How stable is stable to YOU?

M

morfinx

[H]ard|Gawd
Joined
Feb 18, 2005
Messages
1,105
I've been overclocking and doing stability testing on my video card for the last few days. First thing I did was I let ATI Tray Tools (ATT) find my max GPU/MEM speeds, and that seems to be a good starting point. I was able to loop 3DMark05 and 3DMark03 five times each, run FEAR for about an hour with no problems, and loop SpeedTree demo (foliage engine used in Oblivion) for over two hours without incident. However if I try to loop SpeedTree overnight, it locks up when I get up in the morning.

What I have been doing is firing up the ATT Artifact Tester, wait an extended period of time and see if it detects any artifacts. After the GPU is well heated up, ATT did detect artifacts at the max speeds that was previously automatically detected. So when ATT reports artifact, I adjust GPU/MEM down by 5Mhz, and repeat the process until it runs for a looong time without artifacts.

So my question for y'all is, how stable is stable to YOU? Is it enough to be able to run FEAR for a couple hours without problems? Or does stable mean zero artifacts for many hours at a time? Or is it something inbetween, i.e. able to run the Artifact Tester for a specific length of time without artifacts?
 
If it doesn't Prime overnight, I usually won't keep the speed....

Otherwise, though, if I can play my games without issues, that's usually stable enough for me. You have to figure, the abuse most of these benchmarks, prime, etc, are a TON more than you'll ever put your PC through in normal use. If it will run through that many loops of 3dmark, etc, it should be fine for normal use.

I just got into the Prime addiction, though - so now my views on my own PC have changed. In my NEW opinion, if it won't run Prime, etc overnight, it's not 100% perfect, which isn't good enough for me anymore.

Hope that helps, and doesn't confuse your thoughts any more! :D
 
CPU/Memory:

100% error free for 6+ hours at least. I won't settle for any less.

GPU core/mem:

If I don't notice any artifacts in games, that's enough for me.


Generally speaking, overclocking the graphics card won't cause lockups or crashes if eorrs occur, you'll just get visual artifacts. So if I don't notice any, that's good enough. I don't care if ATI tool or whatever says there's technically artifacts, if I don't see any that's all that counts.

Though, ironically, ATI tool will actually claim a higher overclock than I'll find by 'manual' testing and so I'll have to back down a bit from what it suggests when/if I see any artifacts in games.
 
atitool's artifact testing doesnt even work properly on the x1xxx series... either that or the cards artifact so bad you cant see anything(the whole cube is covered in yellow) at stock speed.. i use rthdribl at settings and size that takes the card down to 20 fps... artifacts show up as giant green dots
 
ryuji said:
atitool's artifact testing doesnt even work properly on the x1xxx series... either that or the cards artifact so bad you cant see anything(the whole cube is covered in yellow) at stock speed.. i use rthdribl at settings and size that takes the card down to 20 fps... artifacts show up as giant green dots
Click to expand...
Finding artifacts appears to be different for everyone. There's no single application that'll find problems the best. You just have to discover what apps/games expose problems the most/fastest for your card.

For me, BF2 will be the most revealing.
 
spine said:
Finding artifacts appears to be different for everyone. There's no single application that'll find problems the best. You just have to discover what apps/games expose problems the most/fastest for your card.

For me, BF2 will be the most revealing.
Click to expand...
of course, but if your doing 5 mhz at a time... bf2 is a little tedious... even at 300 fsb bf2 takes about 15 secs to load a new map and about 25 secs to load first time... and you also have to wait like 5 minutes for the account server to log you in and as far as stress... if you tweak rthdribl to max quality, use skull and also resize ... you can stress even the fastest cards down to 20 fps, which in my book is probably the most stressfull and quickest way to do artifact testing... as well as the damn huge ass green dots that artifacts are
 
ryuji said:
of course, but if your doing 5 mhz at a time... bf2 is a little tedious... even at 300 fsb bf2 takes about 15 secs to load a new map and about 25 secs to load first time... and you also have to wait like 5 minutes for the account server to log you in and as far as stress... if you tweak rthdribl to max quality, use skull and also resize ... you can stress even the fastest cards down to 20 fps, which in my book is probably the most stressfull and quickest way to do artifact testing... as well as the damn huge ass green dots that artifacts are
Click to expand...
You never heard of alt-tabbing then? ;)

rthdribl is good for gpu testing but, like ATI tool, it barely touches on all your card's memory so it'll give inflated overclock estimates. This is why I use BF2 which'll suck up all the card's mem easily and so with all the mem in use, you'll quickly find the weakest link and have to lower your overclock.

I'd start with ATI Tool and leave it going for ages, then simply take what it gives as a starting point. I mean, you don't expect to find the best overclock within a few minutes, it's more of a long term thing you just keep on hand as you play games and generally use your graphics card.
 
24 hours prime for cpu/mem. gpu = playing an graphic instensive game for around 4 hours (BF2, F.E.A.R)
 
this is probably really dump question but ati tool doesn't work on nvida cards right..

is there an nvidia clock speed auto detect / artifact finder program ...

i have a pny 6200 unlocked to all 8 and under water ..

cheers,

mal


p.s. also if someone could point me towards a step by step bios edit to unlock the temp sensor that would be great too.. thnx
 
ati tool is able to detect artifacts on nvidia cards, but it cant change clock speeds.
 
at least 30 minutes of prime. Then i run it 24/7, if it is on good behavior, then it is stable. I have run at speeds where prime fails in less than 10 minutes and never seen a problem with anything though.

USUALLY howerver, i test prime to about 2 hours. Then nothing else fails, ever.
 
Couple hours of prime with rthdribl running just to recreate the heat from the graphics cards and the current draw that the system will actually see.

But I am also a rabid memtester before I do any of that. I don't boot into windows with any new settings until it passes a couple runs of memtest86.

As far as graphics cards go, I hit auto detect.....back it down a couple megahertz from that an wallah, graphics cards are overclocked. i.e my 7800GTX's both autodetect @ 500/1320 and I run em at 490/1300 which is a lot over stock but I still feel safe and like they will last a good long time.
 
If my system can survive my encoding habits over a few nights, the OC stays. (100% CPU load of pure math for 12-14 hours).
 
spine said:
You never heard of alt-tabbing then? ;)

rthdribl is good for gpu testing but, like ATI tool, it barely touches on all your card's memory so it'll give inflated overclock estimates. This is why I use BF2 which'll suck up all the card's mem easily and so with all the mem in use, you'll quickly find the weakest link and have to lower your overclock.

I'd start with ATI Tool and leave it going for ages, then simply take what it gives as a starting point. I mean, you don't expect to find the best overclock within a few minutes, it's more of a long term thing you just keep on hand as you play games and generally use your graphics card.
Click to expand...
if you set rthdribl up right and resize it, you can actualy run out of video mem and it crashes out, yet it doesnt touch the ram? :confused:
 
ryuji said:
if you set rthdribl up right and resize it, you can actualy run out of video mem and it crashes out, yet it doesnt touch the ram? :confused:
Click to expand...
Yeah, I've done that in the past. With 6x AA and multi monitors you can stretch it across both screens in a high resolution and you'll find apoint where it uses all video mem. You have to use ATI Tool's OSD for 3D for this so you can tell when you're nearing the video ram limit as you keep increasing the window size.
 
i have had the program tell me it ran out of video ram and it quits... havent been able to do such on the x1800 tho but i have on the 6800gt, my 6600, and i belive also the x850 on the cards it did it on it did it at about 1000x1000 not dual monitors worth of res...
 
When I overclocked my system I let it run Prime95 overnight, it didn't get any errors and I was happy with that. If I was a folder I would have let it run much longer (at least 24 hours), but for what I do I thought 12 hours was enough
 
my problem isn't with how stable is stable it's how hot is too hot

i'm not comfortable with anything in the 50s ... i.e. <50 is mandatory for me

running an xp-m 1.35v 2400+ at 2.5ghz at 1.65vcore and it's primeing around 41-42 on water ...

mal

http://www.hardforum.com/showthread.php?t=1030883
pc1ho.jpg
 
Temps generally don't bother me because the are usually so wildy inaccurate.

My comp right now will peak at 51°C-52°C under Prime95 load, but I watercool and I know it's not anywhere near actually that hot. If I use the latest bios for my motherboard, the temps will drop 6°C because DFI tweaked the temp reading offset.

As long as it doesn't crash, I don't care. Though, having said that, I'd never let my PC get to or over 60°C 'cos that's just insane.
 
You must log in or register to reply here.
Back
Top