Jump to content
ccgllc

Pro Miner firmware for S9 vs. Bitmain 0319 firmware

Recommended Posts

ccgllc

Hi all.  We recently decided to turn on about 100 of our S9s that have been off since late last November.  As part of our restart due diligence we noticed that Bitmain had posted a few new firmware revisions since we had partially shutdown.  Of interest was their March 04, 2019 (called 0319 from this point on) which stated it fixed the memory leak and subsequent regular reboot requirement that has been present since they introduced Asic-Boost to their firmware last fall.

 

Since we are an Awesome Miner shop, it only took a few minutes to upgrade a test machine to that firmware.  That is when we discovered their Low Power Enhanced Mode.  Performance is way down when you use that option, like from 14TH/s to 9.7TH/s, but energy usage drops even faster.  Awesome Miner, which had been configured with our power cost, indicated our profits would double (from about $0.25/day to $0.50/day per miner).   We were thrilled and quickly told it to upgrade about 80 of the miners we had powered on.  Only took a few minutes and our daily profit projection was looking good!

 

However, we quickly noticed that Awesome Miner was no longer able to update pool information on any of the 0319 boxes.  Did a portscan on those machines and discovered only port 80 (web), 4028 (api), and 6060 (no clue - anybody have one?) were open.  Port 22 for SSH was no longer enabled.  Well that sucked, so we tried to downgrade the firmware back to our standard 1117 release.  Not happening, at least not easily.  Apparently the 0319 firmware will not allow you to downgrade via the web interface.  No problem we thought, will just do a software reset.  Nope.  Tried the reset button.  Nope.  Tried the IP Reporter while powering up approach.  NOPE!!!  Now we are getting pissed.  They took away functionality and locked us out of getting it back.

 

A few days of back-and-forth with Bitmain support resulted in a statement that SSH was removed for security purposes (really?  Its been in every S-series we have owned since 2014...) and that the 0319 firmware has a new signature feature (presumably to ensure only Bitmain firmware is loaded, not potentially hacked firmware).  We were provided with a link that does in fact work, but requires disassembly of your machine in order to switch a jumper, booting from an sdhc card, then switching the jumper back.  

 

 

 

We had been seeing Pro Miner firmware struggling to be recognized on Bitcointalk - apparently the moderators there were deleting the post every day and banning the userid.  We have test machines on isolated networks, and were pretty pissed off about 0319 looking like a win but with a huge catch, so opted to give it a go.  We downloaded it from here and proceeded to upgrade to it via the normal firmware update process on a 1117 based machine.  It came up without a hitch.  We did find ourselves reading the instructions a few times.  Probably best to read them through once, before upgrading, and to have them handy as you browse through the new tabs.

 

The first thing we did was run a port-scan against the machine.  Ports 80, 4028, and 22 (yeah!  SSH is back!) were open.  No sign of the mysterious port 6060 Bitmain uses.

 

The main Pro-Miner screen looks very familiar:

 

aaef8a330a.png

 

We are not going to do a screen by screen review, but will point out a few things:

 

1)  Go to the Administration screen and set the SSH password.  "admin" is the default Bitmain uses, whatever you choose is fine, but will need to be updated in whatever management software you use, if you use any (like Awesome Miner).

 

2)  There is a new Regional settings tab with a pull down menu that lets you set your timezone.  Handy for making sense of your logs without having to do mental UTC conversions all the time.

 

3)  The fun starts in the Miner Configuration tab, which we will dive into a bit here:

 

6f25e9c7c8.png

 

Note the only change here is the "Additional settings".   Pro-Miner does charge 1.8% of your hashrate for its firmware.  We are seeing at least a 10% improvement in our hash efficiencies, so we deem that a great deal, however it does this by occasionally shifting your hash to their "DevFee" pool.  I'm not sure how often this happens, but not seeing anything negative on Slushpool's reports.

 

The Mining profiles page allows you to select a pre-defined profile and basically walk away.  We elected not to do that since we are trying to optimize our profit (doesn't eveyone?).

 

bf381749e1.png

 

You may see Pro Miner advertising they can achieve 75J/TH via tuning.  We will get into that at the end of this post, but just know that is a best case number using the best chips and best PSUs.  We are seeing 78.8J/TH on our test machine, which is still pretty good.

 

OK, onto the Chip Freq Settings tab where the fun begins:

 

6a08fa7bc3.png

 

A couple of things here:

 

1)  AsicBoost is clearly labeled as such - no more "Low Power Mode" ambiguities

2)  You can set all 3 boards as we did, or set them individually if you like.  We did all three because we love the Chip Auto Tune feature that tweaks things anyhow.  What is important is that this is where Chip Auto Tune starts.  From here, it will decrease frequency as it searches for stability.

3)  In the instructions page, they have a table where they recommend 550M and 8.3V as a minimum setting.  That voltage is very safe if that is as far as you want to go.  We chose to embrace Chip Auto Tune, discovering that it will downshift any chips that can't run at 550MHz at 8.0Vs.  Works for us, but note this tuning will likely cause your machines to reboot MANY times over the next few hours as it tunes itself.

 

The Chip Freq Settings tab serves two purposes:  Its a display of the current chip speeds, highlighting anything not set to default values, and an input method for manually setting those speeds (again, we prefer Chip Auto Tune).  It should be noted that even if you do not use Chip Auto Tune, and pay the reboot price, some adjustments will apparently occur automatically anyhow.  We only noticed this on our first look, since after that we used Chip Auto Tune and expected it.

 

fff4a4713e.png

 

Chip Auto Tune is where the magic happens, and what we feels makes this firmware worth having, over and above restoring SSH functionality.

 

faccfae05b.png

 

So tuning happens in 3 stages, called "stage 1", "stage 2" and "stage 3" (duh <smile>)

 

Stage 1 is a first pass where we believe some rough settings are calculated.

 

Stage 2 is what happens over the next 2 hours or so

 

Stage 3 is what is enabled by the 4th entry: Timer.  This setting basically tells the firmware how often it can reboot the machine to address chips that need downscaling.  12 hours is probably a good number, we set it to 3 because we are more interested in seeing results than caring about reboots.

 

As each stage is completed, the box will change from "enabled" to "disabled"

 

So, our basic process after uploading the firmware was to:

 

1)  Set our timezone via System->Regional settings

2)  Turn on AsicBoost via Miner Configuration->Chain Freq Settings

3)   Set All chains frequency to 500M and voltage to 8.0V on the same page

4)   Enable Stage 1 and Stage 2 autotuning on the Chip auto tune tab

5)   Set HW downscaling to 25 HW errors (pull down options are Disable, 5, 10, 25, 50, 100, 250) - 25 seemed reasonable to us

6)   Set Minimum Freq to 400 (this is a slow as the firmware will try and run a chip, note this defaults to 650 which would be a problem given our 550 selection)

7)   Enable "Reset chip freq and clear logs" just because we want to have a clean start

 

We then revert back to routine monitoring via Awesome Miner with the knowledge the machine will likely reboot a lot over the next few hours.

 

Here are our results using a test machine with our power meter on it:

image.png.340d012af77fcda52b68d4c8de7c1143.png
 

and for reference, expected results per Bitmain documentation (although those results don't make sense to us):

image.png.1bad215597f32eef79def36e7a86517e.png

 

As you can see, at least on this test machine, we are achieving over 6% better power efficiency and pushing through over 18% more hash using Prominer vs. BItmains 0319 firmware.

 

We are currently in the processing of auto-tuning an additional 20 S9s.  We will post if we see any significant variations from these results.

 

Carpenter Computing Group, LLC

  • Winner 1

Share this post


Link to post
Share on other sites
Jimmy

Nice writeup!

Share this post


Link to post
Share on other sites
ccgllc

After running 20 more machines for a few hours we have a bit of an update:

 

Not all of our hash cards are handling 8.0V's particularly well.  Most are doing just fine.  One, so far, is looking like this:

 

269e92ac9e.png

 

Where the third card is clearly having problems.  We have bumped the voltage to 8.1V and restarted Auto-tune, which will hopefully clear that card up.

 

We are also seeing a few cards that are green, but have a large number of chips running at 406M (presumably the lowest actually frequency vs. our minimum selection of 400M).  We are bumping those cards voltages up to 8.1V as well and restarting Auto-tune.

 

On the auto-tune tab, we have also now changed Downscale Step to 1, since we are doing fine grain tuning.

 

Be advised that this page takes several seconds to load as it reads information from each chip.  Having the freq and voltage show is good sign the page is done loading, or about to be.

 

We are happy to report great customer service from Pro-Miner, with live, knowledgeable (!!!) people manning the Chat box on their website.

  • Like 1

Share this post


Link to post
Share on other sites
pro_miner
3 часа назад, ccgllc сказал:

Hi all.  We recently decided to turn on about 100 of our S9s that have been off since late last November.  As part of our restart due diligence we noticed that Bitmain had posted a few new firmware revisions since we had partially shutdown.  Of interest was their March 04, 2019 (called 0319 from this point on) which stated it fixed the memory leak and subsequent regular reboot requirement that has been present since they introduced Asic-Boost to their firmware last fall.

 

Since we are an Awesome Miner shop, it only took a few minutes to upgrade a test machine to that firmware.  That is when we discovered their Low Power Enhanced Mode.  Performance is way down when you use that option, like from 14TH/s to 9.7TH/s, but energy usage drops even faster.  Awesome Miner, which had been configured with our power cost, indicated our profits would double (from about $0.25/day to $0.50/day per miner).   We were thrilled and quickly told it to upgrade about 80 of the miners we had powered on.  Only took a few minutes and our daily profit projection was looking good!

 

However, we quickly noticed that Awesome Miner was no longer able to update pool information on any of the 0319 boxes.  Did a portscan on those machines and discovered only port 80 (web), 4028 (api), and 6060 (no clue - anybody have one?) were open.  Port 22 for SSH was no longer enabled.  Well that sucked, so we tried to downgrade the firmware back to our standard 1117 release.  Not happening, at least not easily.  Apparently the 0319 firmware will not allow you to downgrade via the web interface.  No problem we thought, will just do a software reset.  Nope.  Tried the reset button.  Nope.  Tried the IP Reporter while powering up approach.  NOPE!!!  Now we are getting pissed.  They took away functionality and locked us out of getting it back.

 

A few days of back-and-forth with Bitmain support resulted in a statement that SSH was removed for security purposes (really?  Its been in every S-series we have owned since 2014...) and that the 0319 firmware has a new signature feature (presumably to ensure only Bitmain firmware is loaded, not potentially hacked firmware).  We were provided with a link that does in fact work, but requires disassembly of your machine in order to switch a jumper, booting from an sdhc card, then switching the jumper back.  

 

 

 

We had been seeing Pro Miner firmware struggling to be recognized on Bitcointalk - apparently the moderators there were deleting the post every day and banning the userid.  We have test machines on isolated networks, and were pretty pissed off about 0319 looking like a win but with a huge catch, so opted to give it a go.  We downloaded it from here and proceeded to upgrade to it via the normal firmware update process on a 1117 based machine.  It came up without a hitch.  We did find ourselves reading the instructions a few times.  Probably best to read them through once, before upgrading, and to have them handy as you browse through the new tabs.

 

The first thing we did was run a port-scan against the machine.  Ports 80, 4028, and 22 (yeah!  SSH is back!) were open.  No sign of the mysterious port 6060 Bitmain uses.

 

The main Pro-Miner screen looks very familiar:

 

aaef8a330a.png

 

We are not going to do a screen by screen review, but will point out a few things:

 

1)  Go to the Administration screen and set the SSH password.  "admin" is the default Bitmain uses, whatever you choose is fine, but will need to be updated in whatever management software you use, if you use any (like Awesome Miner).

 

2)  There is a new Regional settings tab with a pull down menu that lets you set your timezone.  Handy for making sense of your logs without having to do mental UTC conversions all the time.

 

3)  The fun starts in the Miner Configuration tab, which we will dive into a bit here:

 

6f25e9c7c8.png

 

Note the only change here is the "Additional settings".   Pro-Miner does charge 1.8% of your hashrate for its firmware.  We are seeing at least a 10% improvement in our hash efficiencies, so we deem that a great deal, however it does this by occasionally shifting your hash to their "DevFee" pool.  I'm not sure how often this happens, but not seeing anything negative on Slushpool's reports.

 

The Mining profiles page allows you to select a pre-defined profile and basically walk away.  We elected not to do that since we are trying to optimize our profit (doesn't eveyone?).

 

bf381749e1.png

 

You may see Pro Miner advertising they can achieve 75J/TH via tuning.  We will get into that at the end of this post, but just know that is a best case number using the best chips and best PSUs.  We are seeing 78.8J/TH on our test machine, which is still pretty good.

 

OK, onto the Chip Freq Settings tab where the fun begins:

 

6a08fa7bc3.png

 

A couple of things here:

 

1)  AsicBoost is clearly labeled as such - no more "Low Power Mode" ambiguities

2)  You can set all 3 boards as we did, or set them individually if you like.  We did all three because we love the Chip Auto Tune feature that tweaks things anyhow.  What is important is that this is where Chip Auto Tune starts.  From here, it will decrease frequency as it searches for stability.

3)  In the instructions page, they have a table where they recommend 550M and 8.3V as a minimum setting.  That voltage is very safe if that is as far as you want to go.  We chose to embrace Chip Auto Tune, discovering that it will downshift any chips that can't run at 550MHz at 8.0Vs.  Works for us, but note this tuning will likely cause your machines to reboot MANY times over the next few hours as it tunes itself.

 

The Chip Freq Settings tab serves two purposes:  Its a display of the current chip speeds, highlighting anything not set to default values, and an input method for manually setting those speeds (again, we prefer Chip Auto Tune).  It should be noted that even if you do not use Chip Auto Tune, and pay the reboot price, some adjustments will apparently occur automatically anyhow.  We only noticed this on our first look, since after that we used Chip Auto Tune and expected it.

 

fff4a4713e.png

 

Chip Auto Tune is where the magic happens, and what we feels makes this firmware worth having, over and above restoring SSH functionality.

 

faccfae05b.png

 

So tuning happens in 3 stages, called "stage 1", "stage 2" and "stage 3" (duh <smile>)

 

Stage 1 is a first pass where we believe some rough settings are calculated.

 

Stage 2 is what happens over the next 2 hours or so

 

Stage 3 is what is enabled by the 4th entry: Timer.  This setting basically tells the firmware how often it can reboot the machine to address chips that need downscaling.  12 hours is probably a good number, we set it to 3 because we are more interested in seeing results than caring about reboots.

 

As each stage is completed, the box will change from "enabled" to "disabled"

 

So, our basic process after uploading the firmware was to:

 

1)  Set our timezone via System->Regional settings

2)  Turn on AsicBoost via Miner Configuration->Chain Freq Settings

3)   Set All chains frequency to 500M and voltage to 8.0V on the same page

4)   Enable Stage 1 and Stage 2 autotuning on the Chip auto tune tab

5)   Set HW downscaling to 25 HW errors (pull down options are Disable, 5, 10, 25, 50, 100, 250) - 25 seemed reasonable to us

6)   Set Minimum Freq to 400 (this is a slow as the firmware will try and run a chip, note this defaults to 650 which would be a problem given our 550 selection)

7)   Enable "Reset chip freq and clear logs" just because we want to have a clean start

 

We then revert back to routine monitoring via Awesome Miner with the knowledge the machine will likely reboot a lot over the next few hours.

 

Here are our results using a test machine with our power meter on it:

image.png.340d012af77fcda52b68d4c8de7c1143.png
 

and for reference, expected results per Bitmain documentation (although those results don't make sense to us):

image.png.1bad215597f32eef79def36e7a86517e.png

 

As you can see, at least on this test machine, we are achieving over 6% better power efficiency and pushing through over 18% more hash using Prominer vs. BItmains 0319 firmware.

 

We are currently in the processing of auto-tuning an additional 20 S9s.  We will post if we see any significant variations from these results.

 

Carpenter Computing Group, LLC

 

Thank you for great review !!!

  • Like 1

Share this post


Link to post
Share on other sites
ccgllc

Just learned from support that if your seeing any unusually high hashrates per chip, like 872 or 694 or 973 instead of the more typical 60-something, its an indication that your voltage needs to be bumped.  Normal numbers that are underlined are suspect as well.  Alternatively you can manually lower the frequency of that chip.   

 

We have observed that sometimes this problem goes away after a few seconds.  If the number continually drops we have been ignoring it.  However if it bounces around, we adjust the voltage.

 

003ddab6b6.png

 

Note the 553 and 905 numbers on Chain #7.  Will be boosting the voltage on that board from 8.0 to 8.1 momentarily.

 

After making a board level voltage change, don't forget to restart stage 2 testing.  No need to re-run stage 1, its redundant.  Do reset the chip freq and clear the logs.  If you manually just lower the frequency of those chips you won't need to do that.

 

Personally, if we see any problem chips that are already running at our minimum 406, we are likely to bump the voltage, unless, per-chance, its just one chip having problems on the board.  So far that hasn't been the case for us.

  • Like 1

Share this post


Link to post
Share on other sites
ccgllc

Morning of day 3 report:

 

Apparently changing the Voltage (and presumably Freq) on the Chip Freq Settings tab is a temporary thing that is reset when the server reboots.  Proper way, at least for now, is to modify those settings on the Chain Freq Settings tab.  After chatting with the developers I've come to realize this is actually handy:  You can use Chain Freq Settings to establish a stable, rebootable, environment and then play on the Chip Freq Settings tab with new trial values.  Once satisfied with the results, you can make them permanent on the Chain tab.  This is much safer than what I've been doing, which a half-dozen tanked machines will attest too (permanent settings were too low and cold booting failed).

 

VERY few hash cards will run, or run well, at 8.0V @ 550Mhz.  Thought I'd try but its just too little.  Trying about 50 cards at 8.1V now.  We know that 8.3V works fine, but each 0.1V apparently saves about 25W across 3 hash cards, so if your a tweaker (like me), its fun trying.

 

I do have (3) cards on one machine running at 8.0V @ 475Mhz.  This drops the power usage of the machine down to about 770Ws (up to 790W if the fans kick on). That is an efficiency of 76.7J/TH!  That machine is hashing at around 10100 GH/sec.  Amusingly, a single PCI connection per hash card is sufficient to runs stability.  That in turn will allow for my next post this afternoon hopefully.

 

Minor hint:  Always wait for the little blue "saving" popup to appear, disappear, pause for a bit, and then the little blue "Settings saved" popup to appear before changing tabs.

  • Like 1

Share this post


Link to post
Share on other sites
ccgllc

We have had two goals while testing this firmware:  1)  Meet or exceed the Bitmain March 2019 firmware's Low Power Enhanced Mode performance and energy savings and 2) Try and find settings that approached the mythical (at least to us) 75J/TH for an S9 unit.

 

Goal #1 was easy:  Prominer firmware set at 550Mhz at 8.3V.  Pretty much a match for the Bitmain firmware in energy savings (within 1% in our test).

 

Goal #2 is in progress but is likely going to have an S9 run at 475Mhz around 8.0V - 8.1V.  We could downclock even further, like into the Bitmains 417Mhz range, but so far don't see the need.  Might try it someday just out of curiosity.

 

So far we can report that after an auto-tune:

 

[email protected] - few cards will work at all, and when they do the chips are very often downclocked, some machines don't reboot well

[email protected] - most cards work but have significant numbers of downclocked chips, most  machines rebooted fine (all but 1 of 20 in our case, and the one was a known problem child)

[email protected] - mixed results.  Some cards work fine, some have significant numbers of downclocked chips.  Fine place to start if you don't mind tweaking up the cards with issues later.

[email protected] - very stable, only a few chips, if any, downclocked.  Best place to start if you want to set and forget.

 

[email protected] - Mixed results.  Some cards run fine with only a few downclocked chips, some cards downclock almost all the chips.

[email protected] - Better for those cards that don't run at 8.0V, still upwards of a third or so downclocked chips, but most only a few steps.

  • Like 1

Share this post


Link to post
Share on other sites
ccgllc

Day 4 report:

 

This has become our "go-to" firmware of choice.  We currently have 38 machines it and expect to migrate another 80ish over.  Alas, we made the mistake of installing Bitmains 2019 firmware, so those 80 will need to be disassembled and sdhc reverted to factory defaults:

 

https://support.bitmain.com/hc/en-us/articles/360019493654-S9-series-S9-S9i-S9j-S9-Hydro-Control-Board-Program-Recovery

 

At least it will give us a chance to clean them while doing so...

 

We have noticed a few things along the way:

 

1)  Not all antminer hash cards will accept voltage tweaks.  They appear to, but we don't see any improvement in power consumption on those machines.  Its almost like they have a built-in minimum power setting.

 

2)  #1 above complicated a cute thing we saw an image off from the Prominer people and wanted to try:  (2) S9s running off a single APW3++ power supply.  We got lucky with our test unit, it was well behaved and ran at [email protected] drawing about 775W at the wall.   Not all of our 14TH S9s ran well at the voltage, many required a card to two to be bumped up to 8.3V.  Several pulled 900Ws at 475M no matter what we set the cards to.  We did confirm on our test unit that a single PCIe connector per hash card powered that card fine at [email protected]  Please remember to set the speed and voltage BEFORE you try this, and confirm they took with a power meter.  Had we found a 2nd unit that behaved as well, it should have been trivial to run them both off of one PSU.  We expected the increased load on the PSU to improve its efficiency, as per most PSUs, so that perhaps instead of drawing 775W X 2 = 1550W we would see something under 1500W (a few percent better).  If so, we would have been running, combined, over 20TH for that 1500W or 75W/J.  Alas, we ran out of time and moved on.

 

3)  Most of our 14TH rated machines run fine at 8.2 volts, a few cards need 8.3V at 550M.

      Most of our 13.5TH rated machines have a few cards that run well at 8.2V, but most require 8.3V at 550M.

      Most of our later batch 13TH rated machines require at least 8.3V, many require 8.4V, and a few cards require 8.5V at 550M.

 

4) For us, #3 confirms what we have always suspected - the only difference between S9 machines is the quality of the hash cards in them.  This is similar to the way CPU manufactures rate chips - they try them at high speed, if they work they sell them for a premium.  If not, they try them at a lower speed, and price accordingly repeating until they hit a lowest sellable speed.

  • Like 1

Share this post


Link to post
Share on other sites
ccgllc

Spent a few hours converting about a dozen Bitmain 0319 firmware S9s over to Pro-Miner firmware this morning.  Process is time consuming, but works:

 

1) Unrack machine and take to test bench

2) Disassemble removing the back fan and all three hash cards.  Use compress air to clean hash cards and fan assemblies (unplugged fans only!)

3) Slide out the control card, move jumper 4, insert above referenced sdhc card, place control card on something insulting (we use a piece of wood), plug in power and power up PSU

4) Wait for the control card to flash red and green lights together (takes a minute or two, ignore the first blink of the lights)

5) Unplug power from control card, remove sdhc card, restore jumper 4

6) Reassemble machine and power up

7) Watch DHCP server issue a new lease (e.g. IP address)

 

Next steps are via Awesome Miner for us, but can be done other ways

 

7) Run AM's New Miner script search for a new machine.  Casually verify its the same as the new DHCP lease.

8 ) Set default pools for new miner

9)  Set API access for new miner

10) Update firmware to Prominer for new miner

11) Edit profile and set a worker name and profit profile

 

From a browser:  (note that one should pause between the Pro-Miner steps below after a "save" and wait for the "settings saved" confirmation.

 

12) Open a window to the new IP address

13) Set System->Regional settings->TimeZone

14) Save Miner Configuration->General Settings=>Miner General Configuration so that it reformats the configuration file and removes the red "old format" warning

15) Set Miner Configuration->Chain Freq Settings, turning on AsicBoot, and setting the "All Chains" frequency to 550M and 8.2V (for 14TH machines, 13.5s will likely do better at 8.3V and 13TH will likely require at least 8.3V or 8.4V)

16)  Save and wait a minute or so for the "settings saved" popup

17) Set Miner Configuration->Chip auto tune, enabling Stage 2, setting Downscale if HW more (we use 25), set the Stage 3 timer (we use 3 hours, some prefer 12), set the minimum downscale frequency (we use 400M), and set "Reset Chip freq and clear log"

18)  Save

19) Power down and move machine back to the rack

 

20) Monitor Miner Configuration->Chip Freq Settings.  If no chain has many downscales (a few is fine), your done.  Verify acceptable performance via Miner Status.  If no chain has ANY downscales, you may want to try a lower voltage and save about 25W per 0.1V over the 3 chains.  If chains have many or significant downscales, or underlined numbers, bump the voltage up 0.1V and restart auto-tune.  Wait about 10 minutes and then continue monitoring and adjust until all the chains appear healthy.

 

When your done, you will have optimized your machine to be the best it can be.  We are finding many of our machines running at similar J/TH as the Bitmain 0319 firmware's Low Power Enhanced Mode provides but are getting about 20% more hash per miner at that power efficiency level.

 

  • Like 1

Share this post


Link to post
Share on other sites
ccgllc

Pro-miner has released version 3.8.3 of their firmware, available at https://pro-miner.ru/catalog/proshivki/firmware-antminer-s9.html 

 

Besides the normal routine bug fixes and optimizations they have added a "Find Miner" button that causes the Red Fault light to blink in a unique pattern, making it easy to find a miner in a rack.

 

They also added the ability to select which DevFee server should be used by world area for anyone having connection problems (they do charge  2% as a development fee, although they typically get a bit less than 1.8% effectively - see the post below for details).

 

  • Like 2

Share this post


Link to post
Share on other sites
ccgllc

Had a conversation with the firmware developer today around how they collect their fee.  It was very educational.

 

If you look at the Miner Status page and do the math against the DiffA# (the amount of work done for a pool), you will see your defined pools plus two DevFee pools.  Doing the math, it appears that the firmware charges right at 2%.  They do this by running the DevFee pools in parallel with primary pools issuing Getworks to both.  For every 102 units of work done for the primary pool, 2 units of work are done for the DevFee pool AT THE CHIP LEVEL.  However, they only do their work when the chips are idle between primary pool Getwork requests.  So although they request 2%, they often terminate early and tell me they actually effectively get a bit under 1.8% over time.

 

I've recommend they start advertising the 2%, rather than 1.8%, since those are the numbers shown on the status screen.  

 

In other Pro-miner news, I noticed after updating that any miner previously pointed at an alt-coin Yiimp based pool was showing that pool as down.  Turns out that Yiimp doesn't support ASIC Boost.  The Bitmain firmware appears to work against said pools with LPM and/or LPEM mode enabled, but it actually dynamically disables ASIC Boost if it sees a pool is down, in an attempt to get it to work.  This is hazardous to the hardware if your overclocking based on the assumption ASIC Boost is being used.  Although I haven't tried this, yet, they tell me this is easily verified running under Bitmain firmware by simply hooking up a power meter and disconnecting the ethernet cable - power usage will jump as the firmware disables ASIC Boost thinking the pool is rejecting it.

 

The Pro-Miner folks have taken a safer approach, and simply leave such pools down.  Turns out most of their user base is in locations with cheap power and they tend to use the firmware for overclocking, not underclocking as I have been reporting on.  For them, rather than chance a failover to a non-ASIC Boost pool, and burning out a power supply that was previously maxed, it is better to simply mark that pool as dead.  Makes sense to me and is a whole lot more transparent.

  • Like 2

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

Cryptocurrenytalk Logo

 

News, information, and discussions about cryptocurrencies, blockchains, technology, and events. Blockchaintalk is your source for advice on what to mine, technical details, new launch announcements, and advice from trusted members of the community. Cryptocurrencytalk is your source for everything crypto. We love discussing the world of cryptocurrencies.

 

   
×
×
  • Create New...

Important Information

By using CRYPTOCURRENCYTALK.COM, you agree to our Terms of Use.