Artifacts testing != Stability testing?

[gparent]

Hey, I recently overclocked an eVGA 8800GTS 640mb paired with an Accelero XTreme. My temps are typically 51 idle (I keep the fan low, and my room's getting hot) and 62 load.

I raised the clocks from stock to 702 / 1620 / 1004.

These clocks run fine on ATITool for 30 minutes (And probably more), however, I tried playing TF2 the other day and I crashed within a few minutes.

Am I just incredibly unlucky or does the fact that there are no artifacts have no relation to the stability of my overclock?

[Iron Monkey]

ATITool is a poor benchmark for checking the stability of an oc'd card.

[johnerz]

Originally Posted by Iron Monkey

ATITool is a poor benchmark for checking the stability of an oc'd card.

It could however also be down to a unstable cpu overclock crashing out on the game

Have you tried running memtest and prime to check for system stability?

[gparent]

Originally Posted by johnerz

It could however also be down to a unstable cpu overclock crashing out on the game

Have you tried running memtest and prime to check for system stability?

Everything else is stock and hasn't crashed in months.

ATITool is a poor benchmark for checking the stability of an oc'd card.

Any suggestions? I can't really sit there and watch the Crysis benchmark for artifacts 30 minutes.

[Sephious]

Originally Posted by gparent

Everything else is stock and hasn't crashed in months.



Any suggestions? I can't really sit there and watch the Crysis benchmark for artifacts 30 minutes.

The Crysis Benchmark really isn't that great either. World of Warcraft crashes faster than Crysis from what I've found.

I really haven't found a good videocard stability test. ATITool used to be good, back when videocard stability and artifacting were one in the same, but that's just not the case anymore.

[thread]

There unfortunately isn't a really good alternative to game time for testing the graphics overclocks. Mine worked fine in many things, but I got artefacting in crysis after a few minutes, just had to bump down the memory clocks.

So if you find things crashing:
1) try the game again at stock settings - if it still crashes, then there's something else wrong
2) After that You just have to play to find what works.

So, don't feel you need to watch the crysis benchmark for 30seconds - you may as well play crysis and enjoy the time. (or whatever game you need to know the limits for)

[gparent]

So I guess I'll keep playing TF2 and lowering clock speeds until it works?

I assume crashes are more often related to Core speed than RAM/Shaders?

[Iron Monkey]

I find the best test for a GFX card is to just play your favorite games and check for Artifacts, crashes etc. After all that is what you want to be playing. 3dmark is alright also imo

[KopyKat]

Using games is probably the best way if thats what you'll be doing.

Also i used this program called FurMark and it really heats up the card like crazy,but id be careful with that one since ive heard of people blowing their cards with it so keep an eye on it at all times when running the stability test.

Not sure about how it will find artifacting though, or if its a better stability tester than games are. It does have that stability test option and heats up that card more than any games do including the Crysis benchmark program but im not sure that heating it up is all you need to do to check for stability.

Anyone know the answer to this while we are at it???

[gparent]

Originally Posted by KopyKat

Also i used this program called FurMark and it really heats up the card like crazy,but id be careful with that one since ive heard of people blowing their cards with it so keep an eye on it at all times when running the stability test.

I'll check it out. No worries with the blowing up, I've got very appropriate cooling. The Accelero is amazing.

[Th3_uN1Qu3]

Try RTHDRIBL or 3DMark06. 3DMark is especially sensitive to gfx overclocks.

[Sephious]

Originally Posted by Th3_uN1Qu3

Try RTHDRIBL or 3DMark06. 3DMark is especially sensitive to gfx overclocks.

I've found exactly the opposite. I can run for days at incredibly unstable graphics card clocks in both of those applications.

[KopyKat]

Originally Posted by Sephious

I've found exactly the opposite. I can run for days at incredibly unstable graphics card clocks in both of those applications.

Have you tried FurMark yet? Could you give us a bit of insight on how it affects your card by trying it if you haven't yet, and see if it catches those unstable settings?

Thanks

[Th3_uN1Qu3]

Originally Posted by Sephious

I've found exactly the opposite. I can run for days at incredibly unstable graphics card clocks in both of those applications.

I can run RTHDRIBL for ages at unstable clocks but not 3DMark06. Anything 06 stable is gaming stable, at least that's how it always is for me.

[gunbo13]

Graphics card stability testing is all about doing a "suite" of tests. My advice? Benchmark testing from your top three most demanding titles along with 3DMARK and other benchmark only programs.

If it passes a "suite" of tests you are good to go. Single software testing is unreliable in the world of GPUs.

[KopyKat]

Originally Posted by Th3_uN1Qu3

I can run RTHDRIBL for ages at unstable clocks but not 3DMark06. Anything 06 stable is gaming stable, at least that's how it always is for me.

Nice, im gonna try that one out.

Also, can you set your resolutions and all that? Is it best to use the settings you normally use is a powerful game like say Crysis or is it more effective if you crank everything to the max including AA and who cares if the frames go down alot? Does "cranking" everything up heat it more or help with stability testing or does it not matter what its set at?

Or to take it a step further, if the cranking it up thing actually makes a big difference, is it possible to lose out on highest OC potential because of setting it based on such high settings that i will never use in a game like Crysis? Im mostly referring to Antialiasing i guess (AA).

Sorry if that sounded confusing but im a pure noob so i just want to know anything i can.

Thanks

Additional Comment:

Originally Posted by gunbo13

Graphics card stability testing is all about doing a "suite" of tests. My advice? Benchmark testing from your top three most demanding titles along with 3DMARK and other benchmark only programs.

If it passes a "suite" of tests you are good to go. Single software testing is unreliable in the world of GPUs.

Makes sense to me.

Please see my above post and tell me what you think.

[gunbo13]

Originally Posted by KopyKat

Makes sense to me.

Please see my above post and tell me what you think.

Post the list of your latest modern PC game titles and your benchmark progs. available.

I know you listed some but a complete list would be great. Then I'm sure we can put together a "suite" of tests for you. Have them pass and you got stability.

For example my list would be:
1. Oblivion IV
2. Gears of War
3. World in Conflict
4. Crysis
5. Call of Duty 4
6. HL2 EP2

a. 3DMARK06
b. 3DMARK Vantage
c. The rest :P

Honestly I haven't done graphics testing so trust other people on more benchmark progs. My games list covers all modern engines however so I advise I similar approach.

[KopyKat]

Ok, these are my newer ones id say.

1. Crysis
2. Call of Duty 4
3. Command & Conquer 3
4. Bioshock
5. Half Life 2 EP1
6. Ghost Recon 2
7. Rainbow 6 Vegas 2
8. FEAR

A. 3dMark06
B. FurMark
C. rthdribl
D. Crysis Benchmark Tool
E. Prime 95 but thats more for CPU i think

I also have Riva Tuner, EVGA Precision and ATI Tool for OCing the card and a bunch of different temp monitoring programs.

Thanks again!

[gparent]

I havn't had time to test it yet using FurMark (or other games, for that matter), but I will post updates ASAP.

[Th3_uN1Qu3]

Originally Posted by gparent

I havn't had time to test it yet using FurMark (or other games, for that matter), but I will post updates ASAP.

Ah yeah, FurMark is a really tough test. Heats up my 3870 like nothing else.