Gigqbyte GTX 670 OC - what determines % power

rakesh_toms

Distinguished
Jan 6, 2012
26
0
18,530
Guys I have a Gigabyte GTX 670 Windforce OC card running in the below machine

AMD Phenom II X4 @ 3.0 GHz
4 GB DDR3 1600
350 W generic PSU + a dedicated 250 W Thermal Take PSU for graphics card
Win 7 Ultimate 64 bit
Latest drivers from Nvidia website

I have always used Dirt 3 canned benchmark on this same machine across various graphics cards I have owned in the past 3 years with all settings on ultra 1600X900 and below are the numbers

5850 (OC 825 Mhz core, 1125 mem) - avg FPS 56
7870 (OC 1150 core, stock mem) - avg 82
670 (factory OC 980 base, 1163 boost, 3005 mem) - avg 88

Not too happy with the bump from 82 to 88, was expecting more. MSI AB reads 62 C max temp, 97% GPU usage, 1163 core clock, but power usage never goes above 70% during benchmark. The power limit slider is set to 100 max for power in AfterBurner. I am wondering why real time power usage doesnt go to near 100 and if that is causing lower FPS.
What determines Power %? I have seen in videos during 3D load power always hover near 100%. In all games like BF3, Batman Arkham City and Witcher 2 I have noticed my power % never go above 70.
 

rakesh_toms

Distinguished
Jan 6, 2012
26
0
18,530
ok you thinking the power usage is stuck at 70% or below coz the PSU is not providing enough wattage? I have started using this setup in 2008 with a 8800GT and all the other cards i mentioned in the OP and never had issues even OCing them. Just wanted to understand the concept behind what determines max power usage for Kepler cards if anyone knew for sure, before I go buy a new one and it does not solve my problem(if it is really a problem).
 

rakesh_toms

Distinguished
Jan 6, 2012
26
0
18,530
I am talking about power %, not GPU usage. Wish I could post AB screen here...here are the values after running the benchmark

Max Power % - 72
Max GPU Temp - 65
Max GPU usage % - 99
Max fan speed % - 44
Max core clock - 1163
Max mem clock - 3005

I have noticed the reply about changing power supply and it may be a valid suggestion. But I wouldn't have raised any questions if there were driver crashes or the boost clock was conservative. Here there are no crashes and the real time boost clock is a healthy 1163 (no OC from my part over factory set 1059). Again my question is what determines the max power % usage. Temp is a healthy 65 max, otherwise I would have concluded that it is trottling the power usage.