GPU clock not OC'able

[HenryFaber]

I have a Windows 8.1 Lenovo G70-70 laptop with a core i3-4005U and a discrete GF 820M with 2GB of VRAM. In the Render Test in the program GPU-Z the "card" clocks at sustained 950 boost MHz. The laptop is getting warm and loud but nothing critical. With MSI Afterburner I can add e.q. 150 MHz and the "card" goes to 1.100 MHz no problem. In the arena game World of Tanks I can see in garage and during the loading screen that the clock is bouncing between 600 and 950 or 1.100 (OC). In the first 30 seconds of the game where a clock is counting down and you can see your team standing there but no controls work it is bouncing between 720 and 950 MHz or 1.100 (OC). But when the game really starts the clock is hardlimited to 720 MHz. The MSI Afterburner shows an exact flat line of 720 during the game (up to 15 min). After the game in the garage with the post game stats it is 600 - 950 again.

It is the same with EVGA Precision X and the nvidia inspector. No matter how high I OC this "card" in the game it's just 720. The RAM OC works fine. Normally 900 and if I add 200 it's 1.100... in game, out of game, everywhere.

 

What's this all about? Is there an undocumented unchangeable P state that this game uses? I tried a lot in the nvidia control panel. Max quality. Max power. I tried something in the Windows energy control. Nothing. Turned VSYNC in game on / off, triple buffer on / off. Nothing. Does anyone also experience this? 

[HenryFaber]

I'd like to add that the graphics card is used by over 90% and the cpu only to about 60%. I closely watched the fps during the first 30 secons where it still bounces. Didn't move or scrolled the mouse. It goes 70 80 70 80 70 80 70... So there'd be potential for more fps. The starting scene might be calm, nobody moves, nobody shoots. But 10 fps more with 230 MHz more. The temps of the grahics card are 75C after a long game (with 720 MHz). Does the card maybe recognize the game as a threat to overheat if sustained 950 are used?