[xd_1771]

I've had my awesomesauce GDDR3 memory on my GTS 250 clocked up to 1333mhz (from 1120mhz) with no problems at all; lately I have noticed that the stress test tool I use (GPUTool, an improved variant of ATITool) only uses 107MB of the 512MB on board my GPU [Checked with GPU-Z]; is there a better stress tool for memory that [preferrably] has automatic artifact detection? I want to see if I can get up to 1400mhz on my GDDR3 chips!

Unless of course, that is unpractical and unsafe.
I also noticed when I upped my memory clock from 1120 to 1300, my framerates in the GPUTool stress utility increase from ~86 to 96 FPS. at 1333mhz, those framerates increase to slightly over 100.
Wow, I never thought this could get any better

 

[Juggalo23451]

Furmark??
http://www.geeks3d.com/20090703/furm...ities-is-back/

 

[xd_1771]

Quote:

Originally Posted by Juggalo23451 Furmark?? http://www.geeks3d.com/20090703/furm...ities-is-back/

That sounds fair, in fact I've used it sometimes but I find that the fact that you have to detect artifacts that aren't even highlighted with your own eyes makes it a bit tough to tell whether my card is stable or not.

 

[sn0man]

EVGA is working on a utility that sounds exactly like what you want. They say it should provide 'universal' GPU support. Furmark / MSI Kombustor are great for stress testing, but I prefer something that artifact scans as well.

http://www.evga.com/forums/tm.aspx?&m=117563&mpage=1

 

[theCanadian]

Try memtestg80 http://folding.stanford.edu/English/DownloadUtils#ntoc2

 

[xd_1771]

Quote:

Originally Posted by theCanadian Try memtestg80 http://folding.stanford.edu/English/DownloadUtils#ntoc2

Darn, I get a "cudart.dll missing" error trying to run this... sucks because I thought this was the answer to all my problems

EDIT: NVM got it working. How many test iterations should it run and how do I get it to use all my video card memory?
EDIT: This tool is great. After figuring it out and running a test, I find that I have 32 errors after around 8 minutes of testing at 1339mhz using 448MB out of all my 512MB memory; not to mention, this seems to be a good GPU testing tool as well as the GPU core load may vary between 91 and 99 but it does stay above 91 unlike GPUTool and ATITool, though what sucks is that it doesn't report artifact errors. Thanks!

 

[theCanadian]

Run it again at stock clocks using the same amount of memory. You might have just used too much memory. It will always give you an error if you use all the memory, regardless of whether or not it's stable.

 

[xd_1771]

Quote:

Originally Posted by theCanadian Run it again at stock clocks using the same amount of memory. You might have just used too much memory. It will always give you an error if you use all the memory, regardless of whether or not it's stable.

You mean calculation errors or an error message? I got calculation errors; thanks for the tip anyway, I'll try it at stock but I'd rather my whole RAM be tested for stability. How much RAM do you think I should test anyway? 256MB?

 

[MADMAX22]

OCCT has a gpu memory tester in it as well.

 

[xd_1771]

Quote:

Originally Posted by MADMAX22 OCCT has a gpu memory tester in it as well.

Sounds good, I'll check that out. Is it a setting in Linpack?

 

[theCanadian]

Quote:

Originally Posted by xd_1771 You mean calculation errors or an error message? I got calculation errors; thanks for the tip anyway, I'll try it at stock but I'd rather my whole RAM be tested for stability. How much RAM do you think I should test anyway? 256MB?

What I meant was, run the same test, using the same amount of memory, at stock clocks. See what happens. It's just a confirmation that it's your OC and not a limitation of the program. It's a very rudimentary program.