Post Your Mining Rig Efficiency

[BinaryMage]

5870's react very well to undervolting, however finding the otimal settings can and will take long, each card is an individual. I for one am running two with the following settings (I trust these to be spot on with my miner/kernel and other mining conditions, including electricity price, difficulty, exchange rate...):
 #1 core 872Mhz, memory 200Mhz, core voltage 1.000 for 376Mhs
 #2 core 927Mhz, memory 200Mhz, core voltage 1.000 for 395Mhs

You can also go as low as 0.95v on 5870's but I havent started tweaking at that voltage yet, propably will when the block reward gets halved.. depending on the exchange rate and difficulty ofc.

Thanks, that's quite useful. I'll see what I can do.

[rograz]

Switched out the 5870 I had in my best rig since it couldn't do below 0.95V and I wanted to test even lower.

6xSapphire 5850 Xtreme@0,906V 620-650Mhz
HX850 80+ silver
1550MH/S, 464W at the wall, 3,34MH/W

This is where it becomes less profitable for me to undervolt, but damn are these cards cool. Running them with fans on auto (30-35% fan speed) and they wont go above 60C with a room temp of 25C 

Why didn't I buy a 100 of these cards rather than 10 back in may last year 

[BinaryMage]

Switched out the 5870 I had in my best rig since it couldn't do below 0.95V and I wanted to test even lower.

6xSapphire 5850 Xtreme@0,906V 620-650Mhz
HX850 80+ silver
1550MH/S, 464W at the wall, 3,34MH/W

This is where it becomes less profitable for me to undervolt, but damn are these cards cool. Running them with fans on auto (30-35% fan speed) and they wont go above 60C with a room temp of 25C 

Why didn't I buy a 100 of these cards rather than 10 back in may last year 

A question we all ask ourselves.

[Mousepotato]

Switched out the 5870 I had in my best rig since it couldn't do below 0.95V and I wanted to test even lower.

6xSapphire 5850 Xtreme@0,906V 620-650Mhz
HX850 80+ silver
1550MH/S, 464W at the wall, 3,34MH/W

This is where it becomes less profitable for me to undervolt, but damn are these cards cool. Running them with fans on auto (30-35% fan speed) and they wont go above 60C with a room temp of 25C 

Why didn't I buy a 100 of these cards rather than 10 back in may last year 

Damn, that's really good.  What is your power draw at idle before you start mining?

[rograz]

Around 170W

[Mousepotato]

170W is really good for that many cards.  What other components are you using?  CPU/memory/drive?

[rograz]

c2d@800Mhz 0,9V, Some old MSI x48 Platinum something motherboard that is undervolted quite a bit for chipset etc, 1 stick of DDR3@1,1V and a WD Green 500GB. Beats my other rigs quite a bit for cpu/motherboard power usage. with just 1 card it's below 80W idle and that's just crazy considering the efficiency for the PSU at those low loads or the HX850 is simply really good even at low loads.

My other two 6 card setups aren't nearly as impressive for board/cpu. Got one Gigabyte GA-990FXA-UD5 with a sempron that sits at around 110W idle with one card and a x58 motherboard that is around 100w with just one.

Think it has to do with the sapphire xtremes as well though, they draw around 3-5W less than your normal 5850/5870 ref board at idle. Or I should say the first revision of the cards does (6 pin connectors at the back of the card, later revision had them on the side of the card) the later revision is worse iirc, also got shittier VRMs for undervolting (no GPU registers in bios).

[oldDirty]

edit

[BR0KK]

With my GPUminer i archive 2,5mh/w (994 MH/s at 390w/h, 3 downclocked and undervolted 5870 in a Gigabyte GA990FX UD7. CPU: Sempron 145). Have still to tweak on it ... Sempron needs to be adjusted in MHz and Voltage and the HDD should be replaced with PXE or a Bootstick (somehow that board doesn't want to boot from USB Pen drives .....)


With my FPGAs I get 8.3 MH/w (1GH/s at 120w, 5 Ztex 1.15x Cluster on a Mac mini 2011 with a cheap USB Hub and all my other IT equipment - ill change that today to get the real number and I'm adding another GH to that pool at this evening.

[JinTu]

Even being optimistic I don't see how the rig could be bellow 300W with these frequencies and voltage. Something is definitely wrong here.

+1.  Something is wrong somewhere.  Even if AC/DC loss was 0 and MB/CPU was 50W.  We are talking <130W per 6990?  

Agreed. Even for me this didn't pass the smell test.

I'm looking to add an appliance power meter like this one to get real-time power stats (in Watts so no conversion required) for just the mining rig. Currently the UPS power stats are for the entire rack and I subtract out the current draw for the other components in the rack so the error margin is larger than need be.

Revised the original post with more accurate measurements now that I have the means.

MH/s (night time): 968
MH/s (day time): 843
Rig Power (Idle) 188.48W
Rig Power avg (night time):  567.49W
Rig Power avg (day time): 419.92W
Rig MH/W avg (night time): 1.71 MH/W
Rig MH/W avg (day time): 2.01 MH/W

The power in Watts is now measured from the iMeter Solo measuring the at-the-wall power for the mining rig and coupled with the data from my cgminer Cacti template. Thus far I have just a day's worth of power data (and it the weather was a bit warm yesterday) so this is certainly not the best case:




Combining these two yields

[Valalvax]

Mine is a 1 5830 rig

Entire system pulls 204 watts, the card pulls 179 watts, I get 305 MH out of it

1.4878 MH/W for the system

1.694 MH/W for the card alone

[jjiimm_64]

SSonic 1250 + SS 650
crosshairV mobo
phenom 4core
4g ram
windoz 7

930/800/937v 

830 watts (killawatt) 
2770Mh


similar rigs on 220 that probably get 8% better

Code:

cgminer version 2.3.6 - Started: [May 6, 2012, 11:11 pm]    Rig:miner19
(5s):2773.44  (avg): 2762.83 Mh/s  |    H: 83.4  Q:56396   A:87440   R:244   HW:0   E:?%   U:38.36/m
TQ:?   ST:73   SS:?   DW:4198   NB:288   LW:147120   GF:77   RF:32
Connected to http://gpumax.com:8332 with LP as user ?
Value:
GPU 0:	74.0C	2290RPM 	39% 	113	| 555.0/552.7Mh/s | 	99% |	930Mhz 	800Mhz 	1.17V 	A:17635	R:45	HW:0	U:7.74/m 	I: 9
GPU 1:	65.0C	1080RPM 	0% 	65	| 554.6/552.7Mh/s | 	99% |	930Mhz 	800Mhz 	1.17V 	A:17411	R:51	HW:0	U:7.64/m 	I: 9
GPU 2:	74.0C	1217RPM 	33% 	107	| 554.7/552.6Mh/s | 	99% |	930Mhz 	800Mhz 	1.17V 	A:17438	R:49	HW:0	U:7.65/m 	I: 9
GPU 3:	64.0C	1079RPM 	0% 	64	| 554.3/552.1Mh/s | 	99% |	930Mhz 	800Mhz 	1.17V 	A:17630	R:49	HW:0	U:7.74/m 	I: 9
GPU 4:	68.0C	1088RPM 	0% 	68	| 554.8/552.7Mh/s | 	99% |	930Mhz 	800Mhz 	1.17V 	A:17326	R:50	HW:0	U:7.60/m 	I: 9

[jjiimm_64]

My new rig configuration:
4x5970
3.2 GH/s @ 1080W
3.02 MH/W

That is on 240V circuit, 80-Plus Gold PSU, usb drive running linux, optimized BIOS, underclocked & undervolted sempron, and >3 GH/s per rig.

I am pretty sure it isn't possible to get higher efficiency from a GPU rig without undervolting the GPUs.  A 4x7990 (or possibly 3x7990) rig will likely beat that though.

D&T, what clocks are you running to get  3.2G out of 4x5970.  ?

 

[DeathAndTaxes]

My new rig configuration:
4x5970
3.2 GH/s @ 1080W
3.02 MH/W

That is on 240V circuit, 80-Plus Gold PSU, usb drive running linux, optimized BIOS, underclocked & undervolted sempron, and >3 GH/s per rig.

I am pretty sure it isn't possible to get higher efficiency from a GPU rig without undervolting the GPUs.  A 4x7990 (or possibly 3x7990) rig will likely beat that though.

D&T, what clocks are you running to get  3.2G out of 4x5970.  ?

 

High.

They are water cooled.  Each card is clocked differently because once you remove the heat each GPUs individual ASIC quality starts becoming the performance bottleneck.  They average ~860 but no two GPU is clocked the same.  Most are in the range of 830 to 875 range.  I have found you can't get past 900Mhz and keep them stable.  

I think the 3 phase VRM (5870 uses 4 phase) just can't handle it.  I do have some lagards though.  I have some which can't get above 800 even cooled to <40C.  I have two cards which I think have a bad VRM because 1 of the 2 GPUs can't go above 700.    Of the 48 GPUs I had one simply die when I watercooled it.  Explain that?  Maybe I pushed it too hard and without excessive heat to throttle it back it killed itself.  It had been flaky prior to watercooling it though.  I think I am going to remove the waterblock and just use the remaining core for gaming.  

Due to the luck of the draw not all the rigs hit 3.2 GH/s.  Of the 5 rigs they are 3.22 GH/s, 3.20 GH/s, 3.15 GH/s, 3.08 GH/s, 3.01 GH/s.  I got one more rig to convert.  I am hoping it is a 3.3 GH/s+.  The 3.08 and 3.01 likely could be pushed a little harder but building the watercooled server rack was a lot of work so I am happy they are just stable right now.

Ironically I am now kinda "over" mining.  I am moving on to other things hopefully they just mine away for a long time.

[rograz]

So I've been doing some more work on my quests for efficiency and the result is quite impressive.

New rig

Silverstone Strider 1000W +gold
gigabyte 990fxa-ud5
Sempron@1ghz 0.9V
OLD 160GB seagate HDD

6x5870 XFX Reference 0.9V 600-640/300 (fans run at idle speed and cards stay below 65C at these settings)

1640MH/s

Power consumption: 443W after 1h of running (had to test with 2 different brand of kill-a-watt cause it seemed to low but the measured difference was only 2-3W)

Efficiency:  3.7MH/W

[jjiimm_64]

NICE!  i have been running my 5x5870  and 6x5870 rigs at    800/300/0.99.  not sure of exact watts, but I know that each card is approx 100w each

[bitlane]

I just began setting up my new 7970 rigs and so far:

- 4x Gigabyte Reference HD7970 @ stock voltage, 1100mhz GPU, 950mhz Mem
- Gigabyte X58 I7 Motherboard
- Xeon 5506 80W Quad Core CPU
- 4GB DDR3
- SATA HDD
- Win7 x64, CAT 12.3, CGMiner 2.4.1
- 2 Power Supply (OCZ Z-Series) Setup (80+ 850W for Motherboard/Slot Power + 2 Cards Aux Power, 80+ 650W for Aux/External Power to 2 cards + HDD + additional cooling fans)

(655 MH/s each x 4) 2,620 MH/s @ 1,000W at the plug (UPS with LCD Screen/Power Readout)

So, I guess that's approx. 2.62 MH/watt

I am currently trying to find a way to trim the power usage a bit.....

[jjiimm_64]

I just began setting up my new 7970 rigs and so far:

- 4x Gigabyte Reference HD7970 @ stock voltage, 1100mhz GPU, 950mhz Mem
- Gigabyte X58 I7 Motherboard
- Xeon 5506 80W Quad Core CPU
- 4GB DDR3
- SATA HDD
- Win7 x64, CAT 12.3, CGMiner 2.4.1
- 2 Power Supply Setup (80+ 850W for Motherboard + 2 Cards, 80+ 600W for Aux/External Power to 2 cards + HDD + additional cooling fans)

(655 MH/s each x 4) 2,620 MH/s @ 1,000W at the plug (UPS with LCD Screen/Power Readout)

So, I guess that's approx. 2.62 MH/watt

I am currently trying to find a way to trim the power usage a bit.....

its all about voltage....   5x7970 @ 930/800/0.937v   is  830 watts.  2770Mh

im guessing 4x7970 same would be about 650 watts.

[bitlane]

its all about voltage....   5x7970 @ 930/800/0.937v   is  830 watts.  2770Mh

im guessing 4x7970 same would be about 650 watts.

Does Powertune have any effect on your system(s) with 7970's ?

[jjiimm_64]

its all about voltage....   5x7970 @ 930/800/0.937v   is  830 watts.  2770Mh

im guessing 4x7970 same would be about 650 watts.

Does Powertune have any effect on your system(s) with 7970's ?

i have never used powertune..  if it does anything with volting, it wont work