Antminer S9E 16th review

[mikeywith]

So I got 2 S9Es, mainly for the sake of trying them out and to see how they do against the other S9s.

First let's have a look at the design.



Almost the same depth and width as the other S9 versions, but this one has more height on it, it's like 2cm taller , it comes with a closed-top design whereby you have to remove the cover, and these is a small whole for the PSU cables to come in.



You can see the whole on the bottom-left side of the cover, but honestly it's kind of small and it was a pain the ass to try and insert the fat APW7 wires in it, so to make my life easier, I removed the top cover and the miner turned into a beautiful "convertible lambo"

Below is a compassion between

1-S9E
2-an overclocked S9i
3-an S9 running on LPM

in order from top to bottom.



you can see that S9e does 16.17 on average, with temps at 64c , and that is degrees less than an S9i running on LPM, all miners are in the same room, same fan speed and just about same everything except for the fact that this S9E is still brand new and 100% dust free, I however did dust off the other gears to be fair, so I don't expect the S9E to run that much hotter, and it will at least be 10c cooler than the S9i , obviously the S9i runs cooler than the normal S9 13.5th.

I know you probably want to see the power consumption on these, but sorry I didn't do that since I have free power in this location so I don't really bother , but the S9E is advertised as 1280w for 16th.

[1714307799]

1714307799

[1714307799]

1714307799

[DaveF]

It's been a couple of days, anything change or still running good?
Just wondering as sometimes after having new gear for a while you discover something you missed at first.
Thinking of picking up either one of these or the S9k.
The k is slower but a lot less money.

-Dave

[mikeywith]

Yes the S9Es are still running good and cooler than the rest , I do have S9k as well, it has been running just fine, the performance is pretty close to the S9j and S9i , however I heard only bad things about them, someone claimed that bitmain collected all the shitty boards left from the old S9 and put them there, another person said the gear was dead in less than a month.

I can not verify all of these info i read on telegram groups, but i see no reason why would anyone lie about such a thing, also they sell for a bit cheap and they don't go out of stock, they seem suspicious to say the least.  

[thierry4wd]

i thinks the S9K and the S9SE is the SAME MACHINE !!! with new chip (10nm ? 7nm?) because ? just look the freq !!!

S9K and S9SE are 180chips, (the standard S9 are 189chips) ... is the reason i thinks are new chip size ...

For exemple, S9 show arround 14 000ghs with 650mhz
-------------- S9SE show arround 16 000ghs with arround 460mhz
-------------- S9K show arround 14 000ghs with arround 380mhz

For conclusion, i thinks, the S9K and S9SE are same machine, the S9SE are higher frequency and downvoltage compare to S9K !!!

[mikeywith]

This makes a lot of sense, but then if they were the same exact chips, souldn't the S9E run hotter than the S9K? Since they both have the same number of chips but the S9E runs on higher frequency, despite it being undervolted or not, what are the chances that it would run cooler??

[thierry4wd]

be careful to watch the fan ... i have two t9 +, un at 12th, but not the same temperature ... one has the fan has 4000trs and the other t9 + has 5200trs ... and yet the one who has the fan has  4000trs is colder ... it is possible that you post an image of the minor statu of your S9SE and your S9K here?  so that I understand

For exemple ... T9+ and S9 are same chip ... but T9+ run lower and is the temp is higher than S9 ... i thinks on S9se and S9k is same problem ...

I thinks S9se run with high frequency and lower voltage than S9k ...

Just i wait my S9k (ship in 7 days) , and i work to try hack this ...

[mikeywith]

Not really any difference in terms of fans RPM , S9k does 5880 and 5670 , while the S9E does 5280 and 5640 , in fact the fans on the S9k run on higher rpms, both are set to 90% static speed.

The details you requested are below

1-S9E



2-S9K



You are right about them having the same number of the chips, I doubt they are the same chips tho, or it could be, but the S9E is a larger in size , the hash boards on it seem taller ,maybe they have bigger heat sinks on the same chips that's why they run cooler?

[thierry4wd]

Yes, full of parameters comes into account ... the size of the case, the heatsink, the type of fan, the number of propellers, the heatsinks are the same size as usual (the same since the S7 on all minor bitmain) except that there is 2 version of these heatsink, it's the lottery ...

For board size, possible, just like the T9 + and the D3 they are strictly the same !!! Except that there is a row of condenser in addition ... for the reliability? I do not know...

[mikeywith]

I believe the heatsinks on these S9Es are fairly larger than the rest, i removed one of the fans and was able to see the size of the heatsinks, i forgot to take a picture i will do so when i go there again, also i noticed that there is a grill between the fans and the hashboards, it is nothing like the other S9s where you can touch the boards once the intake fan is removed, not sure if that is a good idea but they must have done it for a reason.

They also have a grill attached to the exausht fan which is a bit strange!

[DaveF]

Could be that they are just binning the chips like Intel / AMD / etc do for CPUs or memory manufactures do for RAM
Not good vs bad, but this one runs at "X" speed, this one runs at "Y" speed and this one runs at "Z" speed before producing errors.
This one runs at "X" speed at "A" degrees this one runs at "X" speed at "A+5" degrees etc. due to increased voltage needs.
You then build boards out of those chips that are similar performance & heat generation.

The slower units might have a higher failure rate not because they are "bad" but rather "not as good"

To the end user the result is the same, (a dead miner) but for the manufacturer it's just good policy not to chuck out good but slower chips.

-Dave

[philipma1957]

the lower priced ones could have an 8 month fail rate of 25%  and would be a better deal  then the higher prices one with an 8 month fail rate of 5%

100 units drop to 75 units 100 x 263 =  26,300

100 units drop to 95 units. 100 x 430 = 43,000

So you have a 16,700 head start  with the cheap units.

14th at 1190 watts.
16th at 1280 watts.

14 x100 = 1400  14 x 75 = 1050    1400 + 1050 = 2450/2  you average 1225 ph

16 x 100 = 1600  16 x 95 = 1520   1600 + 1520 = 3120/2 you avaerage 1560 ph

so 1560-1225 = 335th  more hash over 8 months for the s9e

for 8 months  240 x 335 x .00002 = 1.608 btc  or 16562 usd

you would be 138 ahead with the cheap gear.  not counting that you need less power with the cheap gear.

 the cheap gear with a 25% fail rate over 8 months is like 100 + 75 = 175/2  or 87.5 units
the expensive on  with a 5% fail rate over 8 months is like 100 + 95 = 195/2 or 97.5 units.

1.190 x 24 =  28.56 kwatts cheap gear
1250 x 24 =   30 kwatts expensive gear

28.56 x 87.5 x 240 = 599760kwatts
30.00 x 97.5 x 240 = 702000kwatts   lets round this difference to 100000 kwatts in favor of the s9k

at
5 cent power  5000
4 cent power  4000
3 cent power  3000
2 cent power  2000

all favor the s9k    so buying s9k's over s9e  makes sense  if the fail rates are  25% and 5%

[DaveF]

Very good point, that was what I was looking at except with a smaller # of units.

I was just wondering if I was missing something obvious as to how the numbers worked out.

Guess we both had the same idea.

Thanks,
Dave

[philipma1957]

I used the 100 units since the math was easy.

I have 4 s9k's due to ship sept 20-30

2 to my house and 2 to buysolar's place  these should legally beat the trump tax. (i hope)

I would order more  right now but I fear that the second batch gets shipped with the first one and I get trump taxed.

Now in the op's case  I think he has some free power spots  so he could order the s9e's over the s9k's.

[mikeywith]

Now in the op's case  I think he has some free power spots  so he could order the s9e's over the s9k's.

It is actually the exact opposite, It makes more sense to order the S9k over the S9E simply because of cost per TH

S9E = 430$/16th = 26.8$ a TH
S9k = 310$/13.5th = 22.9$ a TH

with shipping, it gets more expensive for the S9E as it's 8kg , its real weight is not 8kg, it's merely a 5.5kg , however most shipping companies like DHL use volume weight which is 8kg for the S9E due to it's height , so this adds about 30-50$ more to the total cost of the S9E.

The only reason I got these 2*S9Es  was to test them and make this review, the same reason why I got the S9k , if it wasn't for the sake of seeing how these new gears work, I would rather stick to S9/S9i/S9j simply because I prefer being able to switch hash boards across miners when needed,I am also thinking of getting into "Fixing S9 hash boards" which means I need to have the same type of chips for most of my gears in order to be able to use the same chips and tools.

 S9/S9i use the same chip
 S9j has it's own chip
S9k/S9E unknown , but most certainly not the same chip as any other S9.

[thierry4wd]

hi, i recevied my S9K today , not power up, but i opened this !!! lol

I Confirm, the hashboard is totaly new disgn (and mikeywith are true, the headsink are diferents) !!! just look this

[mikeywith]

http://www.noelshack.com/2019-38-5-1568993661-20190920-151716.jpg

Looking at this image, it does seem like they might have used the old S9 hash board and filled those empty slots with new chips ? the reason I think so is the color of the heat-sinks, they don't seem to be the same, both size and color are very different, there is really something "strange" about these miners.



There is another strange thing in the price tag as well, the S9k uses 200w less than the S9J with no overt asicboost, and even with Asicboost on the S9j the S9k is still a little bit more efficient , and despite the fact that both can do 14th , the S9j is a little bit more expensive and it goes out of stock real quick, seems like resellers don't want to buy these S9ks for whatever reason !

There are many theories regarding the S9k , the truth however remains hidden somewhere.

[thierry4wd]

Yes of corse ... the S9K / S9SE are newest disgn !!! new board, new chip ... all is new !!!

At this moment, i working for hack this, i take ssh with succes, now i try to hack this... S9 standard is easy ! but S9K is very hard to tweek, is totaly new system, run with CGminer (not BMminer) ...

[madmartyk]

Hoping for a way to adjust the freq on the S9k, they seem to run a bit hot.  Lots of rebooting.

[thierry4wd]

Hum ... i working for this ... i reshearch partership for hack this  

At this time, i'm agree to take control fan with succes (easyest) and agree to enter to "eco mode" ... but not run good, run at only 100mhz (arround 3,5th for 300w) ...

I continue the reshearch , and i found, S9K / S9SE are chip probalby samsung "BM1393" , for comparaison S15 are "BM1391" and S17 ar "BM1397"

PS : here, "chipless" are make a SDcard image for S9K , original firmware with unlock SSH and signature, run perfect   good job for him


UP :

New discover ... i modified the boot ! see my kernel log :

Code:

bring up chain 0
2019-09-30 21:38:19 zynq.c:212:set_reset_hashboard: set_reset_hashboard = 0x0000fff9
2019-09-30 21:38:50 zynq.c:212:set_reset_hashboard: set_reset_hashboard = 0x0000fff8
2019-09-30 21:38:52 driver-btm-soc.c:3810:check_asic_num: without pre_open_core chain 0, find asic num = 60
2019-09-30 21:38:55 driver-btm-soc.c:4633:set_freq_by_chain: chain[0] set freq:425.00
2019-09-30 21:38:55 driver-btm-soc.c:4464:User Mode: chain[0] base freq 425.000000 over the max freq 385.000000 
2019-09-30 21:38:56 driver-btm-soc.c:6133:bring_up_chain: 

I set 425Mhz (overclock) but miner detect error and not hash but if i make clock under 385mhz, miner run good with underclock ! but i don't understand why only chain 0 is down ... i continu my reshearch

[mikeywith]

I set 425Mhz (overclock) but miner detect error and not hash but if i make clock under 385mhz, miner run good with underclock ! but i don't understand why only chain 0 is down ... i continu my reshearch

Would you please throw an update regarding this subject ? I got a few S9ks coming in and I am interested in testing a different firmware, right now I am running the latest Firmware from bitmain > https://service.bitmain.com/support/download?product=Antminer%20S9k , i found out that it works better than the stock firmware but it's has no options to underclock/overclock.