i7 920 overclocked to 3.8 GHz - Prime Stable -- Random Restarting?

[Axon]

This is the culprit: OCZ Gold DDR3 3x2GB 1600 (155x~ mhz) @ 1.66v. What are your timings? I would also reduce the RAM Mhz to 1333, or to any speed below 1600. While that kit is rated at 1600, it rarely can achieve that speed with any form of stability.

It drove me bonkers for three months while I figured it out - I had random restarts, rebooting loops whenever I restarted windows (while shutting down the PC for 90 seconds or more would ensure that it would boot just fine). I am not someone who is inclined to trash a particular piece of hardware, but there are so many stories of this RAM not running at its rated speed across the board.

http://www.newegg.com/Product/Produc...ords%29&Page=2

These are my settings for 3.6 on my ASROCK Extreme/OCZ Gold. 100% stable. Most will translate, but some will not, as I'm sure you already know. Give this a shot then bump for 3.8 accordingly (e.g., blck to 190, mess with QPI).

Overclock Mode

BCLK Frequency (MHz) [180]
PCIE Frequency (MHz) [100]
Boot Failure Guard [Enabled]
Spread Spectrum [Disabled]

CPU Ratio [20]
QPI Frequency [6.480GT]
Uncore Frequency [2880MHz]
DRAM Freqency [720MHz (DDR3 1440)] CRITICAL setting

▶DRAM Timing Control [8-8-8-24-88 (or auto)...auto for the rest, then CR-1T]

ASRock Vdrop Control [Without Vdrop]
CPU Voltage [Manual]
Voltage Value [1.21V]
DRAM Voltage [1.654V]
IOH Voltage [Auto]
VTT Voltage [1.340V]
ICH Voltage [Auto]
IOH CSI Voltage [Auto]
IOH/ICH PCIE Voltage [Auto]
CPU PLL Voltage [1.80V]

CPU Settings

CPU Ratio [20]
Enhanced Halt State [Disabled]
Intel Virutalization Tech[Disabled]
CPU Thermal Throttling [Disabled]
No-Exeute Memory Protection [Disabled]
Hyper Threading Technology [Disabled] (for now)
Active Processor Cores [ALL]
A20M [Disabled]

 

[ItsJono]

Thanks Axon, I'm STILL getting the restarts so I had to search for this post to see if anyone provided any useful information! I will put it to use immediately!

Edit: I was already at 1461 MHz for DRAM Frequency

All I did was lower the DRM Voltage from 1.66v to 1.64v, I doubt it'll change it but it is overspec

 

[Idontcare]

Thanks Axon, I'm STILL getting the restarts so I had to search for this post to see if anyone provided any useful information! I will put it to use immediately!

Edit: I was already at 1461 MHz for DRAM Frequency

All I did was lower the DRM Voltage from 1.66v to 1.64v, I doubt it'll change it but it is overspec

This may be a long shot but some asshat posted this in your thread: http://forums.anandtech.com/showpost.php?p=29808978&postcount=4

Did you give it a read?

 

[Axon]

Thanks Axon, I'm STILL getting the restarts so I had to search for this post to see if anyone provided any useful information! I will put it to use immediately!

Edit: I was already at 1461 MHz for DRAM Frequency

All I did was lower the DRM Voltage from 1.66v to 1.64v, I doubt it'll change it but it is overspec

Okay. Try your timings then. Make sure they are exactly as specified. Lower the DRAM Freq. speed even more if you have to.

Look at IDC's post there, it is helpful and will narrow down the problem. If you run Prime 95's Blend test and get an error, but can easily pass SmallFFT, you've got a RAM (or mobo) issue.

 

[jvroig]

This may be a long shot but some asshat posted this in your thread: http://forums.anandtech.com/showpost...78&postcount=4

Did you give it a read?

ItsJono, this is an issue you could at least answer if you want better help. Depending on what Prime Test you did, the correct solution to your problem may differ significantly.

 

[MJinZ]

You're not stable.

 

[Suspicious-Teach8788]

You know QPI can be bumped up a lot more. You'd be surprised. I don't know about your kit specifically, but many XMP profiles for some Elpida BBSEs require up to 1.6 volts. I have a forum post that cites Corsair, GSkill, Patriot etc for having recommendations up to 1.6 or even higher for QPI. I know Intel's max is 1.35, but when I tried to abide by that I couldn't even boot at the stock 2000 9-9-9 rating... even though Vdram was fine at 1.6 or 1.62v.

 

[mbevolution]

try upping the fsb voltage etc. sometimes the cpu can be hours occt stable but would restart occasionally due to the board itself. mine did that at 1.3v but after i put it to 1.32v the problem disappeared.

 

[ItsJono]

This may be a long shot but some asshat posted this in your thread: http://forums.anandtech.com/showpost.php?p=29808978&postcount=4

Did you give it a read?

I did. BLENDED test. However I believe it's irrelevant because the restarting is random. Using OCCT as per recommendation, testing at full load did not prompt system restarts on two separate tests so its something else I would be inclined to believe.

You're not stable.

You're not helpful and your post-count increasing reply isn't needed.

try upping the fsb voltage etc. sometimes the cpu can be hours occt stable but would restart occasionally due to the board itself. mine did that at 1.3v but after i put it to 1.32v the problem disappeared.

my vTT/QPI/DRAM Core Voltage was set to Auto, I set it to 1.28x v which is about .03 away from my CPU 1.256v~ Hopefully this is the culprit of this instability I have a feeling it'll still randomly restart. I'll go to 1.30-1.35v~

and those who are being helpful, my sincere thanks goes out to ya

 

[Duvie]

Are you small FFT or large FFT prime stable?

If you tested with "blend" then the results don't really mean much.

Small FFT tests the cpu, but you could still be unstable wrt ram or the QPI/memcontroller/L3$ (the uncore on your i7 920).

Large FFT tests your memory, the QPI, and your L3$.

You can't claim prime stable unless you run both the small FFT and large FFT tests and passed both. Recommended test time is 3-6hrs for each FFT type.

I agree with this...

What OS are you running? Winxp has a setting on b ydefault that mask BSOD with a reboot.....It reboots when you have a critical error and you never get to see the BSOD which is actually quite helpful when diagnosing an issue....

I thought of this issue the moment I read this...I also though I bet he only ran small FFT or blend...

Thirdly I thought maybe the power supply....

 

[Nemesis 1]

I had similar problem with my IC7 max3 . Pushing BH1 to the limit . After modding the voltage to memory, no more bsods . Your problem tho differant looks to be memory as stated. But Duvie mentions the PS that also could be the problem .

 

[ItsJono]

I agree with this...

What OS are you running? Winxp has a setting on b ydefault that mask BSOD with a reboot.....It reboots when you have a critical error and you never get to see the BSOD which is actually quite helpful when diagnosing an issue....

I thought of this issue the moment I read this...I also though I bet he only ran small FFT or blend...

Thirdly I thought maybe the power supply....

Windows 7 Ultimate X86-64 (legit)

 

[ItsJono]

What's odd is that this issue especially happens during encoding videos using Sony Vegas 9.

I have HyperThreading enabled, so i'm guessing that encoding is stressing my computer more than a synethetic test. I hope that helps someone help me find out what's going on because the problem still persists!

I'm now at 1.27v for CPU voltage
and 1.33v for VTT/QPI/DRAM Voltage

I don't think it could be my PSU because only my CPU is at full load, not my GPU so gaming should be more stressful and power demanding on an average basis, however, encoding /w HT on seems to really bother my system.

 

[jvroig]

I did. BLENDED test. However I believe it's irrelevant because the restarting is random. Using OCCT as per recommendation, testing at full load did not prompt system restarts on two separate tests so its something else I would be inclined to believe.

How long did you run OCCT? Was it OCCT classic or OCCT: Linpack? And was HT also on?

Also, different stress tests sometimes catch different types of errors. Being stable in one particular stress test does not always guarantee 100% stability. That is why some often use several tests to determine stability, for example P95 Large FFT + P95 Small FFT + LinX + Memtest HCI before calling it good. Anecdotally, I remember one OC I thought was good and stable already, as it passed 4 hours of Prime95 Small FFT. Out of curiosity, I decided to try Large FFT and it produced a rounding error in far less than 1 hour.

If I were in your case, I would personally test using Prime95 Small FFT for 4 hours, then Prime95 Large FFT for another 4 hours, both with HT on. I won't bother with Blend anymore since you said you already passed it. Also, out of all the Prime95 tests, blend is the "easiest" one, generally, unless your RAM OC is unstable. It heats up the CPU less than Small and Large FFT (by experience, the hierarchy is Small FFT > Large FFT > Blend).

 

[ItsJono]

How long did you run OCCT? Was it OCCT classic or OCCT: Linpack? And was HT also on?

Also, different stress tests sometimes catch different types of errors. Being stable in one particular stress test does not always guarantee 100% stability. That is why some often use several tests to determine stability, for example P95 Large FFT + P95 Small FFT + LinX + Memtest HCI before calling it good. Anecdotally, I remember one OC I thought was good and stable already, as it passed 4 hours of Prime95 Small FFT. Out of curiosity, I decided to try Large FFT and it produced a rounding error in far less than 1 hour.

If I were in your case, I would personally test using Prime95 Small FFT for 4 hours, then Prime95 Large FFT for another 4 hours, both with HT on. I won't bother with Blend anymore since you said you already passed it. Also, out of all the Prime95 tests, blend is the "easiest" one, generally, unless your RAM OC is unstable. It heats up the CPU less than Small and Large FFT (by experience, the hierarchy is Small FFT > Large FFT > Blend).

Prime95 Small FFT - 8 Threads for Four Hours - PASSED

 

[Rubycon]

Prime95 Small FFT - 8 Threads for Four Hours - PASSED

I don't use Prime for this very reason! I can pass literally days and yet will get an error on the LATEST linpack libraries (essential for Gulftown/Westmere). Try using LinX 0.6.4, all memory and set it to run for 300 minutes or longer. If THAT passes you should not have issues with other programs shutting down your pc.

 

[Rifter]

You are not prime stable.

 

[biostud]

so, have you disabled TURBO?