Ok, it's "hat in hand" time. I fail. (i7 OC)

E

enlightenedby42

Supreme [H]ardness
Joined
Jan 19, 2005
Messages
4,412
I am not having any luck overclocking my i7 920, and I think maybe I'm just missing something in the settings. First thing's first, here's the setup:

CPU/HSF: i7 920 (c0) w/ TRUE Black and Noctua 120mm 1200rpm fan
MOBO: Asus Rampage II Extreme
VGA: EVGA GTX 275
RAM: 6gb Corsair Dominator 1600
PSU: Corsair HX1000
HDD: Boot Drive: X-25m SSD, Games and storage: WD Black 1TB
OS: Windows 7 x64 RC1 (build 7700)

My results have just been odd. I turn SpeedStep off, of course, and can go to 3.2 on stock voltages with hyperthreading enabled. This is stable in OCCT and 8 instances of cpu stress Orthos.

For some reason, I am hitting a total wall at 3.2 with hyperthreading enabled, and it takes .1-.2 volts just to eek another 1-200 MHz out of it. If I disable it, I can get up to around 3.5-3.6 on 1.4-ish volts and keep it fairly stable.

Temps break into the lower 70's at 1.4, but just barely cross 60 under load at stock. Not GREAT, but seems to be about expected for an overclocked i7 from what I have read. The errors I'm getting on CPU stress tests with HT enabled are basically ALWAYS bsod's.

The only other setting I have messed with is DRAM Bus voltage which I locked down just shy of 1.6. I'm almost certain that RAM isn't the issue, as it will handle 8 instances of blend or RAM stress tests in Prime without any issues at all.

I'm fully aware that overclocking is a gamble and sometimes you don't get lucky, but from what I have read and some of the comments Kyle and Dan have made about their experiences across several chips, I get the distinct impression that maybe I'm missing something here. It seems like my chip is hitting the wall of diminishing returns with voltage increases REALLY fast for an i7.

Is there perhaps another small voltage bump on an obscure component or another setting somewhere that is throwing me off that may be worth trying? Any suggestions in general?

I'm a bit stumped at this point. :confused:
 
1.4 to say 1.55 is not abnormal in trying to overclock these co,s at all. Maybe up qpi vtt to say almost the same and you should get there also pll and so forth. it will get there my co takes 1.5 to get to 4.3 with hyper threading and almost 1.55 to get 4.450. Also make sure qpi and uncore are set right.
 
Last edited:
I don't own an i7 myself, but from what I've seen so far, in order to overlock with hyperthreading enabled you have to put in up to an extra 0.05v-0.07v to keep it stable.

http://www.ocforums.com/showthread.php?p=6097232#post6097232

You could just disable HT, but you'll probably lose more real world performance than you'll gain from a higher raw cpu frequency.
 
I have a 920 at 4.15 GHz.
VCore 1.435
CPU VTT + 350mV

Those are the two key values.
You really need to lean on the C0 920s to get them up in the 185 to 195 BClk range.

I know this is from the EVGA Clasified, but the boards are very similar, give this a read-over:http://www.evga.com/forums/tm.asp?m=100528455
 
Plumber4life said:
1.4 to say 1.55 is not abnormal in trying to overclock these co,s at all. Maybe up qpi vtt to say almost the same and you should get there also pll and so forth. it will get there my co takes 1.5 to get to 4.3 with hyper threading and almost 1.55 to get 4.450. Also make sure qpi and uncore are set right.
Click to expand...
QPI is good, I haven't checked uncore. I will double check it's in the right ratio with the memory multiplier (or however it works, exactly) and such this evening. Good call, thanks.
 
spine said:
I don't own an i7 myself, but from what I've seen so far, in order to overlock with hyperthreading enabled you have to put in up to an extra 0.05v-0.07v to keep it stable.

http://www.ocforums.com/showthread.php?p=6097232#post6097232

You could just disable HT, but you'll probably lose more real world performance than you'll gain from a higher raw cpu frequency.
Click to expand...
That's sort of what I was thinking. Really, my main concern is that it seems that HT doesn't necessarily need to be disabled to get decent clocks a lot of the time and that perhaps it indicated I was doing something wrong that would make things automagically easier.

magoo said:
I have a 920 at 4.15 GHz.
VCore 1.435
CPU VTT + 350mV

Those are the two key values.
You really need to lean on the C0 920s to get them up in the 185 to 195 BClk range.

I know this is from the EVGA Clasified, but the boards are very similar, give this a read-over:http://www.evga.com/forums/tm.asp?m=100528455
Click to expand...
Awesome rundown of mobo features, thanks. Much googling just hasn't turned up much in the way of useful info. It seems like a lot of forum posts around the net have 1-2 posts of useful info and then degenerate into either banter or trolling. That, or people just changing 3 settings and hitting 10GHz on air. :D

It sounds like the overridding theme is that I just need to be less nervous about throwing the coals on the fire. Will definitely punch it tonight with the suggestions in mind and see what I can do. Keep the suggestions coming if you have them, thanks!
 
SuperTroye said:
Don't mean to be a troll here, and don't have much to contribute, but I got more stable results when I had turbo mode turned off. Is this the case for you? A little snipped about Turbo mode here:

Ultimate Core i7 Overclocking Guide -- We Push Nehalem to its Limits
Click to expand...
Not trolling at all, and that reminds me I neglected to mention that in the OP.

My IMPRESSION was that if you turn speed step off, turbo goes automatically. I can't even find a setting for it in the BIOS with SS disabled, so I THINK I've removed that from the equation.

And, of course, thanks for the linkage.
 
Right, yeah I remembered that too, it's the Hyperthreading you are having issues with.
 
I've had an hour or so to try and walk things in this afternoon, and I think I figured out what the problem was. Ever since I locked the uncore frequency down to exactly twice the memory speed things have been a lot more stable with 8 prime instances. That's the only major change, so I'm wondering if Auto was doing something crazy with it. I would imagine it being excessively high could be a problem as well given the things it controls, so maybe that was it.

I got 3.6 at just shy of 1.3vcore stable for around half an hour after starting at 1.400 and moving backward. Getting some blue screens moving forward.

Temps at 1.4 and are breaking 80 under 8 prime threads, so that's the voltage wall without better cooling. I did have a 1600 rpm S-Flex that was a little loud for my tastes, but perhaps it could shave a degree or two off the lower CFM but quieter noctua. Could always try a reseat, too, but that's getting old given that the first mobo was fried and I'm out of Shin Etsu TIM.
 
You must log in or register to reply here.
Back
Top