Final clocks
All of the following clocks have been tested extensively.
Code:
400/625 @ 800mV Uses less power than the default dual monitor profile
650/1000 @ 900mV The power of a 6870 at very low volts
801/1250 @ 990mV 110mV less than stock 6950 @ 6950 clocks
880/1375 @ 1060mV 115mV less than stock 6970 @ 6970 clocks
910/1375 @ 1100mV 110MHz overclock at stock 6950 voltage
940/1375 @ 1175mV 60MHz higher than a 6970 at its default voltage
My clock profiles
400/625 @ 800mV - 24/7 desktop use with dual monitors
650/1000 @ 900mV - Video playback when PC used as HTPC and playing older games
940/1375 @ 1175mV - Playing modern games that need the extra ooomph
Test Bench
- i5 750 @ 3.98ghz Turbo Boost disabled
- Gigabyte P55 UD5
- 4GB G.Skill ECO 7-8-7-24-1T @ 1592mhz
- Corsair HX650W
- Samsung F3 1TB + Samsung F3 500GB
- Windows 7 64bit
- 6950 unlocked shaders with Shaman fitted - Driver version 11.3
- Raven RV02 with side panel removed and only middle fan spinning at 5volts
- Ambient Temperature 20.5c during all testing
Thermalright Shaman info
I have the shaman connected directly to the graphics card using this.
Fan speeds are as follows:
- 25% = 625rpm
- 50% = 930rpm
- 75% = 1170rpm
- 100% = 1280rpm
At 25% the fan is completely inaudible and barely spinning. At 100% the shaman is a fraction quieter than the stock fan at 25%.
Misc
A clockspeed of 801mhz was required to apply voltage adjustments in afterburner. For some reason it will not allow changes when running at the bios stock speed.
I have only included results upto 1.175v. In testing I found 1.225v was enough to activate the limiter in some games. As yet afterburner will not allow us to apply +20% power. Sapphire Trixx Will, but you cannot run below stock volts with that app (1100mV).
Stress testing
Furmark 1.8.2 20minutes @ 1680x1050 4xaa
Crysis Warhead 50 runs @ 1680x1050 DX10 Gamer 0xaa Ambush - higher load with AA disabled
Unigine Heaven 2.5 for 2 hours
4+ hours of actual gaming
- Load power consumption figures represent peak value rounded up to nearest 5watts
- Load temperatures are taken after 30minutes
- Idle power consumption is averaged over 10 minutes of inactivity
- Idle temperatures are taken after a quick furmark to bring temps up, followed by 10minutes of inactivity to allow them to fall
Power & Temps @ 100% fanspeed
Idle single monitor clocks of 250/150 are included. This shows my PC consumes roughly 100watts without a graphics card.
Code:
Core/Mem@mV Idle Crysis Furmark
250/150@900 121watts @ 27c
400/625@800 134watts @ 29c 195watts @ 32c 190watts @ 34c
650/1000@900 147watts @ 31c 230watts @ 37c 225watts @ 39c
801/1250@990 157watts @ 33c 260watts @ 40c 255watts @ 42c
880/1375@1060 164watts @ 37c 275watts @ 44c 285watts @ 55c
910/1375@1100 166watts @ 37c 290watts @ 46c 295watts @ 59c
940/1375@1175 170watts @ 38c 310watts @ 51c 330watts @ 82c
When consuming more 330watts in furmark using lower AA settings, the Shaman would exceed 90c. I have no intention of killing my card, so at this point I abort.
The Shaman appears to have a cooling limit of ~200watts (330watts furmark - 100watts no card with ~88% power effeciency).
Temps by fanspeed
Code:
25% 625rpm 100% 1280rpm
Core/Mem@mV Idle Crysis Idle Crysis Furmark
250/150@900 29 - 27 - -
400/625@800 31 36 29 32 34
650/1000@900 33 41 31 37 39
801/1250@990 36 45 32 40 42
880/1375@1060 37 50 33 44 55
910/1375@1100 37 52 33 46 59
940/1375@1175 38 58 34 51 82
Did I mentioned the Shaman is inaudible at 25% fanspeed? Just look at those temps!
Benchmarks
All of these benchmarks are run at the hexus review settings. Crysis is a standard bench run rather than fraps, but the result is close enough. My scores are slightly higher due to using a newer driver.
Code:
Core/Mem@mV Crysis JC2 Mafia2 3DMark Vantage
400/625@800 30.2 32.6 36.1 6659
650/1000@900 48.5 52.1 57.4 11173
801/1250@990 58.3 62.6 69.4 13434
880/1375@1060 62.8 67.9 75.2 14877
910/1375@1100 64.1 69 76.8 15233
940/1375@1175 65.1 70.1 78.6 15429
Below are the hexus results for various cards to compare against.