warning!! Wall of text bellow!
I hope it helps the latest doubts that ppl were discussing today!
After I got home from work today I spent quite sometime doing some extensive testing with both stock bios, Rev 1 and Rev2 bios and on several drivers.
This is what I found out:
About the dircrepancy in the TDP monitors for Rev2 Bios.
I have this problem too, and I didn't found out earlier because I was checking Riva Tuner Statistic Server (which is the On Scree Monitor server that PrecisionX utilizes, but it SON't use it for the profiler), and not precision X profiler
So I did the following:
Launch PrecisionX' test tool with the Flurry+Tesselation option. It will use the hell out of your GPU and stress it a lot (not to the max of it though, you will see).
I did that with 1.212mv with both 100% and 115% power target.
Precision X presented on the profiler the power target I had set -1%. So if I had set 100% it presented I was using 99% of my TDP (wow!)
But Riva and GPU-Z were both showing aroung 73 to 78% of it. It considered my GPU too hot (because of the missreading of the TDP, probably) though it was at only 68C after several minutes burning. and then the core clocks were being kept real low, bellow 1000mhz.
THEN I found out what was wrong:
I slided the PT back down to 60% and then and ONLY then, PrecisionX profiler, Riva and GPU-Z all showed the same value of ~60% TDP, AND even though I did not stop the burner, my clocks dialed to my full overclock for that voltage which was 1241mhz.
Then I increased back the slider to whatever... PrecisionX allways shows something near the max of the slider, but Riva and GPU-Z were showing around 70-77% TDP and the clocks KEPT at 1241mhz.
Then I concluded:
* GPU-Z and Riva Tuner Statistic Server shows THE SAME TDP VALUES always (damm, GPU-Z might just use Riva Tuner).
* PrecisionX Profiler is the one which is wrong, I will tell you why I think that bellow.
I Have two MSI REFERENCE cooled cards in SLI at 1212mv. I tell you what, IF those cards were NEAR the 340W Skyn3t setted on the custom bios, they would be burning hot. WAAAAAAYY hotter than the 68C mx temp they reached at this test while PrecisionX Profiler was accusing 114% TDP
If you do the math, you will see that the ~73% GPU-Z and Riva were acusing, is 73% of the 340W Skyn3t configured, which is just around the cards original 250W TDP, and imagining those cards, at almost full load at 1212mv at ONLY 68C seems just right for 250W!!!
So I believe the first problem is solved! Skyn3t's bios is NOT BURNING any GPU out there. And for anyone's sake, if you find any discrepancy on these readings, for the TDP reading, GPU-Z is the right one.
Now to other problems I found:
As I said, on the burner load, the clocks were really low, untill I messed with the TP slider and then it maxed. So there IS a missreading from somewhere which is causing some wrong throttling.
The GOOD point is, it only happened in the burners, because with Rev2 Bios (with boost disabled) on every benchmark I had a much steadier run with my clocks locked at 1241 than with REV1 and Stock Bios. So, despite it is still a problem to think about, it might not be a big one.
Second Problem (not Bios related this one I believe):
I was talking to Skyn3t that my Metro Last Light benches were much better on 320.18 than on any 326 Drivers.
On 326 Drivers my overclocks didn't even changed the results of this specific benchmark... I had around 101 fps on 320.18 and aroun 87 on 326.01 and 95 on 326.19
Also on 320.18 On all benchmarks, my overclocks scaled better and more clearly. I could se the scores vary with my clock increases. On 326+ I couldn't, BUT on 326 the results were overall higher.
Today I was in a big doubt If I should stay on 326.01 or 320.18 and settled on 326.01 for my 3Dmark11 had a better graphics result, (and a slightly inferior Physx result which lowered my overall result, but as I'm seeing only Graphics here, that is the parameter I chose).
320.18 with cores at 1241: http://www.3dmark.com/3dm11/6911563
326.01 with cores at 1241: http://www.3dmark.com/3dm11/6911687
Also, on 326, the clocks seems a little bit more stable, but I could see some artifacts on the 3Dmark run, so I might end up lowering it a bit for gaming.
For drivers 314 some strange things happened while I tried to install them, I lost my patience and skiped them.
And lastly, I saw some ppl, Skyn3t included, with clocks comparable to mine (1241 / +400 Mem) achieving score slightly higher than mine on 3Dmark graphics scores and on Valley (I got 5000 he got 5300, on 3dmark11 graphics, I had 29489, he got 29900). I would like to find out what might be wrong on my set ups. I believe it is due to the 314 drivers I couldn't install. I think they might give me the best results I couldn't achieve.