GPU Stress Testing Software?

[Uppercut]

What else can i use besides 3DMark and FurMark?
Also, i hear FurMark can be rather intensive.
What is the recommended amount of time that one should run it?
Thank you.

 

[stevedave]

I run Furmark and Prime95 together overnight to test my setups.

I ran Furmark by itself for 8 and had no system crash, then ran prime for 8 with no system crash. However it would still crash every no and then while gaming so I decided to run both at the same time and boom found out I wasn't stable.

I haven't found a need for anything other than Furmark for GPU stress testing so I haven't looked much.

 

[Bahamut]

Do you play rather intensive games? Does the system crash, games lock up, does the entire system freeze or die at random times when you're playing those games?

No? Then what's the problem?

FurMark as mentioned is a great GPU "heater" these days since you can just loop it forever if you're so inclined. Realize that applications designed to purposely stress a component - the GPU in this case - are going to put that component through far more torture than most people would ever put on it themselves.

I've heard reports from people that say FurMark brought their monster machines to a grinding halt in minutes but they were able to play something like (dare I say it) Crysis for hours upon hours at insane resolutions (like 1920x1200 or higher) and the machine didn't hiccup once.

The example above by that poster of running one for 8 hours (?!?!?!) with no negative results, then running the other stress test for another 8 hours (?!?!?!?!?!?!) wasn't good enough. Not even noticing the system was "stable" through 16 hours (!@?*!#!*^) of "testing" wasn't good enough - had to hit it with BOTH apps at the same time and oh, yeah, there... see, it's not stable. Well, hell man, a human being can do a triathlon just fine, but ask that person to do two triathlons in a row (since they can't do them at the same time) and they'll fail too, just like your previously "stable" system that got murdered by overstressing the components.

Computers can and do "crash" and it's not necessarily because of stress or heat or other issues: software is code, and code fails at times for a variety of reasons. Benchmarking and stress testing software itself can be the cause of most problems, and give false info. 16 hours "stable" and that's not good enough... good lord.

I personally believe people waste entirely too much time on "stress testing" and would probably be better off just playing the games they've spent so much money to play.

Just my opinion, of course, but...

 

[bigdogchris]

I think this program because you can open multiple instances of it, turn on AA on every one and max your gpu out. I usually run 6-8 instances at once.

 

[bexamous]

Computers can and do "crash" and it's not necessarily because of stress or heat or other issues: software is code, and code fails at times for a variety of reasons. Benchmarking and stress testing software itself can be the cause of most problems, and give false info.

That last comment is pretty ridiculous. All the common stress tests, prime95/linpack/furmark/3dmark have been being used long enough to say with pretty good certainty that any system crash is going to be a hardware at fault. To suggest the test app is at fault is simply ridiculous.


Anyways for OP, Furmark & 3dmark are popular for a reason. If you can run 8 hours that says something, if you can run 24 hours that says a little more. I'd say go for 24 hours and see what happens. After 8 hours if its still going maybe thats good enough for you, if it fails at 20 hours obviously your system isn't stable, but maybe you don't care.

Personally I just do stuff for 24hours and call it a day. Actually usually I do 24 hours at some speed, it passes, then I drop a little in speed. But I'm a little anal about stability.

Really though, there is no way to prove anything with absolute certainty. What we're really thinking about is probability of failure. Eg what we wish we could say is "This system ran prime95 for 48hours, the CPU has a 99.997% chance of not producing an error in the next 90 days."

Realistically the best we can do is say "If your system can run prime95/furmark/3dmark for 24 hours, it is pretty unlikely the system will 'fail' in near future when running at a typical lower stress level."

What is 'good enough' is debatable. Again 8 or 24 hours are often used as they're nice numbers. Do whatever you feel is good enough for you.


Oh yeah BTW in general GPU can probably be test much less than CPU. A single CPU error can create big issues, eg crash program/system. A single GPU error will most likely go unnoticed. You need many GPU errors to occur to notice them or have a good chance of acutally hanging system or app. With that in mind, 24 hours for prime95/linpack, 8 hours for furmark/3dmark might be reasonable for most.

 

[Uppercut]

I personally believe people waste entirely too much time on "stress testing" and would probably be better off just playing the games they've spent so much money to play.

I would do this if i had some free time. Unfortunately, gaming isn't on the schedule for this month, so a quick stress test will have to do for now.

 

[Mr. Pedantic]

rthdribl is a good app. And it looks pretty too.

 

[Spazilton]

I think this program because you can open multiple instances of it, turn on AA on every one and max your gpu out. I usually run 6-8 instances at once.

This..

I wonder if the guy realized that his college project would become a stress testing program that has lasted almost 8 years lol.

Think I started using this in 2003

 

[Jon55]

What else can i use besides 3DMark and FurMark?
Also, i hear FurMark can be rather intensive.
What is the recommended amount of time that one should run it?
Thank you.

Why? Is your machine having potential hardware issues? Why beat up your system like that for no reason?