Core i7 day-to-day overclock

[jandlecack]

Originally posted by: Sacraster
seeing all those low vcore and high oc # make me really want to get a d0, my c0 is oc fine, but my vcore is no where near the d0.

Even though the D0s have some nice batches, I wouldn't really upgrade just for a new stepping. A Core i7 920 at plain stock speeds is virtually faster than anything else not i7 out at the moment, sans the QX models. However, even a bump to 3.2GHZ will fix that, which every i7 920 in existence can do

 

[2March]

Originally posted by: Sacraster
seeing all those low vcore and high oc # make me really want to get a d0, my c0 is oc fine, but my vcore is no where near the d0.

Some of them are not D0's

There are definately advantages in those new steppings but you could still end up with one that would need a lot of voltage to OC properly. It's always the luck of the draw, even if the batch is popular.

 

[Shmee]

stress testing my C0 at 3.8 GHz now. Temps are good, should be able to break 4. max temps in p95 currently 51 degrees C. I figured out why I couldnt get higher earlier. turns out I needed to up QPI voltage more. at about 1.440 now I believe. Vcore could probably lowered from the setting of 1.375.

 

[TurK-FX]

i forgot to mention that i used OCCT to test the stability, load is put to all 4 cores. I am suspecting that i may fuck up the putting thethermal paste. I am selling my cpu and getting D0 stepping anyway. i hope i can hit to 4.6-4.7 on D0 stepping.

 

[jandlecack]

Originally posted by: Shmee
stress testing my C0 at 3.8 GHz now. Temps are good, should be able to break 4. max temps in p95 currently 51 degrees C. I figured out why I couldnt get higher earlier. turns out I needed to up QPI voltage more. at about 1.440 now I believe. Vcore could probably lowered from the setting of 1.375.

You get 51C in Prime95 at 1.375Vc and 1.44Vq? Woah. Are those CoreTemps (or RealTemps) or what are you using to measure it?

Also, I'm weary of turning up the QPI past 1.35V. Glad it wasn't necessary so far.

 

[Golgatha]

Originally posted by: AdamK47
I run my i7 920 at 21x180 (locked turbo mode) 24/7. I set it to that speed back in January and it's been 100% stable ever since. A base clock of 200MHz showed slight instability when gaming (particularly games with Unreal Engine 3), but could run Prime95 continuously without error.

3.2Ghz, C0 stepping at 1.225V (need to try lower voltages according to this thread apparently) with Turbo mode disabled 24/7 running 4x Folding@Home clients. Memory is at 1.60V just to be sure everything is solid. I'm not sure why I always give the memory a slight bump, but I do. Every other voltage is stock. I'm using an ASUS P6T.

Passes overnight looping of OCCT, large data set with flying colors. I'm using a Zalman CNPS9900LED to cool it. The Zalman 9500CNPS with adapter bracket didn't cut it once I added another 4870 1GB inside my case for Crossfire X.

At idle the cores average around 42-45°C, with 4x Folding@Home clients running it raises to around 62-65°C, OCCT can get it to around 69-72°C, and 4x Folding@Home clients with a newer game running at the same time makes baby Jesus cry (80-82°C and motherboard beeps at me past 80°C). The first time I heard the motherboard beep due to CPU temps, I about had a heart attack.

 

[jandlecack]

What spec is your RAM? I take it you're not running it in XMP mode. When I run mine at 1.6GHZ (certified) XMP, it sets DRAM volts to 1.65V by itself. I lowered it to 1.6V without stability issues but I have it in the back of my head that they bumped it up for a reason...

 

[1ManArmY]

Originally posted by: jandlecack
What's yours?

After my initial OC to see how far I could get with minimal voltage increase, I settled on 3.2 for a while, then 3.4, and now 3.8 with an insignificant voltage bump. I was able to run 3.2 and 3.4 at 1.18Vc and 1.2Vq, now I'm running 3.8GHZ (201x19) at 1.23Vc and 1.2Vq. No idea how I can keep my Vq that low with a QPI of 7.2GHZ.

Here's some preliminary results, but I'm going to try force down voltages as much as possible next.

http://img179.imageshack.us/img179/783/68189055.jpg

Take this with a grain of salt though as I haven't checked it in a 12 hour benchmark yet which I usually do. Just seeing where I can go in terms of temps and volts at the moment.

what's the correct way to install LINX? I know it's a 7z file and I got a 7z zip program to unzip it to my C drive but I get an error when I try to run it so I have to use OCCT.

What should the setting be at? where should I install it?
thanks

 

[jandlecack]

Far as I remember I just had to extract it and then run Linx. I always set it to 1.5GB, and sometimes 4GB if I have enough time on my hands. Runs usually at 20 because Linx pushes your chip so hard that obvious instabilities will become apparent after even the first run. To get my settings confirmed 24/7 stable I usually run Prime95 anyway, for round about 10 hours. That's sufficient imho.

 

[lopri]

Linpack need not be installed. It runs off a standalone executable. There are packages with easier GUI.

Intel Burn Test

LinX

I thought it's supposed to be run with all available memory and maximum stress? Using more memory will give better scores (Gflops) as well. It takes long, but I think running 5 passes with max RAM would be a better test than 20 passes with less RAM.

 

[jandlecack]

All available memory for me is usually around 4GB.

 

[ectx]

My i7-920 will not be 100% stable at FSB>163 and I need to bump voltage up to 1.35v to get to 163. Is 1.35v too high?

Currently my temp is at 78C after hours of Prime95 large, in-place FFT .. I'd like to lower the temp a little but I doubt it would improve oc much, or would it?


I can run the system and pass all stablity tests at slightly lower voltage, however, it would fail to boot once in a while. Very odd.

 

[Shmee]

It was via realtemp. I saw the recomendation to up the QPI VTT in the original nehalem oc guide. So far I guess my temps are still good

 

[jandlecack]

Originally posted by: ectx
My i7-920 will not be 100% stable at FSB>163 and I need to bump voltage up to 1.35v to get to 163. Is 1.35v too high?

Currently my temp is at 78C after hours of Prime95 large, in-place FFT .. I'd like to lower the temp a little but I doubt it would improve oc much, or would it?


I can run the system and pass all stablity tests at slightly lower voltage, however, it would fail to boot once in a while. Very odd.

1.35V is the highest I will go on air. More than that shouldn't be necessary considering the performance you get out of 3.8GHZ, 4GHZ or even 3.2GHZ alone.

I wouldn't recommend it.

 

[Shmee]

^ IDK, per jacobs guide, as long as ur temps are good, he says 1.4-1.45 last I checked. Perhaps the i7 is more voltage tolerant than the yorkfields?

Keep in mind, I am on air cooling, @ 3.8 Ghz, 1.375 Vcore, and max temps with linx are about 58C per realtemp

 

[2March]

Originally posted by: ectx
My i7-920 will not be 100% stable at FSB>163 and I need to bump voltage up to 1.35v to get to 163. Is 1.35v too high?

Currently my temp is at 78C after hours of Prime95 large, in-place FFT .. I'd like to lower the temp a little but I doubt it would improve oc much, or would it?


I can run the system and pass all stablity tests at slightly lower voltage, however, it would fail to boot once in a while. Very odd.

If my proc would need that much Vcore at that BCLK I would settle for a low OC and call myself unlucky. Maybe sell it and get a D0?

I'm running 180x21 at 1.24Vc/1.2VTT.

 

[kerr]

Here is one of mine. 4.5 ghz

 

[bryanW1995]

Originally posted by: jandlecack

Originally posted by: Sureshot324

Originally posted by: jandlecack

Originally posted by: VirtualLarry
I guess I'm late to the party, I don't even have an i7 rig yet. I'm planning on skipping the 45nm i7 generation, and moving on to the 32nm, whenever that's released.

The 32nm of the Core i5 series? Because I'm positive there are Core i7 32nm chips in the works that will run on 1366/x58.

So is Intel abandoning socket 1366?

No...what gave you that idea?

Originally posted by: aigomorla

Originally posted by: jandlecack
The i7 is a tock...

no the current bloomfield is tick.

The Tock would be the 32nm bloomfield.

Like kentsfield 60nm was Tick.. and yorkfield was Tock.

No.

http://www.intel.com/technology/tick-tock/index.htm

For someone who runs so many i7 rigs, you're a bit misinformed. I like your anime taste though.

don't feel bad, aigo, I used to think the same thing. very couter-intuitive, isn't it? they should have called it tock-tick!

 

[jandlecack]

Um, it's not really counter-intuitive at all. The "tick" is the new microprocessor microarchitecture (i.e. die shrink) that enhances the previously existent processor technology. The "tock" is the same microarchitecture with a new processor technology on it. Makes perfect sense when you think about it. First the size, then the technology. Tick-Tock.

 

[imported_Shaq]

Originally posted by: 2March

Originally posted by: Sacraster
seeing all those low vcore and high oc # make me really want to get a d0, my c0 is oc fine, but my vcore is no where near the d0.

Some of them are not D0's

There are definately advantages in those new steppings but you could still end up with one that would need a lot of voltage to OC properly. It's always the luck of the draw, even if the batch is popular.

I can attest to that as mine needs 1.44 with no droop for 3.8 plus turbo. Turbo off didn't allow me to decrease voltage so I left it there. Even at 3.6 I need about 1.4 vcore. I always get the bad part of the wafer. My q6600 G0 needed 1.54 vcore for 3.3. lol

 

[supertle55]

PRIME95 for 3 hours

Here's my system:
I7 920 Overclocked to 3.8ghz at 1.25v and staying under 70°C

 

[lopri]

Originally posted by: kerr
Here is one of mine. 4.5 ghz

All I can say is WOW. What is your board and cooling? Just the other day looking at my 955BE churning out 48 GFlops, I was thinking how wasteful today's desktop systems had become (at least for me). 60 GFlops w/ just 6GB is absolutely ridiculous. Did you get that with HT on or off? I heard that Linpack was so well optimized for multiple cores, HT would be actually hurting the performance. Anyway, it is simply amazing how much raw power Nehalem brings in the HPC world.

Edit: Reality check (comparison) here

 

[2March]

Originally posted by: Shaq
[Turbo off didn't allow me to decrease voltage so I left it there.

That is odd. Without turbo you run 200MHz slower. Innitially I hadn't figured that out aswell because I thought it would only increase the multiplier when utilizing one core so I tested with 1.2Vc, fired up Prime and crashed miserably

At 180 x 20 I need about 1,2Vc but with Turbo on about 1.24Vc. Maybe yours didn't hit turbo because temps were too high?

 

[jandlecack]

Originally posted by: 2March

Originally posted by: Shaq
[Turbo off didn't allow me to decrease voltage so I left it there.

That is odd. Without turbo you run 200MHz slower. Innitially I hadn't figured that out aswell because I thought it would only increase the multiplier when utilizing one core so I tested with 1.2Vc, fired up Prime and crashed miserably

At 180 x 20 I need about 1,2Vc but with Turbo on about 1.24Vc. Maybe yours didn't hit turbo because temps were too high?

Turbo (21x multi) also runs your CPU at 140W TDP.

 

[2March]

Originally posted by: jandlecack
Um, it's not really counter-intuitive at all. The "tick" is the new microprocessor microarchitecture (i.e. die shrink) that enhances the previously existent processor technology. The "tock" is the same microarchitecture with a new processor technology on it. Makes perfect sense when you think about it. First the size, then the technology. Tick-Tock.

I disagree

Architecture is most important. After that comes the die-shrink.