Passed Prime95 but not OCCT???!

[Hvale2k5]

I know its only 4hrs of *Blend*prime95 but the pc was fine, didn't shut down or anything.

After prime95, i want to test it again /w occt 3.1 and it would restart after 10mins in....

 

[cadaveca]

Well, it's very simple. You never passed prime.


If you didn't run it until each FFT size has been run, and it starts back over again, the testing really isn't done yet. Takes about 9.5 hours @ 3.6ghz, on AMD quads.

 

[twilyth]

Cadaveca makes an excellent point about P95 and it's something I never realized either until he turned me on to it in one of my own "problem" threads.

This may not be relevant depending upon your temps, but OCCT linpack is much more intense than P95 and will generate higher max temps. However the most intense utility I've found thus far is IBT (Intel Burn Test, and another hat tip to Cada).

IBT not only gives you the highest temps, but it does not cycle to the same degree as OCCT - which goes from about 17% for long intervals before returning to 100%. IBT will give you range of temps but it is a much narrower range with the low end being about the same as the max you get from OCCT.

 

[LAN_deRf_HA]

What is the standard OCCT test made of? I just know in an hour with large data sets it consistently finds issues IBT doesn't.

 

[overclocking101]

IBT is not as accurate as people lead it to be. i have ran IBT/Lin-x for 20 passes then just 5-10 min after siccess crashing during web browsing, movies, or games. honestly i only use stress tests now to find my maximum temps because i have noticed more and more you can pass stress tests and then crash during normal use.

 

[cadaveca]

IBT is not as accurate as people lead it to be. i have ran IBT/Lin-x for 20 passes then just 5-10 min after siccess crashing during web browsing, movies, or games. honestly i only use stress tests now to find my maximum temps because i have noticed more and more you can pass stress tests and then crash during normal use.

Then you too, are not testing fully. For me, IBT must use all available ram. ALL. 20 passes at proper settings will take about 2.25 hours...I run IBT for a minimum of 12 hours.

That means, no, you cannot just start it with max settings...you'll probably only get 93% ram usage. You need to start a max test...let it load teh ram up, maybe finish an iteration, and then stop it...wait for ram to free itself back up...and then start the test again. Maybe you gotta do this once...maybe you gotta restart 5 times...maybe the 3rd time you get lower mem available, and can only get more by a reboot...


The way I look at/approach stability testing has evolved over the years, as cpus have evolved too, and one of the biggest changes as of late with cpus, is IMC's, and much larger caches.


But these tools...P95, OCCT, IBT...have been around for years(I think IBT is newest). How could they be designed to test today's platforms, when they weren't available back then?


So, anyway, these tools were really designed by a single person, kinda, who developed the program on a specific platform. So, for me, I think that way...this tool is only designed for THAT specific platform.


So, if I do not have the same platform, I have to use custom settings to make it a match for my platform.


And I don't use ONE tool. I use 'em all. If anything fails, I do not consider my system stable, even if I know that I'll never see a crash during my "normal" usage.


Probably 95% of people I've explained this too say:

"WHAT! That takes too long...I don't have time for that".


Real stability testing isn't benchmarketing. It takes time...and lots of it. It's not short, it's not fun, and it's a big waste of electrical power.

But for me...that's what it takes.

 

[sneekypeet]

Sorry, but linpack testing goes back to the 70's, I think IT is the older of the group

 

[cadaveca]

Sorry, but linpack testing goes back to the 70's, I think IT is the older of the group

Yeah, just the IBT GUI that lets you customize how the test is run is relatively new.

 

[sneekypeet]

agreed, but it does utilize Linpack correct?

 

[cadaveca]

agreed, but it does utilize Linpack correct?

I don't know.


Actually, yeah, pretty sure it does. But I do think it is a relatively new Linpack version...I'm using the May 2010 IBT, but there's no version info on the linpack packages.

 

[overclocking101]

yeah i do a shitload of passes maximum ram usasge, and i do the "xtreme burning mode" its far more stressfull then regular ibt. but it there are times when it just hardlocks or bsod after successfull tests. idk why. i usually add vcore/vtt or lower clocks and go agin.