S17 / S17E / T17+ / T17 custom firmware - Free Download only 1% fee

[mikeywith]

Ok so the custom firmware with the current settings is about 4TH less than Bitmain's stock firmware and up to 9 degrees cooler on some chips and average is about 4-5 degrees cooler on the custom firmware, now assuming these run in the same environment this isn't bad at all, what matters more, however, is the efficiency, what is the power consumption per TH on both firmware?

[1715621447]

1715621447

[1715621447]

1715621447

[1715621447]

1715621447

[1715621447]

1715621447

[1715621447]

1715621447

[1715621447]

1715621447

[philipma1957]

please forgive me for delays on reporting this.

But we lost
power
internet
cable and cell phones.  I simply have not had the time to drive up to clifton and measure power.

BTW I can set fans to 85% fan setting works along with volt setting.

I think sometime this coming week I will get back up to Clifton and measure watts.
I can certainly  work with this on overheating gear.

Don't use the numbers yet as I have not had a 24 hour test.  As clifton has had 2 separate crashes due to the power grid.

Curren settings are

1690mv
85% fan speed

oem settings are

1730mv
100% fan speed

I am hoping to show a 36 hour result  on sunday


here is a longer test.

I am going to load full size and leave links to see them.

what is interesting is the old firmware got a bit hot and did auto down clock to not overheat. So the freq is lower for the old firmware.

the new firmware stayed solid at higher clocks and simply runs fine with out a throttle effect.

https://i.imgur.com/IxXQfTD.png


https://i.imgur.com/1ywg6F0.png

new screen shot link

the new firm ware/software tweak

has 100% victory in all categories


                 new------------old

hashrate.  53.46th _____ 53.35th.  at the pool where it counts
HW errors.  562 _______ 1469
HW %.     0.0000 ______ 0.0001
Fan.          6000 _______ 5040
temp.       73.33 _______ 75.66.  note: I used average of 3 temp chip

rejected.    9 __________ 91      note : did not fit on the screen shot

So new set to
1690mv 85% fan
old set to
1730mv 100% fan

new wins in every way

Today. it will get hot I will post again

I suspect fans on new will pick up to 100%

and freq will not down clock to keep gear cool

the old freq will down clock to keep gear cool

so new will have another win.

the only thing now test is exact power used. I think I will go to Clifton to check that next Saturday.

[mikeywith]

Did thierry4wd make any adjustments after the initial installation?

The custom firmware without a doubt will consume slightly less voltage than the stock, also the rejected shares and hardware errors look a lot better on the custom, I guess in the long test the new one will report a slightly higher hashrate on the pool side a well.

Sorry to hear that things are not going fine in the U.S, are things back to normal yet?

[philipma1957]

on the storm end we are fully back to normal. i packed the generators into storage. cleaned the brush and branches and can move on to helping wife recover from last winters (covid-19?) damage.

we are now making appointments to a ear nose throat doctor 👩‍⚕️ to address her 7.5 month long post nasal drip. which has hindered full lung recovery from what ever she had back in jan 2020.

i have been pretty much overwhelmed this year.

personal health
wifes health
covid-19 shutdowns = this affects all of us
george floyd = black lives matter protesters this affect some of us.
large tropcial storm= mostly north eastern usa.

been quite an interesting year.


new info 08/11/2020

So it got smoking hot again and results were the old original bitmain firmware crashed




so for a hot spot the t17e is a real saver.

the settings of 1690mv and 85% fan save power and prevent over heating of t17e gear.

hashrate just about the same. but do favor the new mod.
errors just about the same.     but do favor the new mod.
rejects just about the same.    but do favor the new mod.

this mod allows fan settings of 85, 90, 100%.  There are more choices I only tested three of them listed.
this mod allows volt settings of 1680mv,1690mv. There are more choices I only tested 2 of them.

It looks like NJ is over the storm damage from last week. I will contact thierry4wd about freq settings

thierry4d asked me to test 1695mv and 75% fan.

I will get back later.

[mikeywith]

Won't it be better to test lower values since it's pretty hot there? Anway I would recommend that thierry4d tries to implement an auto-downclock function for when a miner gets pretty hot (temps should be pre-set by the user), and vice-versa for when it's too cool, these should be optional features and not set by default.

I would also suggest you stick to 100% fans, these gears break easily as you are probably aware of by now, so you want them to run as cool as possible, and according to my experience, 25% which is about 1500 RPMs across all 4 fans makes the miner run a lot cooler.

[philipma1957]

the nice thing is it readjusts so it you set to 75% and it is too hot they go upwards towards 100% and get there mid day then drop down to 75% when the sun sets.

the oem firmware is too aggressive.

so they run very near 100% 24/7

so the oem fans will wear out faster and waste power . it saves more then you think since four fans use close to 100+ watts at full speed.

so far i don't see ill effects from setting fans at 75% or 80% or 90%

as for 1680 vs 1690 vs 1695

testing will tell.

the stock gear is pushing 1730mv at or close to 100% fans.

it is not hashing faster
it hashes a fraction less then the new firmware set at 1680

the stock gear in a warm room like mine can not speed fans to cool better because fans are full spped on it.

so the stock gear downclocks  freq in order to cool it.

if it were really cold like 0c outside the stock firmware would be okay.

right now i bumped the new firmware to just a bit more volts 1695mv. so far the new gear is cooler and hashes higher then the old gear.

as long as it does not drop off 1695mv seems better.

now tomorrow it will get hot. So I will check and see if the 1695 is too high.

I can not change freq yet.  so I have not tried super low volts like 1625 or 1600

I also want to do 24-36 hour tests. so the 1695mv test has about twenty four more hours to go.

I would like to try highest setting I can do maybe 1700mv.

I would like to try lowest setting I can do maybe 1600mv.

As for fans maybe 60% in the winter with maybe 1675mv but that will not be anytime soon.

[thierry4wd]

Hi all, i have updated my first post, is now available for all, free download .

[philipma1957]

So far his firmware has worked fine. I am going to Clifton to test two other t17 and t17+

I need access to them as I have two networks.  one from  optimum and 1 from Verizon I will run his on 1 network as a precaution then shift to my other network where most miners are.

[mikeywith]

Thierry I see that you have changed the pricing for this firmware from a fixed fee to a 2% dev fee, mind at first explaining the reason? was it hard to make the firmware run on certain miners and you were worried about piracy?

Also, how does the firmware mine to the dev pool, is it parallel mining where 2% of the hashrate is always directed to your pool or the miner stops from mining to the owner's pool for 2% of the time?

[favebook]

I think it was changed due to lack of people interested in it. But would love to hear actual reason here from him.

It mines for Y amount of time every X hours so it mines total of 2% of 24h which is ~30min. So to answer your question directly, yes, it stops mining on owner's pool for brief amount of time.

[mikeywith]

This is a bad way of going about, depending on when the dev pool is initiated, it could be a lot more than 2% and of course, could be less (with no direct benefit for the miner) this switching structure "assumes" that the miner is on 24/7 which isn't always the case.

Now in a perfect scenario, if the total fees are 30 mins worth of mining, your gears run for 24/7, and the dev pool is on for 1.5 mins every hour (1.5*24 = 30 mins).

if the dev is greedy, he wants to get the 30 mins as fast as possible, they might want 5 mins per hour, even if they were fair with 1.5 mins that starts at minute 00 of every hour.

Time From : Time To
00:00  : 01:00   = 1.5 mins
01:00  : 02:00   = 1.5 mins
02:00  : 03:00   = 1.5 mins
03:00  : 04:00   = 1.5 mins
04:00  : 05:00   = 1.5 mins
05:00  : 06:00   = 1.5 mins

at 5:30 your miner reboots, it goes back online at 5:35, the dev will take another 1.5 mins, a few reboots here and there add up and if you have a dozen gears, it will get even worse.

If the dev is greedy and wants 5 mins an hour to collect the 2% in the first 6 hours of mining, it gets ugly.

00:00  : 01:00   = 5 mins
01:00  : 02:00   = 5 mins
02:00  : 03:00   = 5 mins
03:00  : 04:00   = 5 mins
04:00  : 05:00   = 5 mins
05:00  : 06:00   = 5 mins

so far you have paid 30 out of 30 mins, you would expect until it's 00:00 next day the dev should get nothing, but this is not how it goes since the dev can't keep a record of this info, all they have is your miner run-time and thus when your miner goes back online at say 07:00 , you will still pay another 30 mins and that makes the dev fee 4% instead of 2%.

It all boils down to how these fees are taken, not sure if you remember Dollemizer firmware for S9, was one of the very early custom firmware for S9, the fees were way higher than advertised not just because they were lying about it, but because of how they collected it,  IIRC that firmware was set to start mining to the dev about 2-3 mins after the miner boots up for about 3-5 mins, at the end of the day if for whatever reason you reboot your miner somewhere after the initial 5-6 hours of them attempting to collect all fees upfront, you end up paying twice, I can't recall all the details but I know dev fees were way higher than what they seemed to be, not sure if the devs came up with some solutions to this issue or not.

Having said that, I am not accusing thierry4wd of anything,in fact, I do trust him a lot and I am sure he would try to structure this in the fairest way possible.

[favebook]

I totally agree with you mikeywith, but afaik there hasn't been a better solution with dev fee other than this. IIRC Vnish uses same structure to get his dev fee. I highly doubt it runs dev fee every hour (this would be way too often IMHO) and I doubt it runs dev fee as soon as you run it. But as I said, we should wait for official statement from thierry4wd.

[thierry4wd]

Hi everyone, and thank you for your interest!

I will try to answer your expectations as clearly as possible, I am here for it;) and I am as transparent as possible!

Here are 2 screens, which should already answer most of your questions;)



1: Regarding the single payment, for a license without devfee, I realize that it does not interest anyone in reality, or at least a very small minority ... that's why I decided to integrate 2% cost, but in return the download is free, which allows you to test it for free, absolutely nothing prevents to revert to another firmware thereafter (original included).

2: the fees are extracted every 2 hours as you can see, my priority is to make your miner more efficient, not to fill my pockets as quickly as possible as some do (be careful, I do not quote anyone, some custom firmwares are very good too), and by design, the minor switch on the dev fees one can more than every 2 hours, for about 2min30.

3: The beginning of the fees, in fact, when you install my firmware, you will need to play with it a bit (like any other custom firmwares), try new parameters, new freqs, votlages .... and by design, each time you press "apply" news settings, the count of these 2 hours is reset to zero, by design, for as long as it takes you to polish it, you will still be on your main pool.

4: Confirmation of fees, look at the 2 screens, I will round off the figures, the miner works at 49.5th, if you do (49.5 / 100) x2 = 1 ... now look at the pool report, over 24H, you have 1th extract, which confirms that you have 2%, and no more.

Now look at the minor side, understand that it is necessary to make the report with the "DiffA" (and not the "Accepted"), because the difficulty of the pools comes into play in the calculation, a pool with 52.5K will have more shares than 'one at 65.5K (to use our example).

Please calculate ... we have "1763283525" on the main pool, and "35061760" on the dev pool side, if you do this calculation:

(1763283525/100) x2 = 35265670.

See that we are very close to "35061760", in this case, in reality the miner has extracted very slightly less than 2%, that is explained, the miner extracts the dev pool, and whatever happens, he Switch back to the pool, even if he hasn't finished his share, and therefore, by design, if a share takes 10s to do so, it will be lost.

I hope I have been the most transparent with you, and if you have new questions, I will answer them without problems! Thank you.

PS: I will modify the first post to add the other firmware. I am also awaiting the test from Philipma1957, who I know, that this self good or bad will give as much detail as possible , I will add a link for the T17 by tonight, with a charge of only 1%.

[philipma1957]

I am heading to Clifton NJ.

I will move the t17 the t17+ to the same test network as the t17e.

I will give access to thierry4wd he will set them up and I will test and post results.

I will soon be able to show overclock on the t17+ and the t17

and we are still working on the t17e so that it can do more then just undervolt.

In the interest of transparency my fee for my work will be my modded gear.

1 t17
1 t17e
1 t17+

will not pay a development fee.

I am hoping to get a lot set up tommorrow and I will prep today by loading sd cards to load into the gear.

At the moment my teamviewer access to  my farm is down. I do have access to the two t17e units. as they are on the test network.

the vast majority of my gear is on the verizon network not the optimum network. When I get there I will figure out why teamviewer is down on the verizon network.

I want to talk about a service or firmware mod like this. It is needed due to bitmain locking voltage fans and freq on the gear.

So think of this from the developer theirry4wd viewpoint, bitmain can end his business with a simple new firmware upgrade.  This is why he found it hard to sell it as one time fee. Pay 100 or whatever and then bitmain does new firmware with options. Miners did not want to go one shot fee. A fee every two hours works.  I would think if he has enough people using it he could alter fee to 1.5% .

I am hoping to get all this setup for testing tomorrow.

I will need to burn some unlocking sd cards today. Move all units to test network tomorrow. And fire away.

To update just a bit both new and old t17e are running on test network.

the new runs cooler and the fans are slower. At this point in time I can only alter voltage and fans.

The weather in NJ is cooler so the benefit of undervolt is less. These are the two t17e.

[thierry4wd]

Hi, regarding the testing of my firmware, I appreciate your commitment to do a review, however, for the T17 and T17 + models I would prefer that you use the public link, so that the review is as close as possible to reality

Of course, the T17E is another topic for the moment, and as promised to those who reached out to me to develop the tests, thereafter, you will have access to my firmware without dev fee, if at all are suitable, and that you want to keep them of course;)

[mikeywith]

IIRC Vnish uses same structure to get his dev fee.

Vinish uses a different structure called "parallel mining" so 2% (or whatever the fee is) of your hash rate always mines to the dev pool as long as your gear is online and hashing, so this is a completely different way of handling the dev fees, at least this is what they claim to be using and I have no evidence to confirm nor deny this claim.

... the minor switch on the dev fees one can more than every 2 hours, for about 2min30.

That's pretty fair unless parallel mining is a thing then doing 24/2/2.5mins is pretty fair, you only get your 30 mins if the miner runs for 24 hours, this is way better than some other custom firmware devs who try to get their fees within a couple of hours.

Thank you so much for your explanation, and good luck.

[favebook]

Vinish uses a different structure called "parallel mining" so 2% (or whatever the fee is) of your hash rate always mines to the dev pool as long as your gear is online and hashing, so this is a completely different way of handling the dev fees, at least this is what they claim to be using and I have no evidence to confirm nor deny this claim.

I stand corrected. Haven't checked for while...

[philipma1957]

Testing is starting

Okay testing t17 and t17+

We are also still playing with t17e as we have fan and volt control but no freq control.

Learning curve on under volt under clock.

but the next week will be 90f every day so now is the time to do it.

I set the t17  to freq   580 volts   1640mv
I set the t17+ to freq 585   volts 1660mv



here is a good tune and a bad tune.  it suggests raising 10 mv from 1660mv to 1670mv.  I did 1665mv

[mikeywith]

Great work phill, why does it say "failed" when the miner is hashing and everything seems to be normal? what criteria does the firmware use to evaluate what is a good tune vs bad tune?

Also, I am very interested in seeing how are you doing with power consumption comparison, assuming your goal is not a better temperature and you aim only for efficiency, would you still recommend the firmware?

[philipma1957]

Okay took me a while to understand terminology.

Also we are doing 3 units.

t17
t17 +
t17 e

Failed means it is under tuning. I e it should do 50th and does 49 th

tune okay means it is at or over  the mark.

I am going to fully rate and test the t17 today.

after 10 minutes my latest test with a good meter says tune is okay.
this is full size click to link: https://i.imgur.com/ljcARud.png

it projects that it should do

36018
36583 is what it did after 10 minute test that is higher then projection so tune is okay

this unit is set to freq 605 fans at 38% and mv at 1645 I did this to keep noise levels acceptable for my wife.

full sized image: https://i.imgur.com/mhNVTP8.png

Now I have an hourly meter that does kwatts

at 4:50pm it read 91.28

so 5:50 pm = ?

it should do 40th at 2200 watts = 55 watts a th

since I am at 36.5 th x 55 = 2007-2008 watts

so the meter should move to 93.28 or 93.29 at 5:50

I have found a 10 hour read to be very accurate for power use.

temps are really good for 38% fan speed.

t17 specs are from here: https://support.bitmain.com/hc/en-us/articles/360020445993-T17-Specifications

I am running the t17 at viabtc

full size link: https://i.imgur.com/5mwhmkO.png

missed the 1 hour meter read.

the 2 hour read is as follows

94.98 6:50pm
91.28 4:50 pm
  3.70

3.70 kwatts in 2 hours =  1.85 kwatts or 1850 watts

36410 gh so

1850/36410 = 50.81 watts a th.  that is well under bitmains 55 watts a th

and viabtc says the hourly avg is 37.26

better then the gui.

if the viabtc number holds you get. 1850/37260 = 49.65 watts a th

that is good. if it keeps it up.

I will check back and do more watt readings.

100.46 at 9:50 this is a  1836 watts per hour avg for 5 hours.
 96.80  at 7:50  this is a 1840 watts per hour avg for 3 hours
 94.98  at 6:50  this is a 1850 watts per hour avg for 2 hours
 91.28  at 4:50

1836/36396 = 50.44 watts a th

pool shows 37.37 th or

1836/37700 = 48.70 watts a th

this firmware charges 1% so far it appears rate has been 1.136%

I will show calculations I did for that.

My garage got too hot and with lowered fans unit was too warm.

So at 11:52 I rest the gear at lower freq and lower mv

my meter read 103.79 kwatts last night and now reads 115.69 kwatts

so I used 11.90 kwatts from 11:52 to 7:52 or 11.9/8 = 1.4875 kwatt an hour.

the gear at freq 500 and  mv of 1635 tunes okay does 29.828th on GUI and 30.08th on viabtc pool

so 1487.5/29.828 = 49.869 watts a th. vs 55 watts a th at bitmain firmware
and 1487.5/30.080 =49.451 watts a th on pool stats

so a good improvement in efficiency is happening on the 36th to the 30th levels of hash

I really like the auto fan setting at this lower hash rate of 30th the fans are running at 2520 to 3120. This miner in my garage with a minimum of sound proofing can be used to solo mine.  As for fees

it did 7172 shares over last 8 hours.

7037 at  32.8 diff
1 at    4.1 diff
108 at   4.1 diff
26 at 65.5 diff

the bolding is the dev fee

7037 x  32.8.  =       2308136
      1 x.  4.1.  =                41
  108 x.  4.1.   =            4428
    26 x 65.5.   =          17030
total adjusted shares 2329635

fee is  17030/2329635 =. 0.0073101 or under 1% after 8 hour test that will alter and flux a bit as he does a take out every 2 hours so you can go over and under the 1% fee depending on when you do a reading of the shares

https://i.imgur.com/PoJF9oF.png
https://i.imgur.com/12cg2Cw.png

full-size images above links



I really like the t17 firmware especially with a 1% fee
you can really do some low settings with it if you want it to solo mine on solo.ckpool.org

It will get really quiet. I won't be able to do high settings until I go to the farm with it.

But I am very tempted to set it to freq 300 and 1620mv
I would like to see if fans drop to under 2000 at that setting along with watts under 1000

this firmware will make your t17's into space heaters that mine solo if those setting work.

Once I move the gear to Clifton I can go with a setting like freq 670 and 1680mv  that should do stock speed but save watts and keep temps lower.

120.14 kwatts at 10:52
103.79 kwatts at 11:52. an 11 hour reading

16.35 kwatts

16.35/11 = 1.486 kwatts to do 29.830. or 49.827 watts a th.

I get 29.97 th at the pool or  49.58 watts a th

dev fee was

9209.    x  32.8         3030552
      1.   x    4.1                  41
  108.    x.  4.1               4428
    36.    x. 65.5             23580

23580/3058601 =   0.77% fee

I am done testing it at:

auto fan
freq 500
1630mv

next test will be

auto fan
freq 360
1540 mv

I will then move to the t17+

I will do full speed tests on the t17 when I bring gear to Clifton NJ

we are doing 21.075th
at freq 360
and 1540mv
fans are 2400__2400__3120__3120

chip temps are

54-60-52-58
52-62-53-60
54-59-51-57

room is hot about 90f or 32c

Meter reads
120.72 kwatt at 11:20
122.67 k-watt at   1:20

I will get back to power reads.

that is 1.95 /2 =  .975 k-watts

so 975/21.073 = 46.2677 watts a th  quite a bit lower then stock 55 watts a th.  of course it does 21th but it makes the t17 a quiet  space heater for the winter.

The t17 mod is very effective on all my under clock and under volt tests.

126.00
5:15

131.35
11:20