[REVIEW] ANTMINER S5 1250GH/s@610W(at wall) + Hi-Res Pics + unofficial Support

[goxed]

Disclaimer: I received a review unit from BITMAIN. Thanks BITMAIN!!
This review is being written after 8days of use.
The unit has been functioning as per expectations (hash-rate >= 1.2TH/s, pulling 610Watts at wall via a 1KW 80Plus gold PSU)
Verdict: Excellent home miner, with class leading power efficiency (2GH/s per W or 0.5W per GH/s) and hashrates in a compact form-factor.

(All copyrights to the following images have been released; these can be freely used by anyone for whatever purpose they deem fit)

[goxed]

Support thread.
https://bitcointalk.org/index.php?topic=935397.0

[goxed]

Hashrate stability chart at the pool @ 393MHz clockspeed. ~1250GH/s:610Watts

[alh]

When I look at your pictures, it appears that your power supply is using one cable, with two PCIE connectore, for each hashing board. That would suggest that each power cable is handling roughly 270 watts. Are the PCIE cables or connectors hot when the S5 is running at full speed. That seems like it's about the limit of what you would want on a PCIe power cable, no? Maybe I don't really understand what a PCIE cable can deliver.

Just a thought.

[Zich]

When I look at your pictures, it appears that your power supply is using one cable, with two PCIE connectore, for each hashing board. That would suggest that each power cable is handling roughly 270 watts. Are the PCIE cables or connectors hot when the S5 is running at full speed. That seems like it's about the limit of what you would want on a PCIe power cable, no? Maybe I don't really understand what a PCIE cable can deliver.

Just a thought.

Same thought here, as far as i know these cable can handle up to 360 watt(3 x 10A).
I guess the cable will be warm.

[dogie]

When I look at your pictures, it appears that your power supply is using one cable, with two PCIE connectore, for each hashing board. That would suggest that each power cable is handling roughly 270 watts. Are the PCIE cables or connectors hot when the S5 is running at full speed. That seems like it's about the limit of what you would want on a PCIe power cable, no? Maybe I don't really understand what a PCIE cable can deliver.

Just a thought.

260W is fine. That cable has 2x PCI-E 8 pin connectors on so must be able to handle at least 2x 150W.

[goxed]

When I look at your pictures, it appears that your power supply is using one cable, with two PCIE connectore, for each hashing board. That would suggest that each power cable is handling roughly 270 watts. Are the PCIE cables or connectors hot when the S5 is running at full speed. That seems like it's about the limit of what you would want on a PCIe power cable, no? Maybe I don't really understand what a PCIE cable can deliver.

Just a thought.

260W is fine. That cable has 2x PCI-E 8 pin connectors on so must be able to handle at least 2x 150W.

Exactly, the power rating of the 8pin connector is 150W per connector. The cable is slightly warm to the the touch.

[TwinTurbo]

Have you tried any clock rates higher than this? What was the highest stable?

What is the wall voltage you are using?
Presumably PSU efficiency will be higher for higher incoming voltages.

Either way, very nice and thanks for the post!

[dogie]

Presumably PSU efficiency will be higher for higher incoming voltages.

Surprisingly, not by that much. You might see 1-3% for 230 over 115V which I think is a testament of how well developed the tech has become.

[goxed]

Presumably PSU efficiency will be higher for higher incoming voltages.

Surprisingly, not by that much. You might see 1-3% for 230 over 115V which I think is a testament of how well developed the tech has become.

Yup ~2% loss of efficiency by going down to 115V. A few other obvious locations of power loss are the DC power cables i.e. the PCIE 6/8 pin power cables.
I have noticed that if I use a dedicated cable for each of the 4 PCI 6pin socket on the miner, the power consumption is reduced by ~8Watt for the Antminer S5.
 

[goxed]

Have you tried any clock rates higher than this? What was the highest stable?

Code:

Tried 400, but HW error rates started to rise considerably

What is the wall voltage you are using?

Code:

220V

Presumably PSU efficiency will be higher for higher incoming voltages.

Code:

~2% higher

Either way, very nice and thanks for the post!

Code:

Thanks!

[toptekk]

After reading about the heat sink trick. i was no way so I tried it i was wow what big a diff  in how cool it runs and how much more quiet it is even with the stock fan which i took off it's still to load for me even with the heats sink applied .Ive been running  one miner now for 10 Hours with the heat sink trick and two of these fans http://www.amazon.com/gp/product/B00FKNAG7Y/ref=oh_aui_detailpage_o06_s00?ie=UTF8&psc=1 doing push pull,  i would link  a Image but can't figure how to on the forums ..

but after 10 Hours with the heat sink trick and the above fans the sound level is like my PC or freezer in the  back ground.



1.13 TH/s
hashrate avg
25 (0%)
hw errors


fan 1 1,920   

fan 2 1,920

temps blade 1  50

temps blade 2  46
   
clocked at  350 both blades.


I did some OC  testing with the heat sink trick and new fans the temps did not go up that much higher but i really don't see a  need to OC so all my S5 are at defaults, I'm very happy with the sounds .

they  run very sable this way, i even ordered more heats sinks and plan to do my other two S5 the same way .

[Ilrale]

Hi ,

I don't understand every think . But , i have one S5 .
I run it into 437.5M  and Have about 1377 GH/S
my temps are about 55 and 56 , but i have some Hardware errors .

What does HW mean ? should a made it run lower and have no HW ?

Thx for your help

[goxed]

Hi ,

I don't understand every think . But , i have one S5 .
I run it into 437.5M  and Have about 1377 GH/S
my temps are about 55 and 56 , but i have some Hardware errors .

What does HW mean ? should a made it run lower and have no HW ?

Thx for your help

Some Hardware error is fine; it arises due to variances in the ASIC silicon. Adjust your operating frequencies so that total hardware error is <0.1%.