i7 920 + Asus P6T Deluxe v2 - Can I go above 3.5GHz?

Amtrak · Sep 30, 2009

You're fine in the 80's. These chips run just fine at high temps. I try and keep mine under 85 under the hottest of tests.

Those stability testing programs are all much harder on your computer than almost any real world applications; that is why they are used for testing stability. If you want to be sure your computer is 24/7 stable, use LinX with a high problem size and/or Prime95 Large FFTs (for CPU) and Blend test (for CPU and memory). If they pass 24 hours of that, you can be relatively sure that your rig is stable.

Also, disabling HT might get you a slightly higher overclock and will certainly lower temps a lot, but it will mean you'll just be running 4 threads. Most programs won't be affected, and some games may even run faster. Well-optimized multi-threaded apps will see a performance hit.

milomak · Sep 30, 2009

Thanks. Is there some CPU Voltage limit I should look to stay under?

baditude_df · Sep 30, 2009

Small fft's stresses the CPU the most. You will never come near that in RL applications, but that's not the point of the stress test. The rationalle is, if you can survive that, you will be able to survive everything else.

For a D0 chip at 4.0GHz you should be ~1.2-1.25v as a rule.
For a C0 you will likely need a bit more. 1.25-1.3v or even a bit more.
Temperatures between the two steppings should be roughly the same regardless of the extra voltage in my experience. Cooler running is not a widely documented benefit of the new stepping, but lower required voltage per clock seems to be.

milomak · Sep 30, 2009

Is there any reason for 24hr stress testing? I oversaw a test for just over an hour (small) and it seemed that it may have been running in cycles. I would think that say after 6 hours (as an arbitrary number), if it hasn't fallen over it is likely stable.

But then again I did initially think that more TIM would make more sense than less TIM.

MIAHALLEN · Sep 30, 2009

baditude_df said:
If you want to push for CPU stability, you need to use small ffts. With that cooler, it will likely mean that you have to back off your OC and Vcore. You're approaching danger levels at 80+degrees.

Amtrak said:
You're fine in the 80's. These chips run just fine at high temps. I try and keep mine under 85 under the hottest of tests.

Those stability testing programs are all much harder on your computer than almost any real world applications; that is why they are used for testing stability. If you want to be sure your computer is 24/7 stable, use LinX with a high problem size and/or Prime95 Large FFTs (for CPU) and Blend test (for CPU and memory). If they pass 24 hours of that, you can be relatively sure that your rig is stable.

Also, disabling HT might get you a slightly higher overclock and will certainly lower temps a lot, but it will mean you'll just be running 4 threads. Most programs won't be affected, and some games may even run faster. Well-optimized multi-threaded apps will see a performance hit.

baditude_df said:
Small fft's stresses the CPU the most. You will never come near that in RL applications, but that's not the point of the stress test. The rationalle is, if you can survive that, you will be able to survive everything else.

For a D0 chip at 4.0GHz you should be ~1.2-1.25v as a rule.
For a C0 you will likely need a bit more. 1.25-1.3v or even a bit more.
Temperatures between the two steppings should be roughly the same regardless of the extra voltage in my experience. Cooler running is not a widely documented benefit of the new stepping, but lower required voltage per clock seems to be.

You guys are too funny. I run my 24/7 rig at 4.2GHz with 1.425V and it hits 95C with Prime95 small FFTs within about 10 minutes

The thing you have to keep in mind is that stress testing puts a heavier load on the CPU than 99% of real world application.

During normal use (and I often encode Blu-ray movies and play games at the same time) my CPU never exceeds 85C. :beer:

Amtrak · Sep 30, 2009

baditude_df said:
Small fft's stresses the CPU the most.

Not for i7. Run 20 minutes of Small FFTs, then 20 minutes of Large FFTs and see which one gives you much higher temps. I've also had immensely stable runs of Small FFTs only to fail soon after starting Large FFTs.

MIAHALLEN said:
You guys are too funny. I run my 24/7 rig at 4.2GHz with 1.425V and it hits 95C with Prime95 small FFTs within about 10 minutes

The thing you have to keep in mind is that stress testing puts a heavier load on the CPU than 99% of real world application.

Everyone has their own preferences about how hot they're willing to run. I just commented that temps in the 80s is in no way dangerous. You're correct in that the vast majority of real world applications won't come close to fully loading the CPU.

baditude_df · Sep 30, 2009

MIAHALLEN said:
You guys are too funny. I run my 24/7 rig at 4.2GHz with 1.425V and it hits 95C with Prime95 small FFTs within about 10 minutes

The thing you have to keep in mind is that stress testing puts a heavier load on the CPU than 99% of real world application.

During normal use (and I often encode Blu-ray movies and play games at the same time) my CPU never exceeds 85C.

I said "approaching" danger levels.

Meathead.

Durring my benches the other night I hit 90deg. plenty of times. Stupid water.

Congrats on the blues MIAH. That's good to see!

milomak · Oct 1, 2009

It seems I may not be able to get this baby to 4GHz (200x20) unless I crank up the CPU voltage to something like 1.45

I've tried at 1.40 and within seconds of starting a Prime Blend test it freezes. I'm pretty happy with 3.8GHz, but there is that little tug inside wanting to hit 4GHz.

When I get back from work this evening, I'll try 1.45 and if it doesn't freeze up immediately; I'll briefly run P95 for an hour before I decide if I'll try to push for stability at 4GHz.

EDIT: Given I manually changed the RAM timings (to specs per Super Talent's website), would an OC stick better if I first tried with Auto settings on the RAM?

MIAHALLEN · Oct 1, 2009

milomak said:
It seems I may not be able to get this baby to 4GHz (200x20) unless I crank up the CPU voltage to something like 1.45

I've tried at 1.40 and within seconds of starting a Prime Blend test it freezes. I'm pretty happy with 3.8GHz, but there is that little tug inside wanting to hit 4GHz.

When I get back from work this evening, I'll try 1.45 and if it doesn't freeze up immediately; I'll briefly run P95 for an hour before I decide if I'll try to push for stability at 4GHz.

EDIT: Given I manually changed the RAM timings (to specs per Super Talent's website), would an OC stick better if I first tried with Auto settings on the RAM?

Overclocking 101....only work on one thing at a time. If you're looking for the CPU limit, change your mem settings to auto. After you've found a CPU speed that you're happy with, then dial in the mem speed/timings :beer:

Back on my first i7 rig, my 920 C0 needed 1.475V to be stable at 4GHz

baditude_df · Oct 1, 2009

Back on my first i7 rig, my 920 C0 needed 1.475V to be stable at 4GHz

Actually, come to think of it I needed 1.42v for 4GHz on my C0 and P6T-D. The 1.25-1.3v I listed above was to run my daily 3.6GHz for Rosetta. Apologies everyone, I was out to lunch there. Miah's comment just rang a bell.

milomak · Oct 1, 2009

1.45V also BSODed within seconds of running P95.

I think I'll just stick to 3.8GHz and try to lower the voltage to 1.375V

milomak · Oct 2, 2009

13hr mark passed on blend test using 1.375V for 3.8GHz (190x20). Highest temps of 84-82-82-77 after 15 minutes. Hopefully when I get back home from work it will still be running. Fingers crossed.

baditude_df · Oct 2, 2009

I'm going to go ahead and see if I can find my stable 4.0Ghz settings when I get home. I had this exact same setup and it Ran 4.0Ghz sooo easilly, I can't help but think you can do it too, but just maybe with a new cooler.

Shadowmonk · Oct 2, 2009

The Corsair h50 works wonders when you switch to a 2 fan setup, the scythe slipstream fans move alot more air than the included fan and it made a huge difference in the temps.

baditude_df · Oct 2, 2009

Yeah, people say that those slipstreams don't work very well on Rads or TRUEs becuase they are fairly low on the Static Pressure side. I though they worked just fine, and they are pretty quiet and relatively inexpensive. I'd still go with a Yate medium though.

milomak · Oct 2, 2009

QPI and CPU Voltage set at 1.375V

IcedOutCpu · Oct 3, 2009

I was able to oc'ed my 920 'till 4.4Ghz on the same mobo but the only difference was I'm on True 120 cooling..

milomak · Oct 4, 2009

Iced - was that with a C0 or D0 stepping?

My main concern with pushing towards 4GHz and beyond is the CPU Voltage. When I tried with 1.45V, the system hung/BSODed (either or would happen) with temps still showing in the 60s. So while I think at 1.45V+ voltage temps will likely creep towards 90, temps are not yet a problem.

baditude_df · Oct 4, 2009

Your Vcore shouldn't need to be that high until you reach the 4.0GHz Range with a C0. But of course all chips are different.
You're going quite a bit over stock on your bclock. Have you raised your IOH voltage at all? If not, you should try bumping it up 2 or 3 notches. I can't remember what stock IOH votage is on that board, but I need 1.3v to run 4.0GHz on this board (920 D0/950 D0).

Unfortunately, I left all of my BIOS settings in a profile on the P6T-D when I sold it. I wrote nothing down, so I don't have a bios listing for you.
I would definitely help us help you better , if you posted a full list of your AI settings. Something might stand out right away.

IcedOutCpu · Oct 4, 2009

milomak said:
Iced - was that with a C0 or D0 stepping?

My main concern with pushing towards 4GHz and beyond is the CPU Voltage. When I tried with 1.45V, the system hung/BSODed (either or would happen) with temps still showing in the 60s. So while I think at 1.45V+ voltage temps will likely creep towards 90, temps are not yet a problem.

I had the D0 stepping.. I was able to hit 4Ghz without going beyond the voltage limit of 1.35v.. Do you have a C0?

i7 920 + Asus P6T Deluxe v2 - Can I go above 3.5GHz?

Member

Member

Northern Senior

Member

Senior Member

Member

Northern Senior

Member

Senior Member

Northern Senior

Member

Member

Northern Senior

Member

Northern Senior

Member

Registered

Member

Northern Senior

Registered

Similar threads