Stable or not stable ?? Need opinion

Since I managed to OC by bios on the g1s i been testing and testing to make the clocks definitely. I saw people with 660 /860 , 600/900, 625/830.. (GPU/MEM)

Now let's talk about GPU/SHADER and forget the MEM.

My only problem is that I and probably everybody that OC , can run a 3dmark without problems since the benchmark last what, 8 min ?? and 2 or 3min of them are on the cpu! And the test is between the SM2 and SM3 benchmarks so the GPU has time to cool down a bit, so you have more or less 5min at top on the GPU.

My consern here is 100% stable. I do not need higher 3dmark scores, but want to have a OC and play without problems for 5 ou 6 hours not 5 min. I've been struggling with atitool to find the 100% stable mark and I think I found it but i do not know what to do.

I did not saw until last night that the shadders work by steps; The poblem with that table is that is does not include shaders bellow 1188 nor it's for mobile card but it shows the ideia of steps:

http://forums.guru3d.com/showthread.php?t=238083

Set core | Resultant frequency
frequency | Core Shader
---------------------------------
509-524 | 513 1188
525-526 | 513 1242
527-547 | 540 1242
548-553 | 540 1296
554-571 | 567 1296
572-584 | 576 1350
585-594 | 594 1350
595-603 | 594 1404
604-616 | 612 1404
617-617 | 621 1404
618-634 | 621 1458
635-641 | 648 1458
642-661 | 648 1512
662-664 | 675 1512
665-679 | 675 1566
680-687 | 684 1566
688-692 | 684 1620
693-711 | 702 1620
712-724 | 720 1674
725-734 | 729 1674
735-742 | 729 1728
743-757 | 756 1728

Problems - since the 8600 comes stock with 1:2 ratio , e.g. (STOCK) 475core 950shader, so (1:2) ratio. That means that the maximum gpu will be cut by the maximum shader. On the other side the desktop cards is the way around, the shader is cut by the gpu. But we are talking about mobile cards.

On nibitor:
The maximum gpu is 564 that gives the "112x " shader step,atitool 3dview does not crash the card, at 566 you have "113x" and that's a step above (i see this with rivatuner) and atitool 3dview after some minutes of running crashes my card giving desktop artifacts. So the problem here IS SHADER.

The funny thing is that I can play hours @ 580 / 1160 but once again atitools 3dview crashes my card. It's the only think that crashes the card at that clocks.
Do you guys know any other burin test that does not install a low level driver ?? really heavy on the gpu ?

At 600 / 1200 you have the shaders on another step and this time hours of gaming crashes the card , but 3dmark runs quite well LLOOLL

The problem here is not temps but shaders limits. Atitool 100% stable 564/112X , not atitool stable but gaming stable 580/116x but I really wanted to find another burin test.

To finish. I was allerted that nibitor has VID (voltage) editor, and indeed it has, but I did not see a single gain about it. I changed vcore values and at least i was expecting different temps but no, so it seams that we are vcore stucked.

 

If you need to be stable enough to play for 5 hours, then it only seems fair that you should stress your gpu + cpu for that long.

Run prime along side a looped 3DMark for at least 5 hours while monitoring it. That's the best indicator of stability, because it's an actual game, not a stress tool. If it doesn't crash nor overheat for those 5 hours+, then you should feel safe enough playing a game for that long. But I would not feel comfortable playing a game for 5 hours or more knowing that I only stress tested my system for 8 minutes.

 

Exactally !! I do not understand why people insist on overcloking a mobile card, run a 3dmark and say that's stable :S but again, why does atitool crash my card on e.g. 5min and 3dmark does not while looping withou the cpu tests for 30 min ??

believe me , this is bothering me ...

"because it's an actual game, not a stress tool" good point !!