Author

Topic: Antminer S9 - 23 TH/s possible? (Read 451 times)

legendary
Activity: 4256
Merit: 8551
'The right to privacy matters'
June 15, 2018, 09:35:24 PM
#5
frankly who cares?

I rather have a 14th miner doing 900 watts.

Never ceases to amaze me pedal to the metal mindset never goes away.

2 14th miners pulling only 1800 watts = sexy.

1 23th miner pulling 2300 watts = hot and loud and no real benefit that I see.

I have physical room for 250 miners easy.

now supply the power and keep them cool = hard to do.
sr. member
Activity: 800
Merit: 294
Created AutoTune to saved the planet! ~USA
June 15, 2018, 07:15:22 PM
#4
The chips would be running at about 1070MHz. If power draw was linear (no voltage changes, which seems unlikely) the chips would be pushing almost 10W each. That's not unheard of - S5 ran 10W per chip around 400MHz and people pushed them even farther with less effective heatsinking than the S9 has.

Of course, the S5 was also about one third the heat density of an S9.

The S5 also didn't have any kind of board-level regulator. 10W per chip is looking like over 60A through the main regulator, which already runs hot at around 40A. If the board is immersion-cooled, heat wouldn't be as much of a factor as the saturation current of the main inductor. Once that level is exceeded, at any temperature, voltage stability goes whackadoo and regulator efficiency drops off.

If pushing the S9 to such levels the 1600w bitmain psu would probably need to be swapped out for something more powerful though right?
legendary
Activity: 3374
Merit: 1859
Curmudgeonly hardware guy
June 15, 2018, 04:08:04 PM
#3
The chips would be running at about 1070MHz. If power draw was linear (no voltage changes, which seems unlikely) the chips would be pushing almost 10W each. That's not unheard of - S5 ran 10W per chip around 400MHz and people pushed them even farther with less effective heatsinking than the S9 has.

Of course, the S5 was also about one third the heat density of an S9.

The S5 also didn't have any kind of board-level regulator. 10W per chip is looking like over 60A through the main regulator, which already runs hot at around 40A. If the board is immersion-cooled, heat wouldn't be as much of a factor as the saturation current of the main inductor. Once that level is exceeded, at any temperature, voltage stability goes whackadoo and regulator efficiency drops off.
member
Activity: 658
Merit: 21
4 s9's 2 821's
June 15, 2018, 03:26:34 PM
#2
Although the title may have persuaded you to simply say 23TH/s is impossible, the rumour was started by Kristy (OhGodAGirl) in this interview;

https://youtu.be/A3i3jpJqgoo?t=10m8s

What do you guys think? Is this even remotely possible? As far as I've seen, anything above 15TH/s needs some exotic form of cooling and has insane power usages with much lower efficiency.

somebody knows something, somewhere.   
member
Activity: 125
Merit: 35
June 15, 2018, 06:37:42 AM
#1
Although the title may have persuaded you to simply say 23TH/s is impossible, the rumour was started by Kristy (OhGodAGirl) in this interview;

https://youtu.be/A3i3jpJqgoo?t=10m8s

What do you guys think? Is this even remotely possible? As far as I've seen, anything above 15TH/s needs some exotic form of cooling and has insane power usages with much lower efficiency.
Jump to: