SPI 32m vs P95 stabilty

[freecableguy]

Lately I am not as trusting of P95 as before (maybe its the recent version upgrade that's buggy?). I can run SPI 32m ALL DAY LONG one after another and never have any problems but I can get errors in P95 anywhere from 1-12+ hours...very random. I can increase volts to get rid of them but right now I am attempting a burn-in session to get that as low as possible. My primary means of testing stability has been SPI 32m NOT Prime95. I was wondering what others thoughts on this was.

-FCG

[sirkuz]

Well...I think that Spi32M is a much better memory diagnostic, and Prime (small ffts) is a much better cpu diagnostic. If you can increase your voltage to reduce errors why would you think there is a bug of any sort? I would think prime is doing a great job on your system diagnosing errors for total stability. If your not shooting for complete stability maxing your CPU for days or weeks at a time there there is no need to worry about prime failing in a 12hr window.

Thats just my take on it tho. To each his/her own :D

[Nossie]

i think OCCT is best, Where prime fails in 6 hours, OCCT fails in 6 minutes.

for me, 30minutes OCCT and 1 hour BF2 = computer 100% stable.

[freecableguy]

passes OCCT forever just like SPI 32m....P95 is the only one that has any problems

AFAI can tell, the system is 100% stable.

[Absolute_0]

SP32M is a pretty poor indicator of stability to me. I can run a 32M up where SP2004 still errors and games still crash.

[don_vercetti]

i wouldnt trust superpi32m either. It's good for RAM, but so are large FFT's in p95.

[Zeus]

I have no idea what to trust and what not.
I have the feeling Prime is the most picky of them all, closely followed by OCCT.
SuperPi 32M will run a lot easier imho but is still a nice indicator though.

If it can do 32M and OCCT test it's stable enough for me, all night long Prime gives just a little more confidence i must admit.

About large or small ffts, small ran a without errors on my system, where large ffts often failed within an hour.

Could it be large ffts is stressing the CPU's cache a bit more?

[angra]

in my view, a system that fails either test is not stable.

each test will stress different setups differently. in some cases test A will be the limiting determinant, and in some cases, test B - however if either detects a fault, the system will not produce deterministic, stable results, and is therefore eligible for crashes and corruption.

All of that said, 100% rock-solid stability is not necessarily what every XSer is after, so expectations should be adjusted to match the requirements.

[bachus_anonym]

SuperPi 32M has never been an indicator of an ultimate stability. It just one of those benchmarks that stress CPU/Memory subsystem greatly but not as hard as P95.
Those that played some with both would know that If your "absolute max" P95 Blend or LargeFFT stable limit (for, say 2hrs) is "X" then 32M will be somewhere at "X+some more".
As far as Memory stability tool, best would still be "Memtest for Windows" as it is VERY sensitive to even slightest changes in timings like TRCD, TREF, TWTR or parameters like MAL, Read Preamble or Drive Strengts. Being able to complete 500-1000% of that test without errors at highly overclocked memory clock is quite a challenge ;)

[pduan87]

P95 for 24 hours is the way to go for me. I like rock stability.

[J-Mag]

I like 32m SPi the best because it is a quick and dirty method. I try to push the max for 32m and once I find max 32m, then I usually just back down 100mhz on the proc and I know that my system will be pretty damn stable. Finding the maximum stable overclock (even rounded to the nearest 10mhz) is a myth IMO anyway. One day your rig is 2.8ghz "stable", the next it isn't. For high resolution gaming it isn't that big a deal anyway 100mhz won't make or break ya.

[cantankerous]

I can memtest in windows and does forever, SuperPI 32M one after another with zero issues however this weekend I had prime95 fail after 11 hours and 18mins. Didn't run infininately but 11+ hours is pretty good I think. I ran OCCT for 9h 45m before I got bored and stopped it however I was having 3dmark01 fail after looping about 2 hours in. Drove me nuts. Upped voltage a bit on my ram and it ran for 9.5 hours straight last night. I guess im pretty good now though after all this hassle I truly think I am at my cpus limit and unfortunately that is at an early 2.48ghz.

[szukalski]

I like bf2 stability. Way more fun than Prime95 or Pi32M.
Nothing crashes my system faster.
I can run Pi32M over and over without fail, run Prime95 for 6hrs and it fails there, but in bf2 it will fail in 20 mins or my money back.

[Plague_oc]

in my view, a system that fails either test is not stable.

each test will stress different setups differently. in some cases test A will be the limiting determinant, and in some cases, test B - however if either detects a fault, the system will not produce deterministic, stable results, and is therefore eligible for crashes and corruption.

All of that said, 100% rock-solid stability is not necessarily what every XSer is after, so expectations should be adjusted to match the requirements.
^^ totaly agree