BSOD when OC and play COH

[imported_spike99]

Hi,

I have a QX6850 and I previously had my CPU OC'd from 3GHz to 3.66GHz. I mainly play 2 games, AOE III & COH. When ever I play COH, about 40% of the time I get BSOD or game crashes.

I don't have any issues with playing AOE III or doing anything else... only when I play COH.

I have stressed tested my PC using Everest Ultimate and after running over 8hours... It is fine.

My PC temp usually gets warmer when running AOE compared to COH so from this, I assume that AOE puts more load on the CPU.

When I go back to 3GHz (voltage = AUTO, multiplier = 9), COH never crashes.

I have an expensive CPU cooler and it's a shame that have to run on 3GHz.

Any suggestions are appreciated.

 

[error8]

First of all where did you see people using Everest stress test, to find instability problems with their system? It's not a very reliable software for that. Use OCCT stress tests or Linpack. If your computer passes 2 hours of OCCT cpu test or 10 tests at least in Linpack 1/2 memory usage, then you can conclude that your system is stable. Be careful at what speed are you running that memory, since that may also be the cause of your problems.
I do hope that you've read the overclocking sticky on this forum.

Using the CoolIT freezone cooler, I don't think you have any temperature problems, just a badly done overclock.

 

[imported_spike99]

Hi error8,

I'll looked up OCCT... and looks like the latest version is 2.0.1. Looks like it's VISTA 64BIT compatible.

I'll check this and stress test with my current voltage of 1.4125V. If it fails then I guess Eversest Ultimate is unreliable.

Be careful at what speed are you running that memory, since that may also be the cause of your problems.

I don't overclock my memory. I have also ran memtest to check memory and it's fine.

No, I don't have any temp problem with Freezone Elite... Idle is at 40C and ~ 60C under load with voltage of 1.4125

I'll get back after I run OCCT and give and update.

Thanks for your suggestion!!

 

[imported_spike99]

Update:

I downloaded OCCT, but when running it did not respond. Since it did not run properly, I downloaded IntelBurn Test v1.8(Linpack) instead.

I ran IntelBurn Test with maximum stress and I am running test 10 times...

My initial voltage was 1.4125 and running test... it BSOD'd immidiately. As I increased my voltage it's been more stable... Unfortunately I'm stable with voltage 1.5375v. Last test with voltage of 1.5125v it BSOD on the 9th test run.

This seems like a very high voltage... and having such a high voltage will translate in a noiser CPU Cooler.

I'm now considering running at 3.3GHz in order to lower voltage which would translate into a quieter CPU cooler.

I had previously used Everest Ultimate to stress test... I quess Everest Ultimate sucks for stress testing.

 

[taltamir]

stress testing with a program does not mean it is stable, it just means it is stable with THAT program... you need more voltage or a lower OC. COH is simply more then it can handle.
Also higher temp does not automatically mean less stable.

Also auto voltage is the worst thing you can do, it is NOT default, by automatically overvolted to match the OC, except it massively overvolts and causes CPU damage quite quickly, or at least instability. So manually set your voltage.
Anyways, first test at default CPU speed with no OC, if COH still crashes then it is a COH bug or a driver bug for your video card or something like that.
If it stops crashes then you overclocked too much, or need more voltage or something, and you mistakingly assumed it was stable due to the tests being perfect.

For example, with my Q6600 I oced to 3ghz AND undervolted it a bit, rock solid in all testing programs... and all programs i used, however i later came upon a program that BSODed and the fix was to raise the voltage.
Interestingly enough the most BSODs i get while ocing, even when it is seems stable, are with browsers, like firefox, firefox 64 and chrome. OFcourse i get none now with a sufficient voltage.

 

[imported_spike99]

stress testing with a program does not mean it is stable, it just means it is stable with THAT program...

Makes perfect sense.

Also auto voltage is the worst thing you can do, it is NOT default, by automatically overvolted to match the OC, except it massively overvolts and causes CPU damage quite quickly, or at least instability. So manually set your voltage.

OK

Anyways, first test at default CPU speed with no OC, if COH still crashes then it is a COH bug or a driver bug for your video card or something like that.
If it stops crashes then you overclocked too much, or need more voltage or something, and you mistakingly assumed it was stable due to the tests being perfect.

COH never crashed when I had CPU normal - with no OC. Only when I OC'd did it crash 40% of the time.

Since I stressed tested with Everest I always assumed that COH was the problem until I stopped OC'ing. I have played enough games to say COH only crashes when I OC (with voltage 1.4125v). After I stress tested with IntelBurn Test and determined that 1.5375v was stable I managed to play 3 games with no crashing... but I still need to play another 10 games to start having some confidence that this was the problem... but I'm pretty sure this was the problem.

The only difference is are my temps. With my CPU cooler (Freezone Elite) you can control temp... The problem is the lower you set your temp the higher RPM for fans which translates to higher noise. I have set my temps to lower fan noise which translates to below temps

Before - voltage = 1.4125v
Idle ~ 40C
Gaming ~ 50C

After - voltage = 1.5375
Idle ~ 60C
Gaming ~ 70C

I guess above temps with voltagae 1.5375v is okay but most of the time I get annoyed with the fan noise... When I initially got PC it came with a CPU cooler not suitable for OC but it was dead silent... (Ah, those were good old days). Sometimes it makes me want to go back to no OC'ing and get a much quieter CPU cooler... But having CPU at 3.66GHz does make a difference in game loading time... and I assume in game play. I do notice the game loads faster but I don't notice any difference while I play.

 

[imported_spike99]

Update:

I have played addtional COH games and no BSOD's or game crashing. From this, looks like my initial voltage of 1.4125v was too low for COH.

Running IntelBurn Test v1.8 (Linpack) to stress test PC is definitely the software to use when stressing PC.

Using the CoolIT freezone cooler, I don't think you have any temperature problems, just a badly done overclock.

error8, Just as you had said... I had badly done overclock using Everest Ultimate. Thanks for your suggestion with Linpack !!

 

[error8]

Originally posted by: spike99

error8, Just as you had said... I had badly done overclock using Everest Ultimate. Thanks for your suggestion with Linpack !!

Glad everything turned out fine after all. Linpack is a great tool for stability testing, finding "weaknesses" in just a couple of minutes.

 

[imported_spike99]

Linpack is a great tool for stability testing, finding "weaknesses" in just a couple of minutes.

Yes, I forgot to mention this... Running Linpack for a few mins vs running other software for several hours to check for stability is a no brainer.

Once again.... Thanks !!

 

[DSF]

Originally posted by: spike99

Linpack is a great tool for stability testing, finding "weaknesses" in just a couple of minutes.

Yes, I forgot to mention this... Running Linpack for a few mins vs running other software for several hours to check for stability is a no brainer.

Once again.... Thanks !!

No program can guarantee stability after a few minutes.

 

[imported_spike99]

No program can guarantee stability after a few minutes.

I'm definitely a noobie in OC'ing, but from my experience running Linpack I must disagree with this statement. After running Linpack and determined the voltage that my PC ran stable, COH no longer BSOD or crashes. Running Linpack, I ran 10 tests and each test must have taken ~ 1min... so in total, it must have taken ~ 10mins for final test run...

As mentioned, I previously used Everest Ultimate and ran their stability test for over 8hours and after running over 8hours (In reality I must have ran it over 10 hours)... I assumed I had the correct voltage of 1.4125v.

So, I agree with taltamir... he said the below

stress testing with a program does not mean it is stable, it just means it is stable with THAT program...

So running Everest Ultimate (although a poor choice for running stability tests), it runs stable with 1.4125v.... but COH crashes with this voltage.

So at this point, my COH crashing has been resolved running Linpack to determine my voltage. But who knows... In the future I may buy a new game and possibly encounter problems again due to incorrect voltage... I doubt it... but It's possible.

 

[DSF]

Not finding an instability and being stable are two different things. That much should be clear from the fact that Everest did not error out, while Linpack did.

That doesn't mean that if you left Linpack running for several hours it wouldn't find an error, just because it didn't in the first 10 minutes.

As an example, I'm running an E4500 overclocked to 2.93 GHz. I could push it farther, but I don't need to, so I instead focused on undervolting the chip. As I was going through the trial and error process, on some settings my processor would run for hours on Prime95 (a respected stress testing program) without a problem, only to error out after 8 or 10 hours.

10 minutes just isn't enough to confirm stability.

 

[error8]

Originally posted by: DSF
Not finding an instability and being stable are two different things. That much should be clear from the fact that Everest did not error out, while Linpack did.

That doesn't mean that if you left Linpack running for several hours it wouldn't find an error, just because it didn't in the first 10 minutes.

As an example, I'm running an E4500 overclocked to 2.93 GHz. I could push it farther, but I don't need to, so I instead focused on undervolting the chip. As I was going through the trial and error process, on some settings my processor would run for hours on Prime95 (a respected stress testing program) without a problem, only to error out after 8 or 10 hours.

10 minutes just isn't enough to confirm stability.

DSF, Linpack is something more then Prime95. One thing that makes it special, is that you have the highest temperature you can possibly attain with your chip. My E7200 at 4 ghz, for example, with 1,46 V, would hit as high as 86 C with Linpack in a couple of minutes, while Prime95, Orthos and OCCT would only give around 65-70 C, under load.

Linpack, or Intelburntest, was created by Intel to test their chips for stability. I guess there is no software creator that knows more about stability testing a cpu, then those that made that specific cpu.

You could try it yourself and compare Linpack results, with what you obtain with Prime95. Linpack would see instability in a couple of minutes, while Prime95 could run for hours and hours without seeing any errors. If your chips is stable in Linpack, then it's stable in anything you throw at it. Now, we can also speak about stability and conclude that maybe you don't really need your cpu as stable as Linpack takes it. But that is another story....

 

[imported_spike99]

One thing that makes it special, is that you have the highest temperature you can possibly attain with your chip. My E7200 at 4 ghz, for example, with 1,46 V, would hit as high as 86 C with Linpack in a couple of minutes, while Prime95, Orthos and OCCT would only give around 65-70 C, under load.

I agree, I ran Everest Ultimate and at one time had used Prime95 too and my CPU never got past 80C. Running Linpack, my CPU temp reached a little over 90C.

Also, I previously said that it took 1 min per test... I re-checked it yesterday and it actually runs for 2 mins every test... so in total it takes ~ 20mins running 10 tests.

 

[error8]

Originally posted by: spike99

One thing that makes it special, is that you have the highest temperature you can possibly attain with your chip. My E7200 at 4 ghz, for example, with 1,46 V, would hit as high as 86 C with Linpack in a couple of minutes, while Prime95, Orthos and OCCT would only give around 65-70 C, under load.

I agree, I ran Everest Ultimate and at one time had used Prime95 too and my CPU never got past 80C. Running Linpack, my CPU temp reached a little over 90C.

Also, I previously said that it took 1 min per test... I re-checked it yesterday and it actually runs for 2 mins every test... so in total it takes ~ 20mins running 10 tests.

Careful there man. Over 90 C is quite a lot. Maybe you should lower that voltage a bit and reduce your clocks.

 

[imported_spike99]

Careful there man. Over 90 C is quite a lot. Maybe you should lower that voltage a bit and reduce your clocks.

Yes, I was starting to get concerned when I was seeing those temps... I was like... Yikes !!

In reality, running Linpack CPU temp was at 90C for about 90 seconds and then would go down for every test. So the temp was not 90C for the whole test which took 20 mins to run.

Also, when using my PC on a daily basis temp is no where near 90C.... it currently idles at 60C and goes to 70C when gaming. I think this is a bit high but I believe it should be fine. I can go a little lower on my temp but it would translate into a higher fan RPM for CPU cooler which in turn generates a noisier CPU cooler.