Author

Topic: Bitmain's Released Antminer S9, World's First 16nm Miner Ready to Order - page 206. (Read 531298 times)

E
full member
Activity: 234
Merit: 100
Has anyone tried the firmware from 11-19? I'm giving it a shot on one machine right now. Curious how it is working for others?

Fan speed is much better handled for me - slight variations in speed, gently ramped, instead of the bang-bang controller in the previous autotune FWs.

Seeing much lower reported temps (~10C lower) but I haven't dug into whether that is just a cosmetic change or a change in setpoint for target temp.

Hashrate is good - 13.904avg over 3025 hours on a batch 1 set to 650MHz (theoretical speed 14.004):


legendary
Activity: 4004
Merit: 4656
So Running the S9 on only two boards.
Right now I'm getting 626 and 626 frequency, very balanced.
The miner is showing 8.8TH Avg
But slush pool is reporting only at 7.8
Nice, loss of a whole Terrahash to the interwebs.

There a better mining pool to use?

typically 7.8 to 8.8 difference in your example is just noise (luck affects). it should fluctuate and sometimes be higher than 8.8 and sometimes lower while getting close to what you see on the miner in the longer average.

Also, check your latency.
legendary
Activity: 1050
Merit: 1000
So Running the S9 on only two boards.
Right now I'm getting 626 and 626 frequency, very balanced.
The miner is showing 8.8TH Avg
But slush pool is reporting only at 7.8
Nice, loss of a whole Terrahash to the interwebs.

There a better mining pool to use?
legendary
Activity: 1050
Merit: 1000
There should be a way,
Someone needs to build a controller board to fake the fans response lol.
I'd buy that, But I'm also attacking ducting to mine to pump it into my furnace, But now I'll pick up a 4 inch duct fan to pull full speed, it should increase air flow dramatically.
legendary
Activity: 4382
Merit: 9330
'The right to privacy matters'
We should be able to stick an extra fan at the back controlled by a rheostat to suck the heat out at a steady pace.

I would think two cheap quiet pwm fans could trick the controller

then just send 11 volts at the oem fans

newegg.com/Product/Product.aspx?Item=N82E16835186072
hero member
Activity: 770
Merit: 523
We should be able to stick an extra fan at the back controlled by a rheostat to suck the heat out at a steady pace.
Are the hashboards different between the old and new or just the controller/beagle?
hero member
Activity: 770
Merit: 523
We should be able to stick an extra fan at the back controlled by a rheostat to suck the heat out at a steady pace.
legendary
Activity: 4382
Merit: 9330
'The right to privacy matters'
I got the same problems here, and just a look at the frequencies and temperatures of the boards tells me, that the large differences of frequencies over the boards is the cause of the problem of the fan, so being able to manually adjust fan would be the place wrong place to change something right now. (imo) And keep in mind that the card with the highest temp defines the fan rpm.

(Maybe large difference in temps causes a problem for the control loop of the fan speed)

Just look at freqs and temps in previous post
...
Looking at the miner stats the following was noted for one miner (the other problematic is similar).

1 card seems to do fine freq. 533MHz and temp of 91C. (slightly high)
1 card seems pretty low in freq. 420MHz and temp of 72C. (pretty low)
1 card seems to be quite high freq. 578MHz and a high temp of 102. (very high)
(I can upload screenshots if needed)
...


IMO. An even distribution of frequencies and thereby hashing power over the cards would help much more. It would decrease the chance of failure of cards with high frequency ==> high temps, and would most probably decrease the strong modulation of fan speeds. Be it cause by cycling of the mining process, or wrong interpretation of temperature by the fan speed control loop.



okay the card with freq 420 should be detached you would then  have a 533 card with a 578 card   the spread would only be  45 mz  and fans would be stable.

merely a guess.
legendary
Activity: 1050
Merit: 1000
So I'll deal with it until their crap warranty is up, then let me adjust my own fans, this is annoying as hell lol.
In it's own room in the basement, and I can hear it scream through the floors
donator
Activity: 4760
Merit: 4323
Leading Crypto Sports Betting & Casino Platform
Honestly the fans on this thing are stupid, I can whisper to someone beside it one minute, and the next minute you gotta yell.
Why not find a bloody balance instead of on off?

The answer from Bitmain is that it must have this behavior due to engineering issues, "at the chip level."  Basically, Bitmain has the temperatures rise high and fall on purpose, and they believe this will reduce the number of failed units.  I imagine it also has the side effect of boosting their R4 sales as home miners won't be able to run S9s with this firmware. 

Bitmain...  How about some sort of coupon or hand out for the people who bought an S9 not knowing you switched the firmware to be unusable for home miners?
legendary
Activity: 1050
Merit: 1000
Honestly the fans on this thing are stupid, I can whisper to someone beside it one minute, and the next minute you gotta yell.
Why not find a bloody balance instead of on off?
member
Activity: 167
Merit: 15
I got the same problems here, and just a look at the frequencies and temperatures of the boards tells me, that the large differences of frequencies over the boards is the cause of the problem of the fan, so being able to manually adjust fan would be the place wrong place to change something right now. (imo) And keep in mind that the card with the highest temp defines the fan rpm. (Maybe large difference in temps causes a problem for the control loop of the fan speed)

Just look at freqs and temps in previous post
...
Looking at the miner stats the following was noted for one miner (the other problematic is similar).

1 card seems to do fine freq. 533MHz and temp of 91C. (slightly high)
1 card seems pretty low in freq. 420MHz and temp of 72C. (pretty low)
1 card seems to be quite high freq. 578MHz and a high temp of 102. (very high)
(I can upload screenshots if needed)
...


IMO. An even distribution of frequencies and thereby hashing power over the cards would help much more. It would decrease the chance of failure of cards with high frequency ==> high temps, and would most probably decrease the strong modulation of fan speeds. Be it cause by cycling of the mining process, or wrong interpretation of temperature by the fan speed control loop.

legendary
Activity: 3822
Merit: 2703
Evil beware: We have waffles!
Has anyone tried the firmware from 11-19? I'm giving it a shot on one machine right now. Curious how it is working for others?
Batch 16 is when the autotune was introduced so any after that are autotune only.

As I've said earlier on here IMHO Bitmain forced autotune boost the # of chips they can actually use instead of scrapping under performers. I have a batch 18 with wildly different speeds per-board but in total the miner runs at advertised 11 TH/s. Barely. My guess is they stuff the boards then test to get speed ranges for binning. After that they pick combinations of fast/slow boards to make the miner reach target speed.

ref that b18 miner to the b23 I just got -- on this newest one all boards are running 649.7 to 649.9 MHz to produce 14.129 THs. Obviously the b23 boards were cherry-picked for them.
legendary
Activity: 1050
Merit: 1000

The fan thing and the freq are related as the fan rpm is adjusted to the highest temp of the 3 boards.


I understand that, But they dial right down, nearly off. than crank back up to make turbine speed.
This is why the damn thing is so noisy, if it idled at 80% or so, it wouldn't need to crank up nearly as often, keeping noise down and temp spikes down. The fluctuating temp/speed is annoying.


Check whether your miner is actually restarting, when it cranks down. Or maybe part of the mining process. Am looking at this atm myself. This would explain it a bit.

In theory, when the miner would be constantly running, and one of the cards has a very high temperature, the fan should be running constantly high. Remark here, cranking the fan down manually (if it would be possible) to 80% would probably cause the high temp card to overheat & break for sure. Better would be to have all cards in same frequency range, so all cards have roughly same temperature range, then the automatic adjustment of the fan would work nicely (and probably at lot less noisy).

No not setting a Max, a Min.
I can hear it, and it doesn't turn off, it spins up, then dies right down, then spins back up a little later.
Miner keeps hasing away constantly.
I'd like to see the minimum speed stay higher, maybe it wouldn't kick into turbine mode so often.
member
Activity: 167
Merit: 15
Yes did try it, same problem, and the one time frequency settings only work for miners before the auto-adjustment miner generation of S9s. Mentioned somewhere on this post before too
sr. member
Activity: 546
Merit: 253
Has anyone tried the firmware from 11-19? I'm giving it a shot on one machine right now. Curious how it is working for others?
member
Activity: 167
Merit: 15

The fan thing and the freq are related as the fan rpm is adjusted to the highest temp of the 3 boards.


I understand that, But they dial right down, nearly off. than crank back up to make turbine speed.
This is why the damn thing is so noisy, if it idled at 80% or so, it wouldn't need to crank up nearly as often, keeping noise down and temp spikes down. The fluctuating temp/speed is annoying.


Check whether your miner is actually restarting, when it cranks down. Or maybe part of the mining process. Am looking at this atm myself. This would explain it a bit.

In theory, when the miner would be constantly running, and one of the cards has a very high temperature, the fan should be running constantly high. Remark here, cranking the fan down manually (if it would be possible) to 80% would probably cause the high temp card to overheat & break for sure. Better would be to have all cards in same frequency range, so all cards have roughly same temperature range, then the automatic adjustment of the fan would work nicely (and probably at lot less noisy).
legendary
Activity: 1050
Merit: 1000
I can understand the Frequency thing, for over clocking.. but should allow underclocking...

But the fans? WTF Bitmain, how many complaints has this brought up?
Fix it!
The fan thing and the freq are related as the fan rpm is adjusted to the highest temp of the 3 boards.


I understand that, But they dial right down, nearly off. than crank back up to make turbine speed.
This is why the damn thing is so noisy, if it idled at 80% or so, it wouldn't need to crank up nearly as often, keeping noise down and temp spikes down. The fluctuating temp/speed is annoying.
member
Activity: 167
Merit: 15
As it looks I can open the first warranty case, as one of the boards of 1 of the 2 problematic miners doesn't start anymore.

Great, and all this to faulty firmware of the manufacturer? Cause it is not able to adjust the frequency correctly. And seeing the warranty of Bitmain this is more than frustrating, since the board didn't even run for whole day. Let's see what their reply will be. =(

So, the unit needs to be send back to China at own expenses. Great. Sending from China to us on Bitmains costs, at least something, but seeing that is is more than obviously caused by their own faulty firmware this is outrageous. (And last but not least not having the unit mine for the time.)
member
Activity: 167
Merit: 15
I can understand the Frequency thing, for over clocking.. but should allow underclocking...

But the fans? WTF Bitmain, how many complaints has this brought up?
Fix it!
The fan thing and the freq are related as the fan rpm is adjusted to the highest temp of the 3 boards.

As it looks I can open the first warranty case, as one of the boards of 1 of the 2 problematic miners doesn't start anymore.
Thanks very much for your answer,

Looks like I can prepare two repair tickets already for the time the boards fail, as I assume that the hot boards will not survive long. =(

It would interesting to know whether this is a hardware issue, or a software issue. If software, then Bitmain should try to come up with new firmware pretty soon, as the current will only destroy the miners.
3 new S9 B22 (11TH/s) arrived yesterday.
First miner plugged in, and working fine. Could be better when looking at stats, but okay =D

However, the other two new miners show strange behavior, even after installing latest firmware while closely following instructions.

It is similar to what other people wrote before, constant up and down of fan speed. (which is rather annoying, but in data-center setup, one doesn't hear it, but more worrying is, it doesn't seem to be enhancing the lifetime of miner and ventilators. especially when looking at the stats)

Looking at the miner stats the following was noted for one miner (the other problematic is similar).

1 card seems to do fine freq. 533MHz and temp of 91C. (slightly high)
1 card seems pretty low in freq. 420MHz and temp of 72C. (pretty low)
1 card seems to be quite high freq. 578MHz and a high temp of 102. (very high)
(I can upload screenshots if needed)

In average the hashrate achieves the ideal hashrate of 11 TH/S. However the frequencies and temperature of the low and high card are pretty worrying. Furthermore, actually one expect that the frequencies of the boards would be similar, something like all in range of a freq. 510MHz.

Honestly, being able to set the frequency manually would help. However, other people already stated that this is not possible with the latest firmware. Or is there already a new firmware again, which sets the right frequencies, or gives manual control?

Or is something seriously wrong with the 2 hothead miners?
Best regards


Great, and all this to faulty firmware of the manufacturer? Cause it is not able to adjust the frequency correctly. And seeing the warranty of Bitmain this is more than frustrating, since the board didn't even run for whole day. Let's see what their reply will be. =(
Jump to: