Full-system stress-testing?

[nullpointerus]

I set Prime95 Blend to run overnight (total: ~16 hr.) w/ zero errors; however, after I saw the green lights were still on in Prime95, I looked over to see an error message that an application had crashed (Gigabyte EasyTune?) at some time during the night...


So... I increase the voltage slightly (+0.01250v), but...
Is there anything that can stress a system more thoroughly than Prime95?


I'd like to run another test today if possible.

Small FFT has been running fine for 3 hr. now...

 

[theprodigalrebel]

Intel's LINPACK

I do know it sends temperatures through the roof. This is from the same machine:

4x Linpack 64 bit - 72C
4x TAT ~Linpack 32bit - 66C
4x Prime Small FFTs - 59C
OCCT cpu test - 55C

I have no idea how to use it for stress testing (i.e. how many hours/iterations before a machine can be considered uber-stable).

 

[nullpointerus]

:Q




With LINPACK, I get a 75C load (67C per core), up from about 63C load (55C per core) from Prime95 Small FFT.

Unfortunately, I believe my chip throttles constantly at those extreme temperatures, so I can't use it for stability testing. (I have the older E4400, the one with 61C max. advisable temp, instead of the Mx version that's supposed to handle 70C load.) Maybe LINPACK will run fine when I get better cooling (i.e. a direct intake for the CPU) and possibly lap the CPU's heatspreader at some point.

For normal usage, my computer seems fine. :thumbsup:


BTW, what do you think of this quote from the thread you linked:

Originally posted by: chew* on XtremeSystems.org Forums

FYI the best way to test stability is a buttload of benches then games then 2 fresh installs of windows.....1 stock 1 oced..........I have a program i use to compare files/folders.......if the OS is the slightest bit different your not stable.......Installing windows OC'ed is the best way to find corruption.......

Windows Setup errors almost always point to hardware instability. chew*'s method is definitely unconventional. I wonder how well this method really works; i.e., how many file copy errors does it actually reveal?

 

[Capitalizt]

TAT > all

 

[VirtualLarry]

Some have said that F@H stresses more than P95 does.

 

[Amaroque]

Originally posted by: VirtualLarry
Some have said that F@H stresses more than P95 does.

IIRC, It does stress RAM a lot more. I don't know for sure, as I run DPAD as a DC project.

 

[Markfw]

I can say from personal experience, that it you want the most stable system for all applications, F@H for a week straight, and no EUE's ? and you are good to go. EUE means it abended, usually to a slight too much of an OC. I may pass orthos for a day, and then 2 days later under F@H, you get an EUE.

 

[nullpointerus]

Originally posted by: Markfw900
I can say from personal experience, that it you want the most stable system for all applications, F@H for a week straight, and no EUE's ? and you are good to go. EUE means it abended, usually to a slight too much of an OC. I may pass orthos for a day, and then 2 days later under F@H, you get an EUE.

Excellent!

I've got this running already. :thumbsup:

 

[CP5670]

Originally posted by: theprodigalrebel
Intel's LINPACK

I do know it sends temperatures through the roof. This is from the same machine:

4x Linpack 64 bit - 72C
4x TAT ~Linpack 32bit - 66C
4x Prime Small FFTs - 59C
OCCT cpu test - 55C

I have no idea how to use it for stress testing (i.e. how many hours/iterations before a machine can be considered uber-stable).

I'll need to try this out. This is actually a real world program in some sense (a newer version of it called Lapack is incorporated in Matlab and some other computational software), so it may be more relevant than something like Orthos.

 

[lopri]

I would try to run as many apps that stress various parts of the system. There are many apps to stress CPU/RAM, but not many for disk controllers or disks themselves, or LAN. Brings me a question - What would be a good application to stress disk controller or disks? Something like.. an app that constantly fragments HDDs and then defragments by moving various sizes/types of files?

 

[Markfw]

Originally posted by: lopri
I would try to run as many apps that stress various parts of the system. There are many apps to stress CPU/RAM, but not many for disk controllers or disks themselves, or LAN. Brings me a question - What would be a good application to stress disk controller or disks? Something like.. an app that constantly fragments HDDs and then defragments by moving various sizes/types of files?

My recomendation was based on the fact that I run many apps, encoding, games, etc, an all will run fine for hours, but when you let something take over 24 hours a day for a week, and it performs flawlessly (IO is part of F@H), then you really know you truly have a stable system.

 

[PolymerTim]

That's it! We should start recommending F@H as the new defacto stress test for breaking in new builds and overclocks. What do you guys should be the recommended stress time to be ensure absolute stability; 3 months; 6 months? Better yet, maybe it should be left continually running for the life of the computer to ensure that no parts have degenerated at some point and no longer allow for absolute stability. It's ingenious!

Just having some fun. I'm hoping to start up F@H myself when I get my new system in a few months.

 

[betasub]

Little documentation on it, but SNM Stress Test app is good for working out individual core components.

 

[nerp]

I think F@H is a good tool for stress testing. Generally, when I'm testing a system for stability, I leave Orthos running for a couple days as I use the machine normally and try to put it through its paces. If it passes all that, I'll run F@H for a week wile I use the machine normally. If that goes well, I just use the machine as if everything is perfect and keep my eyes peeled for any problems.

 

[Amaroque]

Atually almost any DC program is great for stress testing. Back in '02 I was on a DC project called G@H, related to folding, but anyway...

I would find that with too high an OC the machines would freeze after a week or two. It only took like 50MHz difference to make the machines stable. That's not a big difference!

The point is that you can't truely find an unstable system with only 24hrs of loading.

 

[nullpointerus]

F@H seems to run fine on my PC, with two exceptions:

1. When I schedule multiple Media Center recordings, they go on forever and MC doesn't respond. I think I might have fixed this by setting the maximum F@H CPU usage to something like 90% instead of 100%. Video playback is better at this setting, too.

2. Since I installed an 8800GTS and enabled multi-monitor support, the F@H client keeps popping its own window up and minimizing my games. Like, once every second. Is there any way to configure F@H not to create its 3D window? I guess I could run the command-line version of F@H...if I must.

 

[BlueAcolyte]

Hmm... I get a BSOD whenever I use my system during Orthos, because of the 100% CPU load. Is it because mine is overclocked?

E4500 2.93GHz 1.28-1.325V depending on load.

 

[VirtualLarry]

Originally posted by: BlueAcolyte
Hmm... I get a BSOD whenever I use my system during Orthos, because of the 100% CPU load. Is it because mine is overclocked?

E4500 2.93GHz 1.28-1.325V depending on load.

Your system is unstable. You probably need more vcore.

 

[nerp]

Probably. You shouldnt' BSOD with Orthos running. I can play games and stream music across my network while Orthos is running.

 

[superstition]

From what I've seen, Orthos doesn't really shine unless it's being run by itself. Even having something like Speedfan running at the same time makes it less effective.

 

[n7]

Originally posted by: theprodigalrebel
Intel's LINPACK

I do know it sends temperatures through the roof. This is from the same machine:

4x Linpack 64 bit - 72C
4x TAT ~Linpack 32bit - 66C
4x Prime Small FFTs - 59C
OCCT cpu test - 55C

I have no idea how to use it for stress testing (i.e. how many hours/iterations before a machine can be considered uber-stable).

Indeed.