Oddity, not really CHV related.

At the request of my brother I downloaded and ran a synthetic benchmark called Sanctuary 2.2.

I'm trying to figure out why my score would go UP a bit by setting my OCed I7-920 back to the stock 2.66GHz from the 3.2GHz I have it OCed up to.

The scores were:
3.2: 11174
2.66: 11215
3.2: 11201

Thoughts? Speculation? Idle threats? Pics of hot overclocked ladies?

Cuz OCin' is da debil?

[ ProTip: The banner is a link to art refs!! | The Khellection | The HBAS Repository | Brute Guides (4/16/10) | How To Post An Image - A Quick Guide ]
Biggest Troll on the forums? I'll give you a hint:

OK, that's a Dx19/Dx10/OpenGL benchmark. Maybe you're GPU bound?

You have an HD 5870 right? You should, for giggles, give their Heaven benchmark a try and look at all the pretty tessellation.

Someone in game mentioned that perhaps I need to bump the voltage a bit to compensate for the OC?

The RAM should be good. I specifically bought higher clock rated RAM that I needed so at 3.2GHz on the proc, the RAM is now running at its proper rated speed.

I'll check out Heaven, FX, and yes on the 5870.

What are your temperatures like?

Possibly at the higher overclock the cpu temperature is causing it to throttle back with the result that you aren't getting the clock speed you expect.

Temps aren't great but it depends on how hot my computer room gets. The other night while playing CHV I saw a spike on one of the 4 cores of 71c. From what I've been reading, at 75c things have gone bad with the I7.

However, during the bench run I never saw the temps go above around 58c.

Given the small sample size you listed, it looks to me like you may just be within the error margin of that test.

Granted, that's a rather large error margin, but Sanctuary is designed more as a GPU "show-off" type test rather than one of the hard-core testing apps.

The scores were:
3.2: 11174
2.66: 11215
3.2: 11201

I'd want several more runs at stock speed vs OC'd speed before I'd make any conclusions.

As far as upping your voltage, if you are stable (several long Prime95 or other CPU stress test apps), then you shouldn't need to change the voltage.

I usually up the voltage a bit to find either my max (or hit my desired) OC, then start scaling back the voltage until it becomes unstable, so I can use minimum voltage for maximum OC.

Doing so, I've found that "stock" voltage is almost always way too high for "stock" speed stability and you can usually use that extra power for mid-level OCs. (assuming hard-set voltage and not "auto" settings)

I like that line of thought, Soulwind.

Oh, and I'm pretty sure you already know this (since you've already OC'd the CPU) but...

OverClocking the 1366 chips isn't quite as straight forward as OC'ing the old 775 chips.

What with moving the memory controller onto the CPU itself and getting rid of the seperate northbridge and all, there's a few gotcha's that can crop up and hurt performance.

Here's a nice basic guide for OC on the I7's. No earth-shattering revalations in it, but it does list a few of those 'gotcha' things to look out for and does a pretty good job of showing the differences in the 1366's vs the 775's.

http://www.chilledpc.co.uk/forum/showthread.php?t=9960

Interesting. All I did to OC mine was bump the BCLK from 133 to 160. I left everything else alone.

This change took my 2.66 proc to 3.2 and bumped my RAM up to the speed I purchased it at : DDR3 1600

After reading that thread I'm now wondering if I should back down the multiplier on QPI but why should the QPI stay around 2400MHz?