Modern CPU Comparison Database

[Cecil]

A while back I started this, and I think its about time to put it up. Its a work in progress, and will be updated over time.

The Goal,

Compare fairly and evenly all major CPU architectures in recent years. Going to run each one in three stages.

First, pure stock. This will be F5 in BIOS for optimised defaults, and nothing else to show pure out of box experiance.

Second, clock for clock. Will have ram speed even across each platform (or as even as possible). For Intel, QPI/FSB, etc will be even and close to stock as possible, and the same with AMD for NB/HT speeds (2k). CPU speed chosen will be 3Ghz. Most modern CPUs are past that, and a lot of older ones wont do much more then that. If a CPU cant do 3Ghz, then the clock for clock test will be left out for it.

Third, fully optimised system. Will be ran with CPU at what would be considered the daily stable average with air cooling, and memory as fast as possible on the system.

The first test will be to show out of box performance for a non-OCer/tweaker. The second is to show the IPC increase (or decrease) from each generation. And the third is to be able to compare what you can expect from a fully overclocked and tweaked system.


The Why,

I usually get all the new stuff shortly after it comes out. So this will be an easy (and fair anti-biased way) to compare how your current system stands up to newer stuff. Also, its an all in one place to check performance of past CPUs to decide what to get if you like to upgrade to things when they are a gen or more old.



The Test Bed,

The parts that will not change (or at least wont do so unless forced to, ie if something dies and cant find an exact replacement)


Video Card
Nvidia GTX-470, refrence card with stock clocks. Still a good card by todays standards.

Monitor
Yamakasi Catleap Q270 2560x1440 res monitor. Games will be tested at 1280x720, 1920x1080, and 2560x1440.

Cooling
Cooler Master Hyper 212 EVO, it shows pretty high end air cooling performance at a low end price. Also fits just about every platform Ill be testing.

Ram
DDR3 - 2x4GB Crucial Ballistix Tactical Tracer 1866Mhz 9-9-9 1.5V, flexable ram that can be ran with speeds as high as 2200Mhz, and low timings across all speeds.
DDR2 - Gskill PI 1100Mhz 5-5-5 1.9V, again flexable ram that can do high speeds and tight timings.

HDD
WD 160GB VelociRaptor

PSU
Corsair TX-750-V2

Case
Custom built test bench

OS
Windows 7 Home Premium X64 full install, fully updated, with no tweaks


Due to my constant hardware changes, some of the motherboards I used in my early tests, I no longer have. So now, for the most part, all tests will be done on one mobo per socket. Boards I have for the tests are,

AM2/+ Gigabyte 790X-UD4P
AM3/+ ASUS Crosshair V Formula

775 DFI LT P35

1366 ASUS Rampage III Extreme

1155 Gigabyte Z77X-UD5H


The Tests

CPU Tests

Super Pi 1m and 32m
Wprime 32m all cores and single core, and 1024m all cores
Cinebench 11.5
Fritz Chess all cores and single core
X264 bench
UC Bench
AIDA 64 full suite
LinX Gflops, power (read with hwmonitor), and temps
Multi Prime 32m and 128m
Nuclearus ALU, FPU, MT, and Total scores


Memory Tests

MaxxMem
AIDA 64 memory and cache benchmark

Full System Tests
PCMark 7

Game Tests

3DMark Vantage
Left 4 Dead 2
Far Cry 2
Battlefield 3




Some notes,

When I started this, I didnt plan to do so much . So at first, the results are a "rough draft". I was testing only modern stuff, and had clock for clock at 3.8Ghz, and then sold the video card and HDD I was using at that time (6870, WD Black 640), so I will put those results up for now. But they will not be "fully" comparable to the new results I start to put up in this revised "version 2". Once I can get a hold of those CPUs again, I will re-bench them. In the mean time, I will run a fourth set of tests, at 3.8, in as many of the benches as I can (that dont use the HDD or vid card).
Also, some CPUs dont work with every test, so some entries may be "NA".


The Results


CPU Tests



Memory Test



System Tests



Game Tests

[ashbornangel]

love to see this

[Cecil]

Ill put up the results I do have some time tomorrow. Have done 2500K, 3570K, FX-8120, 1090T, and Q9650 so far. Have a i7 920, 3770K, PHI-9850, PHII-X4 + PHII-X2 here to do.

[Urbanfox]

The rollover on what is tested isn't strong, but this isn't bad: http://www.anandtech.com/bench/CPU/2

[leekirlew]

massive commitment bud good luck =)

[Cecil]

Made a last minute trip to ny, can't add anything until I get back Mon/Tuesday.

[ShrimpBrime]

Which Cpu's was that again?

NVM I see the list.

The 2.5ghz 9850 is a waste of time and hardly modern. If your going that low in mhz, why not test LLano instead?

[Cecil]

Working on the 920 right now. Once its done, and I get my 3770K set back up, going to do that one, then put the initial results up.

[PhilZJ]

Is that my old Q9650?

Btw, just suggestion, but you already have a massive amount of test but some people here I saw are interested in video conversion, so I was thinking some arbitrary high def test clip run through handbrake.
Unfortunatly, to my knowledge handbrake doesn't use built in video instructions in the sandy bridge, unlike some more professional converters.

All the best with this, I'll be interested in the results.

Oh for sandy and ivy bridge will you run a test using built in graphics for gaming as well?

[Cecil]

Yes that was your Q9650. Dont have it any more though.

Ill see if I can work in transcoding some how. It already takes about 1-3 hours to run through the benches Im doing though at each speed, and testing 3-4 different speeds of each chip.

[PhilZJ]

Ya I completely understand how time consuming this already is so, I think it's great you're going through all the effort and your tests are pretty thorrow already.

All the best!

[Cecil]

Something Ive noticed so far, is the CPU has no effect at all with Battlefield 3, so I think Im going to stop testing that one.

With an i7 920 at 3Ghz, and 3.8Ghz, and a Thuban at 3Ghz the largest difference was at 720P medium settings of 2FPS. Everything else (avg and min @ 1080P high and 1440P Ultra) are all equal or 1FPS different.

Additional Comment:

Until I have some time to pretty it up some more, I added screen shots of the results so far.

Firstly, the 2500K/3570K results are a little off with memory related stuff, because I had used two different boards (Gigabyte Z68 and ASUS Z77). ASUS really loosened a lot of stuff up for more memory speeds, but killed bandwidth doing so. So the 2500K has much better memory results then the 3570K.

Im going to rebench both chips on the Z77X-UD5H that I used for the 3770K, as soon as I can.


Second, the highlighted results are with the new setup with the GTX-470 and VR instead of the WD black and 6870. So you can compare results that are highlighted with each other, but not the ones that arent.

[DarkTower87]

Originally Posted by Cecil

Something Ive noticed so far, is the CPU has no effect at all with Battlefield 3, so I think Im going to stop testing that one.

With an i7 920 at 3Ghz, and 3.8Ghz, and a Thuban at 3Ghz the largest difference was at 720P medium settings of 2FPS. Everything else (avg and min @ 1080P high and 1440P Ultra) are all equal or 1FPS different.

Additional Comment:

Until I have some time to pretty it up some more, I added screen shots of the results so far.

Firstly, the 2500K/3570K results are a little off with memory related stuff, because I had used two different boards (Gigabyte Z68 and ASUS Z77). ASUS really loosened a lot of stuff up for more memory speeds, but killed bandwidth doing so. So the 2500K has much better memory results then the 3570K.

Im going to rebench both chips on the Z77X-UD5H that I used for the 3770K, as soon as I can.


Second, the highlighted results are with the new setup with the GTX-470 and VR instead of the WD black and 6870. So you can compare results that are highlighted with each other, but not the ones that arent.

Looking good. Thanks for all of the hard work!!! Perhaps sticky at some point??? Are you going to be keeping up with this? You change hardware constantly.