Watt usage of an undervolted 5850?

[Leezyist]

Does anyone have an idea of the watts used by a 5850 running 825/150/1v? I'm wondering if it would be plausible to run 6x 5850 on a GD70 (+ sempron 145) with these settings on a 1000w gold rated power supply.

I figure 50-60w for mobo? (guess)

that leaves ~940 watts

940/6 = 156.6 a piece for each card.

So, I'm hoping I can get them down to ~140 watts per card or lower. Would 825/150/1v give wattage in this range?

[1714904185]

1714904185

[1714904185]

1714904185

[1714904185]

1714904185

[imsaguy]

Might need to go .95v

[Leezyist]

Might need to go .95v

I thought about that. I think I heard that 825/150/1v were the most efficient settings for profitability. I pay next to nothing for power consumption, so the electric cost isn't a concern, but I wanted to get the best bang for the buck out of the rig. I may end up running 5 GPUs on the rig and mildly overclocking them.

940/5 = 188. I think I could get 5 cards slightly OC'd and meet that range. maybe 875/150/1.025v?

anyone know the best volt settings for 875-900 core?
 

[Dargo]

I can tell you that 5850s @ 0.95 v = about 90w. So you can definitely run 6 at that voltage. Not sure about anything higher though. I my experience, I get 300-305 Mh at 0.95. Incrementing the voltage up (i.e. anything between 0.96 and 1.038) gives me maybe 315 Mh before instability starts to set in. So in my experience the higher voltage wasn't really worth it. Better IMO to stay at 0.95 and have better efficiency.

[Leezyist]

I can tell you that 5850s @ 0.95 v = about 90w. So you can definitely run 6 at that voltage. Not sure about anything higher though. I my experience, I get 300-305 Mh at 0.95. Incrementing the voltage up (i.e. anything between 0.96 and 1.038) gives me maybe 315 Mh before instability starts to set in. So in my experience the higher voltage wasn't really worth it. Better IMO to stay at 0.95 and have better efficiency.

I agree with you on that. I kept getting the 127.5c/0 rpm error until I undervolted the cards. Now it runs stable with no hangs @ 1250 MHash/s. I don't know about watt usage, but I have a kill a watt on the way from newegg now. so, I'll know soon.

[n4l3hp]

My sapphires are quite stable at 0.85v up to 780MHz. Helped the GPUs to not go above 75C at 60% fan speed. I can't get past 800MHz @ 0.95v though, had to use 1.0v for it to be stable up to 840MHz.

[Leezyist]

My sapphires are quite stable at 0.85v up to 780MHz. Helped the GPUs to not go above 75C at 60% fan speed. I can't get past 800MHz @ 0.95v though, had to use 1.0v for it to be stable up to 840MHz.

Wow, your temperatures seem extremely high to me.



I just remoted into the rig and this was it's current temp readings. 3x 5850 1x5830 all 825/150/1v

http://i49.tinypic.com/2u8a236.png

[crazyates]

My sapphires are quite stable at 0.85v up to 780MHz. Helped the GPUs to not go above 75C at 60% fan speed. I can't get past 800MHz @ 0.95v though, had to use 1.0v for it to be stable up to 840MHz.

Wow, your temperatures seem extremely high to me.



I just remoted into the rig and this was it's current temp readings. 3x 5850 1x5830 all 825/150/1v

Ya they are really high. The only time I had temps like that is when a bunch of cards weren't using riser cables, and there was no space between them for the fans to work. The fans were running at 75%, but hardly moving any air.

[oldDirty]

I got a ex-rig with 4x 5850 850mhz - 1.05xV, PSU was a Enermax 600W 87+ pro.
( AMD Quad, GD70, hdd, ssd, 4gb ram and i do play with that rig to ).

[n4l3hp]

Ambient temp is about 36-39C so its normal for me (window is about 1.5M above the rig), and yes the cards are not using risers. I can lower temps a little bit if I use 150 mem but the 5970 (using a riser and above the 5850s) is not okay with anything lower than 300 mem. Underclocking/undervolting are all done via CGMiner (except the 6870). Using a desk fan to blow air over the rig will only decrease the efficiency. My cards are more than 2 yrs old now, used them for BOINC before I got into Bitcoin. Coolers, fans, TIMs and pads haven't been replaced yet.

Just wondering, does "E" on the cgminer screen stats stand for efficiency or something else? A day of mining results in E:360% but if I increase the core clocks, E drops down to 200%+ and drops even more to about 140% if I go above the default core clock. Using a kill-a-watt, the settings with E:360% gives me a better MH/w.

Specs:
Athlon II 170u (CnQ enabled, 0.9v, 0.7v @ idle)
Gigabyte 890GPA-UD3H
2Gb RAM
16Gb DOM
1x 5970 (625/300 @ 0.95v)
2x 5850 (780/300 @ 0.85v)
1x 6870 (775/300 @ 1.10v)
800w Corsair GS
Windows XP 64, 11.12/2.5, CGMiner 2.4.4

4 x6500 running MPBM is also attached to the rig.

[Leezyist]

Ambient temp is about 36-39C so its normal for me (window is about 1.5M above the rig), and yes the cards are not using risers. I can lower temps a little bit if I use 150 mem but the 5970 (using a riser and above the 5850s) is not okay with anything lower than 300 mem. Underclocking/undervolting are all done via CGMiner (except the 6870). Using a desk fan to blow air over the rig will only decrease the efficiency. My cards are more than 2 yrs old now, used them for BOINC before I got into Bitcoin. Coolers, fans, TIMs and pads haven't been replaced yet.

Just wondering, does "E" on the cgminer screen stats stand for efficiency or something else? A day of mining results in E:360% but if I increase the core clocks, E drops down to 200%+ and drops even more to about 140% if I go above the default core clock. Using a kill-a-watt, the settings with E:360% gives me a better MH/w.

Specs:
Athlon II 170u (CnQ enabled, 0.9v, 0.7v @ idle)
Gigabyte 890GPA-UD3H
2Gb RAM
16Gb DOM
1x 5970 (625/300 @ 0.95v)
2x 5850 (780/300 @ 0.85v)
1x 6870 (775/300 @ 1.10v)
800w Corsair GS
Windows XP 64, 11.12/2.5, CGMiner 2.4.4

4 x6500 running MPBM is also attached to the rig.

The E in CGMiner is a simple percentage of Requested work divided by Accepted shares. It has no bearing on how efficient the actual hardware itself is.

[crazyates]

Ambient temp is about 36-39C so its normal for me (window is about 1.5M above the rig), and yes the cards are not using risers. I can lower temps a little bit if I use 150 mem but the 5970 (using a riser and above the 5850s) is not okay with anything lower than 300 mem. Underclocking/undervolting are all done via CGMiner (except the 6870). Using a desk fan to blow air over the rig will only decrease the efficiency. My cards are more than 2 yrs old now, used them for BOINC before I got into Bitcoin. Coolers, fans, TIMs and pads haven't been replaced yet.

Just wondering, does "E" on the cgminer screen stats stand for efficiency or something else? A day of mining results in E:360% but if I increase the core clocks, E drops down to 200%+ and drops even more to about 140% if I go above the default core clock. Using a kill-a-watt, the settings with E:360% gives me a better MH/w.

Specs:
Athlon II 170u (CnQ enabled, 0.9v, 0.7v @ idle)
Gigabyte 890GPA-UD3H
2Gb RAM
16Gb DOM
1x 5970 (625/300 @ 0.95v)
2x 5850 (780/300 @ 0.85v)
1x 6870 (775/300 @ 1.10v)
800w Corsair GS
Windows XP 64, 11.12/2.5, CGMiner 2.4.4

4 x6500 running MPBM is also attached to the rig.

The E in CGMiner is a simple percentage of Requested work divided by Accepted shares. It has no bearing on how efficient the actual hardware itself is.

E:   The Efficiency defined as number of shares returned / work item
U:   The Utility defined as the number of shares / minute