --so thats what you been working on --Excellent Stuff MF. Very Nice and TY
United States  Offline
    Register to Reply to This Post
Old 05-15-2011, 10:43 AM   #5
apw63
PWM heart rate
apw63's Avatar
Senior Member
 
Posts: 450
Last Seen: 10-26-2014
Age: 51
From: Washington
iTrader: 0 / 0%
Way to go MF, as always top notch presentation of data. This should be extremely help full to any one wanting to put together a WC system. I vote for sticky
United States  Offline
    Register to Reply to This Post
Old 05-17-2011, 05:41 PM   #6
Mike597
Learning To Overclock
Regular Member
 
Posts: 51
Last Seen: 02-22-2013
From: Michigan
iTrader: 1 / 100%
This is a fantastic resource for anyone working with a w/c system. That is a wealth of information.

This should be made into a sticky!
United States  Offline
    Register to Reply to This Post
Old 05-17-2011, 05:52 PM   #7
musicfan
ambient water-cooling
musicfan's Avatar
Senior Member
 
Posts: 2,745
Last Seen: Today
Age: 59
From: Oregon, USA
iTrader: 0 / 0%
Some of you let me know what to add. Thanks to you, I added Fans, Fan controllers, 140.3 rads, and everything below kits.

It's harder for me as the components get harder to test. It's easier to be objective with math.

I will add a bit on Reservoirs, T-Lines, and Drains for completeness. Although it's not number-crunching performance, it gets asked a lot and may save us time.

I am procrastinating on cases because it seems so daunting...especially for a guy who never spent > $200 on one.

That's the forecast over the next week or month or more. Please continue to tell me what needs fixed or would help the community. The goal if it ever happens is for someone to read ricecrispi's thread to 2009, then be able to jump over to this for 2010-2011.

In the last week, Skinnee Labs assembled a CPU shootout #3 and we had comparative flow data. The AC Kuplex Cryos now fell 1 SD outside the mean so I changed the table and Cryos went from green to blue flow. Then yesterday, Martinm retested the Apogee XT rev.2 and the restriction was down 30% but not as good as some had hoped. The flow numbers changed again so I changed the table another time as the AC Kuplex Cryos fell back within the mean (became green again), the Apogee XT rev.2, finally tested independently, fell > 1 SD (became blue) from the mean, and the rev.1 now sold only as the XTL remains 2 SD from the mean (so is still red). So if I mess something up at Image Shack getting my tables correct, please let me know. The testers are keeping me busy changing them already and that is good.

I am using the colors like a stoplight with Green for a buy and Red to stop and reconsider. Yellow would be a caution but is hard to read so I picked something easy to read, blue. Even wanted to go Gold, Silver, Bronze to keep it simple but can't see those colors at all.

Thanks again for your help.

Last edited by musicfan : 05-18-2011 at 12:02 PM.
United States  Offline
    Register to Reply to This Post
Old 05-19-2011, 04:33 AM   #8
mimart7
Je plie, donc je suis
mimart7's Avatar
Senior Member
 
Posts: 1,653
Last Seen: Today
Age: 49
From: NYC
iTrader: 1 / 100%
I find this guide to be excellent. If I put my cpu under water, I will be consulting this guide before asking an questions. This should be stickied.
United States  mimart7 Folds For EOC!  Online
    Register to Reply to This Post
Old 05-19-2011, 01:13 PM   #9
Llunker
Absolute Zero !!
Llunker's Avatar
Regular Member
 
Posts: 368
Last Seen: 03-23-2012
Age: 49
From: Abilene, Tx
iTrader: 0 / 0%
It might be noted --that the Koolance 402x2 dual 5.25 inch reservoir- Also accepts 1 or 2 Swiftech MCP35X Pumps as well since I know we spent alot of time on choosing this pump in another thread.

Excellent Job with the Resources MF--just plain AWESOME !!!

Update for MF, The
Danger Den Monsoon link does not work? Didn't know if you meant this Res or not? Thanks MF !!


Additional Comment:

Yo hey Mods--could we please Stick this at the TOP !!! it needs to go there if you guys don't mind.

Last edited by Llunker : 05-19-2011 at 01:13 PM. Reason: Automerged Doublepost
United States  Offline
    Register to Reply to This Post
Old 05-19-2011, 03:57 PM   #10
musicfan
ambient water-cooling
musicfan's Avatar
Senior Member
 
Posts: 2,745
Last Seen: Today
Age: 59
From: Oregon, USA
iTrader: 0 / 0%
Thanks mimart7. I appreciate your post. I think we have most of the bugs out so will ask our forum moderator for guidance soon.

Llunker, thank you so much for finding that broken link and making the guide more clear. I think we are back in business. That is exactly the type of editing this guide needs.

I need to add something on motherboard blocks - kind of forgot about them since I don't use them.

Please keep finding things that don't make sense or broken links or new tests or something I forgot that belongs in there. Thanks.
United States  Offline
    Register to Reply to This Post
Old 05-19-2011, 04:41 PM   #11
nova777
Liquid Chef
nova777's Avatar
Senior Member
 
Posts: 197
Last Seen: 10-02-2014
iTrader: 0 / 0%
Awesome Post

Nice Work MF
Canada  Offline
    Register to Reply to This Post
Old 05-22-2011, 03:44 AM   #12
Llunker
Absolute Zero !!
Llunker's Avatar
Regular Member
 
Posts: 368
Last Seen: 03-23-2012
Age: 49
From: Abilene, Tx
iTrader: 0 / 0%
Now Lets Stick this at the Top please -Mods...TY
United States  Offline
    Register to Reply to This Post
Old 05-24-2011, 05:01 PM   #13
musicfan
ambient water-cooling
musicfan's Avatar
Senior Member
 
Posts: 2,745
Last Seen: Today
Age: 59
From: Oregon, USA
iTrader: 0 / 0%
Whew! Spent the last few days re-working the radiator section. I was not excited about the statistical breakdown of the tiny fractions in the C/W part. Fortunately the magic of calculators helped. I rewrote the table several times so the information is easier to extract, the performance colors have actual statistical meaning (don't buy red unless you know what you are doing), there is a results section and short discussion. The last is for those who like text added to numbers.

I cannot think of anything else to add without starting a beginner's guide (assembly, layout, leak testing, maintenance, etc.). But that is different guide for a different day.

This is about as much "performance per dollar" information as we can jam in a single-post Buyer's Guide...isn't it?

May 26, 2011

Whew again! The information was there - the difficulty was finding it.

It is a writing challenge for a geek like me to write something a beginner can use, but remains useful to a pro who only needs one new part. That required major editing. What we have currently is aggressive color coding for an clearer road map and the suggestion for beginners and the busy. Please let me know what you think.

We want to stay objective and let the data speak (the scientist part), but not leave out our lifeblood, the beginner who wants to learn to water-cool. They notice we are having way too much fun. So who can blame anyone from wanting to water-cool?

There was data relocated to the bottom of the page but none was removed. Data seekers have it to understand ratings and suggestions.


It looks like this is a Read First Sticky at the moment so let's work together to keep it there. Special thanks to all EOCF members who posted immediately above and sent PM for ideas; to our moderator WiCKeD for important insights; and to those who teach me daily.

Last edited by musicfan : 05-26-2011 at 05:52 PM. Reason: clarity
United States  Offline
    Register to Reply to This Post
Old 05-29-2011, 07:36 PM   #14
Doglips
Mad Warranty Voider
Doglips's Avatar
Senior Member
 
Posts: 2,032
Last Seen: 08-10-2014
From: Oregon
iTrader: 12 / 100%
Congrates on the Sticky!!!! Well deserved!!
United States  Offline
    Register to Reply to This Post
Old 05-31-2011, 05:59 PM   #15
parth
Learning To Overclock
Senior Member
 
Posts: 123
Last Seen: 06-23-2011
Age: 25
iTrader: 0 / 0%
Excellent guide! I haven't worried much about watercooling since 2007 (first post here in the last 2 years and second in the last 4) when I built my first loop, but now it's time for an upgrade so this information is really useful!
United States  Offline
    Register to Reply to This Post
Old 06-03-2011, 03:39 PM   #16
musicfan
ambient water-cooling
musicfan's Avatar
Senior Member
 
Posts: 2,745
Last Seen: Today
Age: 59
From: Oregon, USA
iTrader: 0 / 0%
Thanks for the kind words. This Guide was created to save busy EOCF members some time. Please ask on the forum for any specifics.

I will try to do at least a monthly roundup to keep the front page or Post #1 updated. At this time I am watching Skinnee Labs' CPU roundup #3 carefully since they are testing new blocks in sequence so new results will trickle in.

I was a bit disappointed that the Aqua Computer Cuplex Kryos HF did not exhibit "high flow" but now I am not sure that disappointment is warranted. The AC Cuplex Kryos Pro, XT, and HF all have the same internals only different covers - brass, Nickel plate, or copper. So the fact that the HF performed like the XT with good thermal but a little restriction should be expected. Doh!!!

Otherwise Skinnee Labs confirmed what Martinm found on the Apogee XT. The thermal performance remains with the rev.2 improvement; but the flow improvement in independent testing is not quite as high as it appeared in the proprietary testing. Always seems to be that way doesn't it? And that is why we need more independent testing from more independent testers to produce less performance bias.
United States  Offline
    Register to Reply to This Post
Old 06-04-2011, 02:48 PM   #17
Llunker
Absolute Zero !!
Llunker's Avatar
Regular Member
 
Posts: 368
Last Seen: 03-23-2012
Age: 49
From: Abilene, Tx
iTrader: 0 / 0%
Excellent as always MF !! Good job
United States  Offline
    Register to Reply to This Post
Old 06-08-2011, 10:58 AM   #18
Mike597
Learning To Overclock
Regular Member
 
Posts: 51
Last Seen: 02-22-2013
From: Michigan
iTrader: 1 / 100%
Sweet! An official sticky for MF!
United States  Offline
    Register to Reply to This Post
Old 06-28-2011, 09:28 AM   #19
StephenP2003
It's creamed corn
StephenP2003's Avatar
Senior Member
 
Posts: 440
Last Seen: 10-27-2014
Age: 29
From: Baton Rouge, LA
iTrader: 4 / 100%
I haven't built a loop since 2004, when I was running a socket 478 P4 and an AGP Radeon 9800XT, and I must say, between this guide and your beginner guide, I feel like I'm up to speed with current WC. I've been itching to do another build from scratch.

Question though: If I were to use the Swiftech MCR-QP radiator, given its small footprint, would I need to use two for a loop with CPU and VGA? Back in the day, I only used a single-fan radiator with a loop that went straight from the CPU block to GPU block, and the GPU temps suffered a little, but not so much that I couldn't do what I needed to do (OC and flash 9800Pro to XT). But now I have a 6950 that runs pretty hot.
United States  Offline
    Register to Reply to This Post
Old 07-14-2011, 11:17 AM   #20
musicfan
ambient water-cooling
musicfan's Avatar
Senior Member
 
Posts: 2,745
Last Seen: Today
Age: 59
From: Oregon, USA
iTrader: 0 / 0%
Quote:
Originally Posted by StephenP2003 View Post
I haven't built a loop since 2004, when I was running a socket 478 P4 and an AGP Radeon 9800XT, and I must say, between this guide and your beginner guide, I feel like I'm up to speed with current WC. I've been itching to do another build from scratch.

Question though: If I were to use the Swiftech MCR-QP radiator, given its small footprint, would I need to use two for a loop with CPU and VGA? Back in the day, I only used a single-fan radiator with a loop that went straight from the CPU block to GPU block, and the GPU temps suffered a little, but not so much that I couldn't do what I needed to do (OC and flash 9800Pro to XT). But now I have a 6950 that runs pretty hot.
Glad the guides are doing what they are supposed to. Yes, let's plan another build. It's fun.

The 6950 uses about 160W under load. If you get a CPU and run it at stock, you should have no trouble cooling with one 120.3 MCR-QP since that will only be about 100W per 120.1 unit or less.

If you over-clock or X-fire your GPU, then you might need more cooling. If you want, please start your own fresh thread with your parts and cooling gear listed. More people will see a new thread. There are several good water-coolers on this forum. We can discuss many options for you and answer what we can. Thank you for your interest.

14 July 2011

After PM exchanges with Martinm210, I edited under 7. Fans and Controllers Fans...

Push/pull advocates typically recommend identical fans but there are no tests proving that identical models are necessary for best performance or least noise.

23 July 2011

As per suggestion of Doglips, Danger Den tubing was added to 10. Water Tubing Performance. I don't have experience with this but it looks nice. Also added links to MasterKleer and Feser colored and Duralene clear. Anyone use any of these? Easy bending? Durable? Do you like it? Thanks.

Last edited by musicfan : 07-23-2011 at 02:33 PM. Reason: New information
United States  Offline
    Register to Reply to This Post
Sponsored Links:
Please Register to Post a Reply



Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

vB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Forum Jump

All times are GMT -6. The time now is 02:31 PM.

Copyright ©2000 - 2011, Jelsoft Enterprises Ltd.
Powered by vBulletin
Copyright ©2000 - 2011, EXTREME Overclocking