Topic: Antminer S5 - Underclock - Undervolt - Best J/GH - page 9. (Read 31123 times)

I think the limit might now be being set by parameters other than frequency? I think I may have hit bottom limit on either the Core Voltage or the voltage for the IO or PLL? However it would be interesting to try lower if someone can tell me the file that I need to SSH into to add an additional frequency?

In addition to this I have identified another difference in the V1.91 PCB which may play a part in the board being able to undervolt? Not quantified yet but they seem to have done away with the LDO regulators for the chips? Although this could be a benefit at "normal" voltages it could also be a problem when they are significantly lowered? More measurements needed will report back when I have more data.


Rich

Anything around 0.3 is pretty fantastic regardless. Did you try under 100mhz just for fun?

Might have just set a new S5 Record?

I cannot get Hashing to start reliably with less than 9.3V Supply, 0.62V Core. However once hashing you can wind the Voltage down.
So at 100MHz which is 330GH I got the Supply down to 8.4V, 0.56V Core. Watts at the wall were 97W which gives a J/GH of 0.294 
No use in practice as it only ran for 10 Minutes before one board stopped hashing, but first time under 100W & 0.3J/GH

Rich
 

Look back at this post and you can see the different versions of the Hash board.

https://bitcointalksearch.org/topic/m.12164839

It would be possible to upgrade the V1.9 to V1.91 as it has the pads for the additional components, however a tricky job unless you are equipped to solder surface mount components.

Upgrading the earlier versions would involve making a small daughter board with the additional oscillators and then patching them into the circuit. Again a tricky job.

Also remember it is only a theory at the moment that it was the addition of the oscillators that enables the boards to Hash at a lower voltage.

Rich

Is it possible to upgrade V1.3 board to v1.9, are the board different or only parts missing from older boards?
Any detailed photos?

I had older firmware and fan procent thing wasn't there, I upgraded but I don't have any clue what this setting does?

"Customize the fan speed percentage" percentage is counted from what?
If miner is very cold I have the same 2040 speed still.
  

Undervolted S5 don't like to bee treated  as farming tool (heating tomatoes) and rebels all the time.
I had to adjust volts to 11 and Mhz 306, 3G mobile network for this miner don't make things easier either.

OK let's put aside for the moment PSU efficiency and have a look at some hard data on the 4 individual S5 Hash boards that I have now characterised.

First the measurement Technique.

For all tests I have used the Pair of Meanwell HRP-150-5 5V PSU's shown in the earlier picture. These are connected in series and the Power at the Wall measured on the Meter also in the picture.

I measured each Hash Board in turn with the second card disconnected, so the measurements will slightly undercall what is possible with 2 Cards as the Controller Board & Fan are just being used for a single board.

I adjusted the voltage at each test point to give about 0.01% to 0.02% HW error rate. I found it much easier to go for this in time terms rather than trying to adjust for 0%. In practice you might want to up the voltage a notch and have a lower HW error rate?

So there are 4 tables below showing for each board a range of Frequencies, Supply Voltage, Core Voltage, Power at the Wall, Hash Rate & J/GH.


There is some variance  between the boards, which in practice in a 2 Board S5 means you would have to adjust the voltage for the worst board of the pair. Ideally you need a matched pair in a miner so looking at the results I will probably pair up S5n1 Side 1 with S5n2 Side 2 & S5n2 Side 2 with S5n2 Side 1.

So useful J/GH gains at all frequencies by optimising the voltage. Best J/GH being 0.343 on S5n2 Side 2 at 175Mhz with a supply voltage of 9.2V & a Core Voltage of 0.61V

At some point it would be good to try and understand the Board variance. However I think this will be difficult with the string design as it make isolating chips very difficult.

Rich

I see, the minimum and maximum must be coded-in. It is probably something akin to 35-80% or something. For instance setting it to 100% does not raise it over 3840 iirc. And 3840 is about 80%~.

Beside doing some digging in the firmware to lower it further, your other option would be a better noise/cooling ratio fan. Or lower the voltage to the fan.

 Nope. each supply does half the power, so you take the efficiency of each supply and average them for the overall efficiency.

 You would only MULTIPLY efficiency if you were feeding the output of one supply into the next supply, like feeding one of the "buck converters" from early in the thread from an ATX or server PS.

2040 fan speed is usually the lowest I see even when the miner is very undervolted & underclocked. I have seen 1920 but only when the miner is very cool.

Setting to 20% does not set it any lower, but have not fully understood the % fan speed algorithm, however the miner does run progressively hotter as you reduce below 50%.

Rich

When you set fan to 20%, it doesn't go any lower?

Is there a hacked firmware for S5 which allowes fan to turn slower than 2040 rpm?

I threw underwolted underclocked S5 to porch for tomato heating and reported miner temp is 32/27 C and fan speed is 2040, same as non hashing miner had at 50W power draw.

you combined the 2

5's into a 9-11 volt.  I think it  multiplies the efficiency   which lowers you closer to .80 not .87


Not sure .

Why have I got to do .87 x .87 ?

yeah then do .87 x .87 = 0.7559

and it    kills the efficiency    you said you were at .8 or .81   I am pretty sure this is why.

on a completely different take for sound control….  which goes hand in hand with under clock..  I found this


http://www.ebay.com/itm/8-Inch-Hydroponics-Duct-Muffler-Inline-Fan-Blower-Filter-Silencer-Noise-Reducer-/261492562812?

this is too much money as the s-5 needs one

and the s-7 need two  



but this http://www.ebay.com/itm/220659184713?

inside this

http://www.ebay.com/itm/Sonoco-Sonotube-cardboard-warehouse-storage-tubes-concrete-forms-/321172021836?

or inside this

http://www.ebay.com/itm/692201-8x4-Concrete-Form-Tube/391158475313?


should kill off a lot of noise.

I am thinking the s-7 will need two  2ft long 8 inch tubes

The s-5 will need one 2ft long  8 inch tubes.

I can get tubes at home depot for 6 bucks  each tubs makes 2 mufflers.  I ordered enough foam to build. 18 feet of mufflers or about nine 2 footers.

I am pretty excited as I think I can get s-7's to run quiet and at 4000gh at 900 watts.

This means I can do 5 in house.  1 at my friends office and 1 at my other friends wood working shop.

that would be 7 x 4th = 28th

Agreed the efficiency is lower at the lower voltages, the ones I have (HRP-150-5) are rated at 87%, although I measure them a bit less than their rated efficiency as I do all the PSU's I have.

They are however the only off the shelf way I have found to get a voltage range of 9V to 11.5V which covers the full undervolt range. Also I was very lucky with an ebay purchase and bought 4 for $45 delivered which was a bargain I could not pass up.  


Rich

the 5 volt meanwells are a bad choice.  all of them  are not efficient.


 this one is good  my goal is 2 of these at 10.2 volts or 10.5 volts  running an s-7

if I can drop from 12 to 10.2  that is 15%      since no plat psu does better the 94%  this one is 88%

   so I am better then 11% power improvement. over a high end plat

http://www.trcelectronics.com/View/Mean-Well/HRP-600-12.shtml

https://www.trcelectronics.com/ecomm/pdf/hrp600.pdf

I guess if you are near the limit of something just working or not, then you only need to change another variable like temperature a little and that's enough to throw it either way. Another factor is that the voltage needed to start hashing is slightly higher than the voltage needed to keep hashing. Anytime you have a system that is hashing if you have adjusted the voltage down since it started you need to reboot to check if it will actually start at that voltage.

I have been making measurements on individual S5 Hash boards and there is some variance between them, will post up the results when I have finished.

Rich

S5 V 1.3 seems to be very moody, no start when cold at morning 300Mhz , volts were 10,9.

11,3V were enought for std 350Mhz hashspeed 1100 530W, fan 3700 and loud as std.

Now early at evening 10,7V is enought for 293,75 Mhz 930Gh/s and 415W 2400rpm fan.

 Better efficiency on the Meanwells, if you grab the right ones.
 Cost vs. efficiency is usually a tradeoff in any power supply situation.



 S1 has onboard voltage regulation, no need to use an external "undervolt capable" power supply, and defeats the purpose anyway.
 S5 is a string design with NO onboard voltage regulation, you HAVE to undervolt them externally.


 While the 12V Meanwells usually spec to go down to 10.8, they'll usually go a LITTLE lower - not to 9V though.