My Core i7 920 Overclocking Experience

[exar333]

Installed the Noctua cooler today and got the cpu up to 3.6 with a .06v bump (stock is 1.23 and I am running 1.295).

BCLK 180 x 20 = 3.6ghz

Memory voltage was bumped from 1.5 to 1.56 to keep the memory stable at 1443mhz (8-8-8-24 1T).

With the new cooler, the cpu idles at 41/40/49/37 and at 100% load (after 2 hours) 70/68/67/66.

I can't even imagine what this could do on water...

3DMark vantage cpu score and superpi coming later tonight; I will post these shortly!

Edit:typo

 

[AdamK47]

Originally posted by: ExarKun333
Installed the Noctua cooler today and got the cpu up to 3.6 with a .06v bump (stock is 1.23 and I am running 1.925).

1.29V you mean?

I have mine at 1.3V 3.6GHz (200x18) with the memory at 1.66V 1600DDR. Went with the Cooler Master V8 instead. It seems great so far.

 

[exar333]

Nice OC Adamk47! Thanks for the head's up on the typo; fixed!

First rounds of testing:

3DMark Vantage (performance): 44778

Super Pi 1M: 11sec

SiSoft Memory Bandwidth: 25268 MB/s

Fallout3 (via FRAPS): 1900x1200 Ultra quality DX10 Min: 52 Max:54

I was really happy with the amazing minimum framerates. The system even allowed me to record FRAPS video and stay above 30fps; thats crazy!

Edit: Added Fallout3 and SiSoft data

 

[samw5]

well if that makes you feel any better I've hit a wall at 188 BLCK.

I'm running my 920 on a P6T Deluxe OC Palm board, with 6GB of Mushkin 1600 DDR3 (7-8-7-20).

I've tried a bunch of stuff on this but cannot seem to get it up past 190 even while reducing the CPU multiplier to 19 or the DDR speed to the 1200 range (it's rated at 1600@1.65V so I know it isn't the issue).

From what I can tell I think the CPU can get higher but the motherboard is limiting how high I can push the BCLK (maybe not enough QPI Voltage, I'm not sure, I've also tried a bunch of stuff here).

The CPU is H2O cooled with a DD TDX block, black ice dual radiator and Swiftec pump (1/2 tubing) so cooling isn't the issue here.

Now under full load the CPU will hit 75-80C but from what I can tell this isn't too uncommon those chips have a rating of 92C or so. Others have had the chip running around 85C under 100% load. Now to get the chip at 100% takes some efforts (even under heavy multitasking I can barely get it above 50%). Temps under normal heavy use (not benchmarks) will hover around 60-65C which isn't bad.

Now voltages... The reason I think the CPU will go higher is that I've been able to reduce the Vcore down to around 1.30V while running it at the same BLCK (188) with it being pretty much stable (I did plant under heavy benchmark but the system was stable otherwise). Now reducing the QPI (currently set at 1.3475V) make the board no longer post.
The RAM is at 1.64V and PLL at 1.96 though I found the PLL voltage didn't do much to keep the system stable.
Amplitude set at 800mV as other's have recommended... seems to work ok.

I've upped the Vcore to 1.35V and QPI to 1.35V and back down on the BCLK to 185 (that's 3.7Ghz which is fine so far). Temps haven't increased much (idle a bit more). DDR is running around 1492Mhz at 7-8-7-20-1N. QPI is set to middle value (I believe it's around 8292 or so). Turbo mode and speedstep off of course.

I've seen some people pushing 200 BCLK... I don't think my batch will do this, even upping voltages (I've gone to the first "red" voltage setting option for both Vcore, DRAM, and QPI w/o any increase in BCLK stable).

Hope this will give you some insight...

Oh ya Super Pi 1M @3.7 will take around 9seconds... Didn't run any other benchmarks but Crysis runs smoothly @1680x1050 (with an Asus EAH4870 1GB).

 

[aznxk3vi17]

My i7 920 is only doing Super PI 1M in 11.537s, even at 3.8 GHz. Memory is 1600 MHz at 8-8-8-21. Any reason why this would be?

 

[AdamK47]

samw5, try to set your PLL voltage to Auto and set your QPI to 1.35V. See if you can get a stable 200MHz base clock then. That's what I have on my P6T Dlx. It's stable at 19x200 for 3.8GHz.

 

[samw5]

will give it a shot over the weekend... w/o much hope... had QPI all the way up to 1.4V w/o being able to push over 190 (even with a 19multi).

I think the CPU has some serious potential... motherboard looks like it's holding it back. Maybe a new BIOS will help.

Will post the results.

 

[JohnVM]

exarkun: Do you have a cold room or osmething? At 3.6Ghz I have like 75-77C under full load w/ Prime95, and I have the Noctua too.

samw5: At 3.7 I had SuperPi 1M of 11.07s, so I dunno how you got yours down to 9. Doesn't sound right to me.

 

[Tempered81]

Originally posted by: aznxk3vi17
My i7 920 is only doing Super PI 1M in 11.537s, even at 3.8 GHz. Memory is 1600 MHz at 8-8-8-21. Any reason why this would be?

You feel 11 seconds is too slow?

 

[aznxk3vi17]

Originally posted by: jaredpace

Originally posted by: aznxk3vi17
My i7 920 is only doing Super PI 1M in 11.537s, even at 3.8 GHz. Memory is 1600 MHz at 8-8-8-21. Any reason why this would be?

You feel 11 seconds is too slow?

It isn't that it was too slow, it's that it didn't match up with samw5's claim of a 9 second run.

 

[samw5]

i'll run it again tonight, looking at all you guys figures i might be wrong... i'll take the screeny of course... been running too many benchmarks to remember them all.

 

[exar333]

Originally posted by: JohnVM
exarkun: Do you have a cold room or osmething? At 3.6Ghz I have like 75-77C under full load w/ Prime95, and I have the Noctua too.

samw5: At 3.7 I had SuperPi 1M of 11.07s, so I dunno how you got yours down to 9. Doesn't sound right to me.

No, actually it's in my warmest room in the house. Even with a instance of F@H for the cpu and gpu, my processor idle's at 39C; it is closer to 32-33 at "true" idle.

My temps continued to go down after a few days of operation and my current 8-instance load of Prime is right in the mid to high-60's. I have not seen any of the core temps hit 70C over the past few weeks.

In general usage (10-50% util), the temps stay in the mid 40's or so.

The samw5 Superpi score of 9 seconds is amazing; With my setup, I re-ran the test and disabled some background tasks and hit in the 10.8's, but cannot get it faster at the moment.

 

[exar333]

As a side note; I just noticed this thread was revided, so I apologize for my slow response!

I definitely feel I have the headroom to hit 3.8-4.0, but I am very comfortable with my current temps and everything I am running is working great. My goal is to run for a another few weeks and then experiment at 3.8 and see if I can keep the system stable.

This has been a great system and I definitely happy I went the i7 route; very satisfied.

 

[SickBeast]

If these i7 chips can only hit 3.6ghz and the Phenom II really can hit 3.8-4.0ghz, then things are really going to get interesting, especially if AMD's new chips cost 1/2 as much as the i7. :Q

 

[Idontcare]

Originally posted by: SickBeast
If these i7 chips can only hit 3.6ghz and the Phenom II really can hit 3.8-4.0ghz, then things are really going to get interesting, especially if AMD's new chips cost 1/2 as much as the i7. :Q

And the mobo's cost less, and the DDR2 ram costs less

(or doesn't need to be purchased at all for some upgraders)

 

[samw5]

ok you guys are right, redone my Super Pi 1M runs and I'm getting all in lows 11s (11.163).

Must have ran another test that was in the 9s... my apologies.

I'm also tried the recommendations to set PLL to auto QPI @1.35V but no go past 190 BCLK even at 19 multi.

I even tried raising the IOH to 1.35V w/o changes.

I guess I'll have to settle with this. I'm going to try a new BIOS as well see if that makes a difference. Noticed I was still on 0703 (that shipped with the board).

 

[anindrew]

This thread is rather helpful. So far, I have my i7 920 at 3.6Ghz (180Bclk x 20). I tested with Prime95 and received no errors. My load temp was about 49C while running it. Idle temp is 33C. System temp is 36C. I increased the CPU Vcore to 1.2875v. I will eventually shoot for 3.8Ghz, but a 1Ghz overclock is REALLY good, I think.

Here's what I'm running:

Intel i7 920
Noctua NH-U12P SE1366
Gigabyte GA-EX58-UD5 F3 BIOS
6GB G.Skill DDR3-1600 8-8-8-21 F3-12800CL8T-6GBHK
Evga Geforce GTX260 Core 216
Antec 900
Raptor 74GB and 150GB
Gateway FHD2400 24" LCD

 

[palladium]

*Ignores spam post above*

OP, did you manage to get a stable system with 150*20 @1.09xxV while you're using the stock HSF? I'm Priming as I'm typing this and after 2 hours I'm getting 66/67/66/64 on stock HSF ( 133*21=2.8GHz @1.136V). Maybe I'll try 140 BCLK after another hour or 2. Third party coolers are out of stock in my region ( NZ) at the moment , so I guess I have to stick with stock HSF for now. I'm aiming for ~3.0GHz ( so that my GTX280 doesn't get bottlenecked too much, lol). This is my first time OCing so I don't want to push it too far.

 

[exar333]

Originally posted by: palladium
*Ignores spam post above*

OP, did you manage to get a stable system with 150*20 @1.09xxV while you're using the stock HSF? I'm Priming as I'm typing this and after 2 hours I'm getting 66/67/66/64 on stock HSF ( 133*21=2.8GHz @1.136V). Maybe I'll try 140 BCLK after another hour or 2. Third party coolers are out of stock in my region ( NZ) at the moment , so I guess I have to stick with stock HSF for now. I'm aiming for ~3.0GHz ( so that my GTX280 doesn't get bottlenecked too much, lol). This is my first time OCing so I don't want to push it too far.

Yes, I was able to get 3.0 stable with undervolting the cpu on the stock HSF. When I went above 3.0-3.2, it was just getting too hot for me to be comfortable with. IMHO, the stock HSF is pretty crappy and should have a bigger copper core to dissipate heat. I understand that is more $$ for intel, but it should better...

anindred - I am glad this thread has been helpful for you; you have a very similar setup to mine so you definitely should be shooting for 3.6+ eventually as you continue to tinker.

 

[palladium]

I agree - I think it's probably one of Intel's plans to make overclocking harder and less affordable to the end user.

BTW, I have a small problem now. I finished 3 hours of prime95 without any errors ( at 2.8GHz, 1.136V). I tried running super pi, but it keeps crashing after I click OK to the dialog box ' now beginning calculation' , be it 16k, 1M or 32M. Tried restarting and bumping up Vcore to 1.2V, did not help. Ran LinX( Lapack with GUI) for 30 mins after reboot ( still running it as I'm typing this), no errors . Have I damaged my CPU?

 

[jdogg707]

Originally posted by: palladium
I agree - I think it's probably one of Intel's plans to make overclocking harder and less affordable to the end user.

BTW, I have a small problem now. I finished 3 hours of prime95 without any errors ( at 2.8GHz, 1.136V). I tried running super pi, but it keeps crashing after I click OK to the dialog box ' now beginning calculation' , be it 16k, 1M or 32M. Tried restarting and bumping up Vcore to 1.2V, did not help. Ran LinX( Lapack with GUI) for 30 mins after reboot ( still running it as I'm typing this), no errors . Have I damaged my CPU?

Does it successfully complete at stock?

 

[palladium]

Cleared CMOS, still didn't work. No prime 95 errors after 4 hours ( after clearing CMOS), either.

I should probably start this in a new thread.

EDIT: Worked after I dragged the window to the top right corner, weird.

 

[Dadofamunky]

Originally posted by: palladium
I agree - I think it's probably one of Intel's plans to make overclocking harder and less affordable to the end user.

BTW, I have a small problem now. I finished 3 hours of prime95 without any errors ( at 2.8GHz, 1.136V). I tried running super pi, but it keeps crashing after I click OK to the dialog box ' now beginning calculation' , be it 16k, 1M or 32M. Tried restarting and bumping up Vcore to 1.2V, did not help. Ran LinX( Lapack with GUI) for 30 mins after reboot ( still running it as I'm typing this), no errors . Have I damaged my CPU?

Really, really doubt it. You have admirably conservative voltages. But run Prime for at least twelve hours. You're also on the right track using Linpack. I also like to use OCCT as part of the test battery. Sounds like a software bug somewhere.

 

[palladium]

Originally posted by: Dadofamunky

Originally posted by: palladium
I agree - I think it's probably one of Intel's plans to make overclocking harder and less affordable to the end user.

BTW, I have a small problem now. I finished 3 hours of prime95 without any errors ( at 2.8GHz, 1.136V). I tried running super pi, but it keeps crashing after I click OK to the dialog box ' now beginning calculation' , be it 16k, 1M or 32M. Tried restarting and bumping up Vcore to 1.2V, did not help. Ran LinX( Lapack with GUI) for 30 mins after reboot ( still running it as I'm typing this), no errors . Have I damaged my CPU?

Really, really doubt it. You have admirably conservative voltages. But run Prime for at least twelve hours. You're also on the right track using Linpack. I also like to use OCCT as part of the test battery. Sounds like a software bug somewhere.

Yeah thought so too. Will do it tonight. So, I take it that low voltage will not damage your CPU?

BTW Dadofamunky, do you mind doing me a small favor? Can you try running super pi 32M (on vista 64) to see if it will work? Thanks

EDIT: Really sorry OP for hijacking your tread.

 

[exar333]

Originally posted by: palladium
Cleared CMOS, still didn't work. No prime 95 errors after 4 hours ( after clearing CMOS), either.

I should probably start this in a new thread.

EDIT: Worked after I dragged the window to the top right corner, weird.

I just finished completing the 32M Superpi; I didn't have any issues running this or getting the results once it finished.

Is this working for your now, or are you still having issues?