New firmware for antminer S9 - less power consumption

[MoparMiningLLC]

This is amusing:

https://puu.sh/BR0Fs/051d5f93cd.png

I put a new controller into one of my early batch (maybe batch 1)  S9s (12.93 rated TH), upgraded to 1023, and pushed it until I started seeing issues, then backed down a bit.  e.g.  It has serious problems at 750, but not 725.  No other changes, no voltage tweaks, etc.

Basically those early models rock.

how much power is it pulling at that speed?

[ccgllc]

how much power is it pulling at that speed?

With LPM enabled, 1298 watts at the wall (0.086 J/GH)

Side note:  Pulling EXACTLY the same if I use one of the original APW3 power supplies (the long thin ones) or a new APW3++.  To the watt.

[Hockeybum]

what do you mean? doesnt the 10/23 firmware completely replace the prior version? I dont think they both are being used once you update to the LPM version.

The only reason I call it a patch is because the regular firmware updates were all over 60 MB and this one is less than 1 MB.  It replaces bmminer and a few other files.

I have had no issues at all with the November firmware - other than when the miners get cleaned, they have run nonstop since April with zero issues. I am currently using the 10/23 firmware as well right now. I dont have a way to measure the power consumption yet but I will for sure the difference in my next electric bill. I need to get inline meter set up asap as well.

I've heard of people having issues with the November 17 firmware since it was released but I personally haven't had any issues with it except for a very few problematic miners that only run on fixed frequency.

[ccgllc]

I am running 10/23.  That release is a replacement for 10/22 with 2 known changes:  Its works with Slushpool (not an issue for me, see my sig), and the LPM button actually works.

I also agree with Fanatic26 - if I have ANY problems with a machine, the first thing I did was downgrade to the May 3rd firmware.  Almost always fixed whatever was wrong.

[r01k]

Power Consumption after applying LPM firmware:

• S9 13.5TH/s: -13%
• S9 13.5TH/s: -3% (This miner has alway been power hungry despite being bought together with the other)
• S9 14TH/s:    -8%

Hashrate is about the same, maybe a fraction lower. I'd say the power savings make the hash rate loss worth it.

With the winter approaching here in western NY temps will be consistently below freezing and humidity low, which makes the rigs very happy.

[manomai]

Can't download latest firmware: Antminer s9  

Antminer-S9-LPM-20181023.tar.gz    

https://service.bitmain.com/support/download?product=Antminer%20S9  It Dead

[ccgllc]

Can't download latest firmware: Antminer s9  

Antminer-S9-LPM-20181023.tar.gz    

https://service.bitmain.com/support/download?product=Antminer%20S9  It Dead

Make sure your AV isn't killing it.  Avast does...  (which was the final straw for me, switch to Webroot today)

[MoparMiningLLC]

I've heard of people having issues with the November 17 firmware since it was released but I personally haven't had any issues with it except for a very few problematic miners that only run on fixed frequency.

well that makes sense then, the November one is not for miners with fixed frequency - at least that is how I understand it, as the other versions are for the fixed frequency miners.

[manomai]

Please send Antminer-S9-LPM-20181023.tar.gz to me

thai1812@gmail.com

Thank you

[DefiantMiner]

Bitmain firmware download is working fine here...just say'n.

I converted my S9s on the 23rd using the newest firmware that came out that day.  All run fine with reduced power etc...HOWEVER...I have noticed that a few of them consistently keep loosing hash after many hours running.  I have one 14TH unit that will be down to 11.2TH by then end of a 24hr cycle.  Others can be down 1-1.5TH at that same period.  A reboot brings them back up, but there is certainly a bug in the firmware release, or the ASICboost software.

Anyone else seeing the same phenomenon?

[daniel232388]

Bitmain firmware download is working fine here...just say'n.

I converted my S9s on the 23rd using the newest firmware that came out that day.  All run fine with reduced power etc...HOWEVER...I have noticed that a few of them consistently keep loosing hash after many hours running.  I have one 14TH unit that will be down to 11.2TH by then end of a 24hr cycle.  Others can be down 1-1.5TH at that same period.  A reboot brings them back up, but there is certainly a bug in the firmware release, or the ASICboost software.

Anyone else seeing the same phenomenon?

a few of urs here is facing this issue....

[ccgllc]

Anyone else seeing the same phenomenon?

Not to be cruel, but try reading the thread... its all been stated and restated.

[DefiantMiner]

Note to be cruel, but try reading the thread... its all been stated and restated.

Eh, skipped that last page.  Most were not specific on timeframe for speed loss or which firmware used anyway.  The gist being that fractional yet cumulative hash rate is lost over time, more on some machines than others, automated reboot is only fix for now.

Let me rephrase - anyone have more specific data to share on this phenomenon?

I also noticed that the hash loss was different on each board in the machines affected with some being down more than the other two...and seemed to be consistent over time.  One machine had a single board that was responsible for almost 100% of the loss.  No unusual HW errors or rejects reported.

Anyone still under warranty taking this up with Bitmain support?   (almost choked typing that)

[Hockeybum]

NOV 17 Firmware 1023 LPM on a 13.5 TH/S S9. Here are the readings I have so far.  Definitely decaying hash rate after a few hours.

GH/S RT   GH/S AVG   W/TH      WATTS AT WALL   TIME
13597   13458   87.6      1,274                   START
13600   13607   86.5      1,273                 2 HOURS
13455   13613   86.5      1,273                   3 HOURS
13555   13576   87.0      1,277                   4 HOURS
13404   13571   86.3      1,266                   5 HOURS
13306   13489   86.7      1,265                   9 HOURS
13302   13482   86.7      1,264                  10 HOURS

Edit: Update for 20 Hour data point.  Sorry about the formatting issues, not sure how to fix that

13076        13343        87.5          1,262                20 HOURS

[DefiantMiner]

That's the drop I'm seeing too.  Worse on some machines.

However, I have also data showing some machines running at rated hash (or very close) for 20hrs then dropping off quickly by .5-1.5TH over the following 2hrs.   Other show a steady drop from the start.

It's very strange.

[ccgllc]

Let me rephrase - anyone have more specific data to share on this phenomenon?

When measured over 100s of S9s we are seeing about a 15% power drop with a gradual drop of about 6% hashing.  Individual units vary, so not sure what good "specific data" on individual machines does anyone other than satisfy curiosity.  That, and most either have a single power meter, or measure at the pod level (across hundreds of boxes).  You are correct that a reboot restores the hash rate.

Borderline cards are failing pretty frequently as well with the new firmware.

This is based on my personal experience with (115) S9s and communication with another mining farm testing 10X+ that number.

[MantaMine]

When measured over 100s of S9s we are seeing about a 15% power drop with a gradual drop of about 6% hashing.  Individual units vary, so not sure what good "specific data" on individual machines does anyone other than satisfy curiosity.  That, and most either have a single power meter, or measure at the pod level (across hundreds of boxes).  You are correct that a reboot restores the hash rate.

Borderline cards are failing pretty frequently as well with the new firmware.

This is based on my personal experience with (115) S9s and communication with another mining farm testing 10X+ that number.

I've had five boards fail across four machines (total of seven machines updated) with this update.  Seems way too high.  You try anything that worked to restore them?

[ccgllc]

I've had five boards fail across four machines (total of seven machines updated) with this update.  Seems way too high.  You try anything that worked to restore them?

Try a factory reset (via the reset button).  Accidentally had that happen to a 1023 box last night while it was being moved and it reset back to a July firmware level.  It was on the November firmware prior to upgrading to 1023, so it really does reset it completely (including pointing back to antpool).

[MoparMiningLLC]

Here are my miners that are on Kano.is after 12 hours - they all started at around 13.9 - 14.2 but are now closer to 13.6 - 13.7 (the LPM FW is installed on the first 2 only.)



I have others on other pools but the same effect, so no need for multiple screen shots showing the same thing. They all start losing hashrate after 12 hours or so. And they are not running hot - temps are in the mid 60's to mid 70's

[torontobitcoinmanager]

Here are my findings. See images.
I only tested 3 machines all s9s.
All produced similar savings.
12 times (111.7 A- 109.2 A)=30W savings per machine
or in % 2.2% savings

https://imgur.com/a/7z47inN

Finally we figured out the issue with LP not working. ViaBTC pool is the issue. As soon as we switched 4 test miners to Kano all desired 1 amp savings achieved.

Before AC After AC
6.34           5.57
6.37           5.21
6.49           5.32
6.61           5.57

Before: https://imgur.com/a/8Y0msG7
After: https://imgur.com/a/1DTA5Ug


UPDATE for S9i

We tested the firmware on S9is and it does produce similar savings as for S9s. Moreover, it works with different power supplies.

ViaBTC pool not supported for the October 23rd version of the firmware.