Just a little info: Your motherboard DOES effect your gpu overclocking.

[Dil-tech]

I know some already know this, but alas, it's a topic rarely spoken on. Your choice of motherboard can, and likely does, effect your overall overclock. I recently upgraded my system to my current i7 930 on eVGA x58 SLi3. The only changes to the system in the process was the cpu, mobo, and ram. Everything else remained the same: same powersupply, same case, same case cooling, same operating system, even same cable management. How much can it vary? Here's the number change.

Test Setup (what varied):
i7 setup/q6600 set up
i7 930 @ 4.1ghz/Q6600 G0@3.6ghz
Zalman 9900ALED/Scythe ninja infinity
eVGA x58 SLi3/Gigabyte P35 DS3L
12gb Samsung Green low-profile DDR3/4gb AData 800mhz ddr2

The rest of the system is as follows:
Antec 900 case
Antec 850W TruePower Quattro(using non-modular PCI-E cables for video-cards)
Windows 7 Ultimate x64
Twin 7200 rpm seagate baracuda, 500gb along with a 7200 rpm 300gb WD
Galaxy GTX 480 Super OC model (equipped with artic xtreme plus, cooling isn't an issue)
All stability testing done with OCCT's gpu torture testing. 20 minute minimum.

Using the Q6600 set up, I found myself instantly having to up the voltage to even go up 15 mhz. The stock voltage was 1.038, but to hit 800mhz took 1.1. Even at 1.138 I couldn't do 820 mhz, making me believe I was lied to about the card's abilities by the guy I bought it from (who managed 895mhz on the core with the exact same card). The ram wouldn't even go up 10mhz without issue. At stock there was no problems what-so-ever.

Upon switching to the i7 set up, I decided to start clocking it up again. 800mhz came free, without any voltage increases at all. 830mhz came at 1.05 without issue. Haven't pushed further yet as I've been more than content since. The ram will now do 2.1 (4.2 ghz effective) without issue.

Temps weren't an issue on either set up, as my card has never seen a temp higher than 66C even if my livingroom was 82F.

As a side-note, my GTX 460 wouldn't go much above it's stock frequency of 800mhz either. Now I see it was the board and NOT the card.

The moral of the story? Buying a budget board doesn't just hold back your cpu overclock (which for gaming, a few hundred mhz doesn't make a difference), it can drastically hold back your gpu overclocking. A small voltage increase brought me higher than the maximum allowed brought me to with my previous system. Even with a high-end power-supply (the Antec 850 TPQ is one of the highest rating psu's ever) the pci-e slot power structure means more than you would think, your cables can't make up for that, it needs to pull it's fair share from the slot.

p.s. This is why so many cards are clocked far below what they're capable of; they have to factor in boards lacking the power structure to go above 75W to the PCI-E slot.

[Almost Tactf]

Interesting, however I think to really prove this you'd need to use a cheap a** X58 board as well as a top of the line X58 because the results here could have been due to the differences between P35 and X58

[Dil-tech]

Originally Posted by Almost Tactf

Interesting, however I think to really prove this you'd need to use a cheap a** X58 board as well as a top of the line X58 because the results here could have been due to the differences between P35 and X58

Or I could just test it with the guy who bought my q6600 as he now has 2 q6600 set ups, one with my old set up and one with a much nicer mobo. It effected my 460, and he got that from me too, which would be enough to prove my theory. I'll see if he's willing to let me try it out on his set up since he lives 3 miles from me.

[Almost Tactf]

That would be epic, I don't doubt ur theory at all, this would just completely solidify it

[Deluded]

EDIT: I didn't read the post clearly enough.

I wonder if it's due to the PCI-E versions between the two motherboards. Perhaps the 1366 supplied more voltage/could sustain more current through the PCI-E port that allowed the graphic card to overclock better.

[Dil-tech]

I'll see what I can do. I'll also hit up some buddies at a few tech sites to see if any of them are willing to do a much larger scale on gpu clocking since they'd have a far larger line of boards to test.

Originally Posted by Deluded

EDIT: I didn't read the post clearly enough.

I wonder if it's due to the PCI-E versions between the two motherboards. Perhaps the 1366 supplied more voltage/could sustain more current through the PCI-E port that allowed the graphic card to overclock better.

PCI-E standards all the way down the line support 75W through the slot. Generally speaking, upping the voltage can actually fry your card, so that can't be the case. It's the current that's the problem, and that's handled by the voltage regulators. Cheap motherboards can't sustain that higher current compared to higher end motherboards, which will effect the current pullable by the pci-e slot, which is ~1/3rd of its overall power draw.

[Deluded]

No, I understood that. What I meant was that the difference between the PCI-E versions 1-3 might have some variance in power/current. Or it could be due to the PCI-E controller.

s775 had the NB controller on the motherboard, but it is in my understanding that socketr 1366, the CPU has the PCI-E controller on die, and thus the chances for signal degradation is much lower?

(I might be wrong... I have to reread the slides again)

[Dil-tech]

Originally Posted by Deluded

No, I understood that. What I meant was that the difference between the PCI-E versions 1-3 might have some variance in power/current. Or it could be due to the PCI-E controller.

s775 had the NB controller on the motherboard, but it is in my understanding that socketr 1366, the CPU has the PCI-E controller on die, and thus the chances for signal degradation is much lower?

(I might be wrong... I have to reread the slides again)

On-die PCI-E wasn't added until socket 1156. 1366 had it on the north-bridge.

[Deluded]

Well, I see it now.

Could be improvement across tech. 1366 did have better memory performance.

[Dil-tech]

Originally Posted by Deluded

Well, I see it now.

Could be improvement across tech. 1366 did have better memory performance.

1366 has an on-die memory controller, of course it had much better memory performance.

I'll see if I can do the testing on my buddy's two Q6600 boards and see if my findings remain conclusive.

[FX5200]

P35 is PCi-E 1.1 IIRC... with somewhat differently designed power circuitry. I've always noticed, higher end and newer-gen mobo's tend to be more happy with my GPU overclocks.

[Dil-tech]

Originally Posted by FX5200

P35 is PCi-E 1.1 IIRC... with somewhat differently designed power circuitry. I've always noticed, higher end and newer-gen mobo's tend to be more happy with my GPU overclocks.

It was PCI-E 2.1 that changed power structure IIRC, and PCI-E 1.1 boards had a bios update to make the cards backwards compatible. I could be wrong on that though.

Either way, I'll see what I can do about further testing. That's a huge difference in terms of overclock potential though.

Additional Comment:

Well guys, same difference happened for him between the gigabyte board I had and his MSI 650i Sli board. 840 mhz was the max he could squeeze at 1.087 on the gigabyte, the MSI pulled 880 on the same card. Keep in mind that the 6x0 chipsets are actually inferior to their intel competition. That was on a GTX 460, using a 700w coolermaster psu, which is FAR more than enough for a q6600 and gtx 460. OCCT was again used for stability testing (to make 100% sure it was stable).

Looks like the vreg quality effects more than just your cpu overclock after all.

[Maniac]

My old DFI Lanparty UT P35-2TR had the ability to plug one of those floppy power connectors into the mobo to supply more power to the GPU... I never did try it.

[Sheep221]

Originally Posted by Maniac

My old DFI Lanparty UT P35-2TR had the ability to plug one of those floppy power connectors into the mobo to supply more power to the GPU... I never did try it.

There were several boards with this feature, you connected molex to board when running high end card.

The problem by motherboard limiting the OC of card is pretty much obvious, because although external supply to the card is important, the control circuitry is still fed from PCIE slot. Although maybe it's the failure of your particular board. It can varry from piece to piece especially when it comes to overclocking.

[Dil-tech]

Originally Posted by Sheep221

There were several boards with this feature, you connected molex to board when running high end card.

The problem by motherboard limiting the OC of card is pretty much obvious, because although external supply to the card is important, the control circuitry is still fed from PCIE slot. Although maybe it's the failure of your particular board. It can varry from piece to piece especially when it comes to overclocking.

Hence why I'd like to see an actual review site pick up the article as they could do multiple boards with the same chipset (from budget to high-end) and see how much things vary.

[Sheep221]

Originally Posted by Dil-tech

Hence why I'd like to see an actual review site pick up the article as they could do multiple boards with the same chipset (from budget to high-end) and see how much things vary.

I think this is something that extreme - enthusiast boards resolve. Depending on the chipset, but virtually any MB can overclock today, but having overclocking features is not automatically having overclocking capabilities. You can buy OC board for as low as $70 which is not the same as $250+ board.

[Dil-tech]

Originally Posted by Sheep221

I think this is something that extreme - enthusiast boards resolve. Depending on the chipset, but virtually any MB can overclock today, but having overclocking features is not automatically having overclocking capabilities. You can buy OC board for as low as $70 which is not the same as $250+ board.

I'm talking about measuring what difference it makes on video card overclocking, which is a feature that isn't technically part of the motherboard's features but rather can be held back by the vregs on the motherboard itself that run to the pci-e slots.

[K31TH3R]

I've experienced the same thing, and you're right on the money with the cause. Power circuitry with extensive ripple going to PCI-E slots will degrade overclocks.

Demanding scenes which require a spike in GPU processing can and do cause dips on lower end PCI-E circuitry and this will absolutely cause TDR's, and driver crashes/resets.

Sometimes you can get lucky by replacing poor quality capacitors in the PCI-E power circuitry, most other times there's nothing you can do but get a better board since the entire PCI-E power circuitry was built from the ground up to only meet the PCI-E standard without providing any overhead.