Overclock.net banner

13hrs Prime95 stable...apparently isn't stable

3K views 74 replies 17 participants last post by  munaim1 
#1 ·
#2 ·
i had a similar issue when overclocking my 955 BE. I only took it to 3.8Ghz and my NB to i believe it was 2700Mhz. Everything was prime 95 stable for 24 hours. However simple gaming would crash(game would freeze and i would get an error saying the .exe has stopped working) And other display issues would arise untill i did a cold reboot.

Took it back to stock to see if it was the OC and at stock everything run perfect. to this day I havent been able to figure out what it might have been and havent ran into anyone with a similar issue with all the searching and research i have done.
 
#3 ·
I'd say just bump your voltage up one notch and call it a day.
 
#5 ·
Quote:
Originally Posted by Behemoth777;14242330
I'd say just bump your voltage up one notch and call it a day.
yup, rinse and repeat if necessary. I have seen blend fail after 16 hours before.

101 isnt 100% all the time voltage either. Try bumping up vccio if you havent already. IME with sb low vcore results in a 124 bsod almost all of the time, and in a few cases 101.

101 could also mean memory timings are unstable. Have you overclocked your memory? if so try returning memory frequency and timings back to stock and runing x.264 again see what that does.
 
#7 ·
Quote:
Originally Posted by munaim1;14242737
It could even be the os related, bsod countless times while finding your overclock or gpu clock or whatever and you could more than likely corrupt the OS. Just something to think about.
Yup, that is very likely.

op: if you revert back to stock settings for your cpu and no blue screen, then it is most likely your overclock. If you revert back to stock settings and continue to blue screen you may consider re-installing your os.
 
#8 ·
Quote:
Originally Posted by AliceInChains;14242627
yup, rinse and repeat if necessary. I have seen blend fail after 16 hours before.

101 isnt 100% all the time voltage either. Try bumping up vccio if you havent already. IME with sb low vcore results in a 124 bsod almost all of the time, and in a few cases 101.

101 could also mean memory timings are unstable. Have you overclocked your memory? if so try returning memory frequency and timings back to stock and runing x.264 again see what that does.
Memory is set to spec, didn't overclock it.

I bumped it from 4.6 down to 4.5 and I was able to do 2 conversions with no issues, though the monitor flickr still exists. Not all the time, only when I'm starting up a video on itunes, and a few other times I haven't pinpointed. Never while browsing. Horizontal lines.

I forgot to drop the vcore back down for 4.5 though so I'm probably over volting it. Will probably push it back to 4.6 and bump up .01.
 
#11 ·
Quote:
Originally Posted by munaim1;14243160
Just curious what nvidia driver are you using?
275.33
Quote:
Originally Posted by trulsrohk;14243186
I thought it was pretty common knowledge that prime95 isn't the best test for SB...try IBT instead
thumb.gif
Haven't been doing this long enough to have common knowledge. But will also give IBT a try, thx.
 
#14 ·
Quote:


Originally Posted by trulsrohk
View Post

I thought it was pretty common knowledge that prime95 isn't the best test for SB...try IBT instead


I disagree. I think blend is the best for sandy bridge, and also think most of ocn would agree. Not trying to be rude or anything. Just want to provide the op with the best information available. If im wrong, ill admit it.

On the other hand, it is common knowledge that small fft and large fft do nothing for sb. You can pass 12 hours of small fft just to fail after 1 hour of blend
 
#15 ·
@ 4.6ghz, 1.35V my Prime95 temps were around 72 and IBT 76 but would BSOD after the 2nd run. If I were to push it any higher, I'd be in the 80c mark so I decided to drop it down to 4.5ghz @ 1.31 Prime95 stable. After 1 IBT run it BSOD again so I bumped it up to 1.315V. I stepped away for about 90 mins and it completed 20 IBT runs 76C, highest GFlop at 111. These were all run at Maximum level.

I'm hoping this fixes any stability issues.

*edit* and it looks like the voltage bump also fixed the display flicker problem. nice.
 
#18 ·
Quote:
Originally Posted by ocococ;14249616
I'm not arguing one vs the other...but after 12hrs blend, it still failed IBT after 1 run.
Did you use LLC? Most of us who overclock with the ASUS P8P67/P8Z68 series use LLC at Ultra High. If so, Load Volts will be higher for IBT/AVX VS Prime95.

So If you're not using any LLC setting, or a very low one, the opposite could happen, and make IBT/AVX unstable. Even if this was true, your computer would never crash in anything else, as long as you where Blend stable. This is because nothing else, except IBT/AVX would drop Load Vcore lower than Blend.
 
#19 ·
Quote:
Originally Posted by AliceInChains;14243294
I disagree. I think blend is the best for sandy bridge, and also think most of ocn would agree. Not trying to be rude or anything. Just want to provide the op with the best information available. If im wrong, ill admit it.

On the other hand, it is common knowledge that small fft and large fft do nothing for sb. You can pass 12 hours of small fft just to fail after 1 hour of blend
This is just a backwards statement.
Blend consists of small, and large fft...
 
#21 ·
Quote:
Originally Posted by turrican9;14249708
Did you use LLC? Most of us who overclock with the ASUS P8P67/P8Z68 series use LLC at Ultra High. If so, Load Volts will be higher for IBT/AVX VS Prime95.

So If you're not using any LLC setting, or a very low one, the opposite could happen, and make IBT/AVX unstable. Even if this was true, your computer would never crash in anything else, as long as you where Blend stable. This is because nothing else, except IBT/AVX would drop Load Vcore lower than Blend.
LLC is disabled because of the sleep/resume issue. Until they fix that I won't enable LLC, unless there's another fix.

My Blend test was 12+hrs stable, but still crashed when encoding. It also crashed after 1 IBT run.
 
#22 ·
Quote:
Originally Posted by ocococ;14249792
LLC is disabled because of the sleep/resume issue. Until they fix that I won't enable LLC, unless there's another fix.

My Blend test was 12+hrs stable, but still crashed when encoding. It also crashed after 1 IBT run.
Has LLC something to do with the Sleep feature? I thought it was the Internal PLL Overvoltage when enabled that affected Hibernate.

Also, alot of us have to disable C3/C6 to avoid idle/light load BSOD when using Offset Vcore.
 
#23 ·
Quote:
Originally Posted by Precision_PC;14249721
This is just a backwards statement.
Blend consists of small, and large fft...
Blend does indeed consist of small and large fft. However, unlike the small and large fft tests, blend uses 1600mb of ram by default. If you just choose small ffts and click run it will use no ram; if you choose large ffts and click run it will use 8mb of ram. Now if you choose blend and click run it will use 1600mb of ram, making it a better overall stability test than either of the other two.
 
#24 ·
Quote:
Originally Posted by turrican9;14249858
Has LLC something to do with the Sleep feature? I thought it was the Internal PLL Overvoltage when enabled that affected Hibernate.

Also, alot of us have to disable C3/C6 to avoid idle/light load BSOD when using Offset Vcore.
sorry my mistake, i was confusing the two. remind me what LLC is?

any downside to disabling C3/C6?
 
#25 ·
There is a much better free stability program than intel burn with avx which is TOO stressful, and prime 95, which doesnt cause blue screen fast enough,

its called Team Fortress 2 32 man bot match

I tested my 2500k with many programs, including ludicrous 12+ hour prime and 100 pass avx intel burn test, but tf2 causes blue screen in usually less than 2 hours, and without heating up the processor to 70+c.
 
#26 ·
Quote:
Originally Posted by SyveRson;14249881
Blend does indeed consist of small and large fft. However, unlike the small and large fft tests, blend uses 1600mb of ram by default. If you just choose small ffts and click run it will use no ram; if you choose large ffts and click run it will use 8mb of ram. Now if you choose blend and click run it will use 1600mb of ram, making it a better overall stability test than either of the other two.
Nice. Good to know.
So what is it actually doing with the RAM that is being used in the blend test?
What sort of test is running on it?
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top