Author

Topic: S-11 Speculation Thread. (Read 1464 times)

legendary
Activity: 1498
Merit: 1030
June 04, 2016, 01:51:45 AM
#19
45 TH in ~2600 W , or whatever a 2880 can handle or 2x 1600 / 3x 1200 / 4x 750  without issue.

price of , day of release,  3100. this is like Keno , but fun to see end of story.

could go with .05 watt/Giga.


 Why would Bitmain care about what a used 2880 Bladecenter PS can handle? They don't sell those.

I suspect their new chip might be capable of that - at the bottom end of it's operating voltage range, where Bitmain NEVER runs their miners.


Oh yeah, my speculation.

 S11 will continue to be a 3-board design in a form factor similar to the S7/S9.

 25.3 TH plus/minus 5% at 1487 watts +7% at the wall using their 1600 watt PS (93% efficienty).
 Initial shipping date 16 Febuary 2020.
 Initial batch cost 4.3 Bitcoin, or whatever that equates to in Dollars = or $3200 dollars, whichever is MORE.

 The CHIP (BM1389) will be capable of 0.045 J/GH but they won't run it at it's highest efficiency point or anywhere close to that.


 I'm giving it 50/50 they'll come up with a BM1388 chip in late 2017, squeezing another 25-35% efficiency from it through improvements in process node tech and more work at optimisation - in which case there will be an S11 much sooner, but it will be more of an S5 to S7 transition on the same node with quite a bit lower percentage improvement on specs.


 BTW - anyone else notice that the BM1386 is missing in action? Gotta wonder if they started work on a 14/16nm not-full-optimised chip design and aborted it when they realised it wouldn't be efficient enough over the BM1385 to be worth producing (Hello LK-1401)....


I think the 3 board design is here to stay aswell for them to have it on S7 and now S9.   I personally like it with a mining area, but can see why home miners would want a more S3'ish miner.    But likely remain odd numbers aswell I don't see a full all in one like S4's coming again.  I think selling AntMiner APW3-12-1600 separate is a pretty good bet for them as that is heck of a add-on sale if you can get people to buy PSU with miner.

But as far as efficiency I think they will manage to squeeze some more efficiency out of S9 chips.   Going to guess 30-45ish percent increase enough where it where people want it  basically.

And i'm pretty sure Bitmain will stick to current TDP. 1400W~ make sense and that by itself take up a whole typical 120v circuit.

I also think they will come up with something else before the end of the year. Maybe they will keep churning out S9's... but i don't know. I'd be inclined to see them release a S9+ thats the same thing, but lower voltage, so they can farm more money off silly miners.

could be even lower, you know S11 won't be as fast as mid 2017, it will be within mid 2018 as they are probably introducing a S? model each year.
i mean, the watt power: assume that kind of miner could go lower than 1200 Kw.
[/quote]

 I believe they might come up with a S9 lite model, 2 boards like the current S7 varient.

 I'm pretty SURE they'll do a S9+ as a "triple" miner similar to the S5+ since they'll have plenty of time to do varients on the S9 theme.

 They probably won't release a "single unit" form factor miner every that can't be powered by a single one of their 1600 watt PS, but multi-unit types they've already done and probably will do again.

 However, if they come up with a new 14/16nm chip with 25%+ better efficiency than their current BM1387, it won't go into an "S9" model - that's inconsistant with their past practice. It'll be a "supprise we're early" S11 model.


 Gotta wonder how many of those S7 lite model things they managed to sell before the S9 announcement pretty much killed that design.

legendary
Activity: 1582
Merit: 1006
beware of your keys.
June 03, 2016, 07:09:54 PM
#18
could be even lower, you know S11 won't be as fast as mid 2017, it will be within mid 2018 as they are probably introducing a S? model each year.
i mean, the watt power: assume that kind of miner could go lower than 1200 Kw.

The Chinese tend to always go for the bang for buck approach. So i'd assume they would raise the voltage to bring the AC draw to 1.4kW~
I don't think S11's are 2 years away. I'm pretty sure they could make a better chip on 14nm still. Or just print tons of them and make high chip low voltage units. 0.06J/GH is mostly likely possible already.

14nm is gonna be mainstream even when 10nm is carried out to the majority, look at intel which terminated tick-tock into processor architecture optimization in early 2016 IIRC. however the high-end machine will need 10nm, so i assume that the S11 uses 10nm.

45 TH in ~2600 W , or whatever a 2880 can handle or 2x 1600 / 3x 1200 / 4x 750  without issue.

price of , day of release,  3100. this is like Keno , but fun to see end of story.

could go with .05 watt/Giga.

Oh yeah, my speculation.

 S11 will continue to be a 3-board design in a form factor similar to the S7/S9.

 25.3 TH plus/minus 5% at 1487 watts +7% at the wall using their 1600 watt PS (93% efficienty).
 Initial shipping date 16 Febuary 2020.
 Initial batch cost 4.3 Bitcoin, or whatever that equates to in Dollars = or $3200 dollars, whichever is MORE.

 The CHIP (BM1389) will be capable of 0.045 J/GH but they won't run it at it's highest efficiency point or anywhere close to that.


 I'm giving it 50/50 they'll come up with a BM1388 chip in late 2017, squeezing another 25-35% efficiency from it through improvements in process node tech and more work at optimisation - in which case there will be an S11 much sooner, but it will be more of an S5 to S7 transition on the same node with quite a bit lower percentage improvement on specs.


 BTW - anyone else notice that the BM1386 is missing in action? Gotta wonder if they started work on a 14/16nm not-full-optimised chip design and aborted it when they realised it wouldn't be efficient enough over the BM1385 to be worth producing (Hello LK-1401)....


I think the 3 board design is here to stay aswell for them to have it on S7 and now S9.   I personally like it with a mining area, but can see why home miners would want a more S3'ish miner.    But likely remain odd numbers aswell I don't see a full all in one like S4's coming again.  I think selling AntMiner APW3-12-1600 separate is a pretty good bet for them as that is heck of a add-on sale if you can get people to buy PSU with miner.

But as far as efficiency I think they will manage to squeeze some more efficiency out of S9 chips.   Going to guess 30-45ish percent increase enough where it where people want it  basically.

And i'm pretty sure Bitmain will stick to current TDP. 1400W~ make sense and that by itself take up a whole typical 120v circuit.

I also think they will come up with something else before the end of the year. Maybe they will keep churning out S9's... but i don't know. I'd be inclined to see them release a S9+ thats the same thing, but lower voltage, so they can farm more money off silly miners.

could be even lower, you know S11 won't be as fast as mid 2017, it will be within mid 2018 as they are probably introducing a S? model each year.
i mean, the watt power: assume that kind of miner could go lower than 1200 Kw.

question is , is it chip advances ? board ? combo of ? and how much of that is power savings vs "delivered strength in acceptable envelope" ?

not sure, but look forward with 10nm as mentioned above.
full member
Activity: 219
Merit: 100
Bitcoin Mining Hosting
June 03, 2016, 07:01:16 PM
#17
45 TH in ~2600 W , or whatever a 2880 can handle or 2x 1600 / 3x 1200 / 4x 750  without issue.

price of , day of release,  3100. this is like Keno , but fun to see end of story.

could go with .05 watt/Giga.

Oh yeah, my speculation.

 S11 will continue to be a 3-board design in a form factor similar to the S7/S9.

 25.3 TH plus/minus 5% at 1487 watts +7% at the wall using their 1600 watt PS (93% efficienty).
 Initial shipping date 16 Febuary 2020.
 Initial batch cost 4.3 Bitcoin, or whatever that equates to in Dollars = or $3200 dollars, whichever is MORE.

 The CHIP (BM1389) will be capable of 0.045 J/GH but they won't run it at it's highest efficiency point or anywhere close to that.


 I'm giving it 50/50 they'll come up with a BM1388 chip in late 2017, squeezing another 25-35% efficiency from it through improvements in process node tech and more work at optimisation - in which case there will be an S11 much sooner, but it will be more of an S5 to S7 transition on the same node with quite a bit lower percentage improvement on specs.


 BTW - anyone else notice that the BM1386 is missing in action? Gotta wonder if they started work on a 14/16nm not-full-optimised chip design and aborted it when they realised it wouldn't be efficient enough over the BM1385 to be worth producing (Hello LK-1401)....


I think the 3 board design is here to stay aswell for them to have it on S7 and now S9.   I personally like it with a mining area, but can see why home miners would want a more S3'ish miner.    But likely remain odd numbers aswell I don't see a full all in one like S4's coming again.  I think selling AntMiner APW3-12-1600 separate is a pretty good bet for them as that is heck of a add-on sale if you can get people to buy PSU with miner.

But as far as efficiency I think they will manage to squeeze some more efficiency out of S9 chips.   Going to guess 30-45ish percent increase enough where it where people want it  basically.

And i'm pretty sure Bitmain will stick to current TDP. 1400W~ make sense and that by itself take up a whole typical 120v circuit.

I also think they will come up with something else before the end of the year. Maybe they will keep churning out S9's... but i don't know. I'd be inclined to see them release a S9+ thats the same thing, but lower voltage, so they can farm more money off silly miners.

could be even lower, you know S11 won't be as fast as mid 2017, it will be within mid 2018 as they are probably introducing a S? model each year.
i mean, the watt power: assume that kind of miner could go lower than 1200 Kw.

question is , is it chip advances ? board ? combo of ? and how much of that is power savings vs "delivered strength in acceptable envelope" ?
legendary
Activity: 1302
Merit: 1068
June 03, 2016, 06:57:47 PM
#16
could be even lower, you know S11 won't be as fast as mid 2017, it will be within mid 2018 as they are probably introducing a S? model each year.
i mean, the watt power: assume that kind of miner could go lower than 1200 Kw.

The Chinese tend to always go for the bang for buck approach. So i'd assume they would raise the voltage to bring the AC draw to 1.4kW~
I don't think S11's are 2 years away. I'm pretty sure they could make a better chip on 14nm still. Or just print tons of them and make high chip low voltage units. 0.06J/GH is mostly likely possible already.
legendary
Activity: 1582
Merit: 1006
beware of your keys.
June 03, 2016, 06:54:30 PM
#15
45 TH in ~2600 W , or whatever a 2880 can handle or 2x 1600 / 3x 1200 / 4x 750  without issue.

price of , day of release,  3100. this is like Keno , but fun to see end of story.

could go with .05 watt/Giga.

Oh yeah, my speculation.

 S11 will continue to be a 3-board design in a form factor similar to the S7/S9.

 25.3 TH plus/minus 5% at 1487 watts +7% at the wall using their 1600 watt PS (93% efficienty).
 Initial shipping date 16 Febuary 2020.
 Initial batch cost 4.3 Bitcoin, or whatever that equates to in Dollars = or $3200 dollars, whichever is MORE.

 The CHIP (BM1389) will be capable of 0.045 J/GH but they won't run it at it's highest efficiency point or anywhere close to that.


 I'm giving it 50/50 they'll come up with a BM1388 chip in late 2017, squeezing another 25-35% efficiency from it through improvements in process node tech and more work at optimisation - in which case there will be an S11 much sooner, but it will be more of an S5 to S7 transition on the same node with quite a bit lower percentage improvement on specs.


 BTW - anyone else notice that the BM1386 is missing in action? Gotta wonder if they started work on a 14/16nm not-full-optimised chip design and aborted it when they realised it wouldn't be efficient enough over the BM1385 to be worth producing (Hello LK-1401)....


I think the 3 board design is here to stay aswell for them to have it on S7 and now S9.   I personally like it with a mining area, but can see why home miners would want a more S3'ish miner.    But likely remain odd numbers aswell I don't see a full all in one like S4's coming again.  I think selling AntMiner APW3-12-1600 separate is a pretty good bet for them as that is heck of a add-on sale if you can get people to buy PSU with miner.

But as far as efficiency I think they will manage to squeeze some more efficiency out of S9 chips.   Going to guess 30-45ish percent increase enough where it where people want it  basically.

And i'm pretty sure Bitmain will stick to current TDP. 1400W~ make sense and that by itself take up a whole typical 120v circuit.

I also think they will come up with something else before the end of the year. Maybe they will keep churning out S9's... but i don't know. I'd be inclined to see them release a S9+ thats the same thing, but lower voltage, so they can farm more money off silly miners.

could be even lower, you know S11 won't be as fast as mid 2017, it will be within mid 2018 as they are probably introducing a S? model each year.
i mean, the watt power: assume that kind of miner could go lower than 1200 Kw.
full member
Activity: 219
Merit: 100
Bitcoin Mining Hosting
June 03, 2016, 06:52:22 PM
#14
45 TH in ~2600 W , or whatever a 2880 can handle or 2x 1600 / 3x 1200 / 4x 750  without issue.

price of , day of release,  3100. this is like Keno , but fun to see end of story.
legendary
Activity: 1302
Merit: 1068
June 03, 2016, 06:49:21 PM
#13
Oh yeah, my speculation.

 S11 will continue to be a 3-board design in a form factor similar to the S7/S9.

 25.3 TH plus/minus 5% at 1487 watts +7% at the wall using their 1600 watt PS (93% efficienty).
 Initial shipping date 16 Febuary 2020.
 Initial batch cost 4.3 Bitcoin, or whatever that equates to in Dollars = or $3200 dollars, whichever is MORE.

 The CHIP (BM1389) will be capable of 0.045 J/GH but they won't run it at it's highest efficiency point or anywhere close to that.


 I'm giving it 50/50 they'll come up with a BM1388 chip in late 2017, squeezing another 25-35% efficiency from it through improvements in process node tech and more work at optimisation - in which case there will be an S11 much sooner, but it will be more of an S5 to S7 transition on the same node with quite a bit lower percentage improvement on specs.


 BTW - anyone else notice that the BM1386 is missing in action? Gotta wonder if they started work on a 14/16nm not-full-optimised chip design and aborted it when they realised it wouldn't be efficient enough over the BM1385 to be worth producing (Hello LK-1401)....


I think the 3 board design is here to stay aswell for them to have it on S7 and now S9.   I personally like it with a mining area, but can see why home miners would want a more S3'ish miner.    But likely remain odd numbers aswell I don't see a full all in one like S4's coming again.  I think selling AntMiner APW3-12-1600 separate is a pretty good bet for them as that is heck of a add-on sale if you can get people to buy PSU with miner.

But as far as efficiency I think they will manage to squeeze some more efficiency out of S9 chips.   Going to guess 30-45ish percent increase enough where it where people want it  basically.

And i'm pretty sure Bitmain will stick to current TDP. 1400W~ make sense and that by itself take up a whole typical 120v circuit.

I also think they will come up with something else before the end of the year. Maybe they will keep churning out S9's... but i don't know. I'd be inclined to see them release a S9+ thats the same thing, but lower voltage, so they can farm more money off silly miners.
legendary
Activity: 1456
Merit: 1000
June 03, 2016, 02:58:50 PM
#12
Oh yeah, my speculation.

 S11 will continue to be a 3-board design in a form factor similar to the S7/S9.

 25.3 TH plus/minus 5% at 1487 watts +7% at the wall using their 1600 watt PS (93% efficienty).
 Initial shipping date 16 Febuary 2020.
 Initial batch cost 4.3 Bitcoin, or whatever that equates to in Dollars = or $3200 dollars, whichever is MORE.

 The CHIP (BM1389) will be capable of 0.045 J/GH but they won't run it at it's highest efficiency point or anywhere close to that.


 I'm giving it 50/50 they'll come up with a BM1388 chip in late 2017, squeezing another 25-35% efficiency from it through improvements in process node tech and more work at optimisation - in which case there will be an S11 much sooner, but it will be more of an S5 to S7 transition on the same node with quite a bit lower percentage improvement on specs.


 BTW - anyone else notice that the BM1386 is missing in action? Gotta wonder if they started work on a 14/16nm not-full-optimised chip design and aborted it when they realised it wouldn't be efficient enough over the BM1385 to be worth producing (Hello LK-1401)....


I think the 3 board design is here to stay aswell for them to have it on S7 and now S9.   I personally like it with a mining area, but can see why home miners would want a more S3'ish miner.    But likely remain odd numbers aswell I don't see a full all in one like S4's coming again.  I think selling AntMiner APW3-12-1600 separate is a pretty good bet for them as that is heck of a add-on sale if you can get people to buy PSU with miner.

But as far as efficiency I think they will manage to squeeze some more efficiency out of S9 chips.   Going to guess 30-45ish percent increase enough where it where people want it  basically.
legendary
Activity: 1498
Merit: 1030
June 03, 2016, 12:52:33 AM
#11
Oh yeah, my speculation.

 S11 will continue to be a 3-board design in a form factor similar to the S7/S9.

 25.3 TH plus/minus 5% at 1487 watts +7% at the wall using their 1600 watt PS (93% efficienty).
 Initial shipping date 16 Febuary 2020.
 Initial batch cost 4.3 Bitcoin, or whatever that equates to in Dollars = or $3200 dollars, whichever is MORE.

 The CHIP (BM1389) will be capable of 0.045 J/GH but they won't run it at it's highest efficiency point or anywhere close to that.


 I'm giving it 50/50 they'll come up with a BM1388 chip in late 2017, squeezing another 25-35% efficiency from it through improvements in process node tech and more work at optimisation - in which case there will be an S11 much sooner, but it will be more of an S5 to S7 transition on the same node with quite a bit lower percentage improvement on specs.


 BTW - anyone else notice that the BM1386 is missing in action? Gotta wonder if they started work on a 14/16nm not-full-optimised chip design and aborted it when they realised it wouldn't be efficient enough over the BM1385 to be worth producing (Hello LK-1401)....

legendary
Activity: 4256
Merit: 8551
'The right to privacy matters'
June 02, 2016, 07:32:36 PM
#10
Link to this alleged S9 controller board?

(Edit)

 Nevermind, found it - only 4 connectors populated, but I do see the holes for the other 16. I DO wonder why they populated 4 connectors though if they only need 3 for the hashboards?

I would guess some in house ones may have all 20 populated.



Maybe leaving four is easy compared to leaving 3 due to assembly techniques.



Yes 4 and 20 both strange numbers as neither is divisible by 3? 4 populated could just be so that you have a spare in case of connection problems?

Rich

yeah i would have thought  3 rows of 3.  but until we really look at them we won't know for sure.
hero member
Activity: 588
Merit: 500
June 02, 2016, 03:02:20 PM
#9
Link to this alleged S9 controller board?

(Edit)

 Nevermind, found it - only 4 connectors populated, but I do see the holes for the other 16. I DO wonder why they populated 4 connectors though if they only need 3 for the hashboards?

I would guess some in house ones may have all 20 populated.



Maybe leaving four is easy compared to leaving 3 due to assembly techniques.



Yes 4 and 20 both strange numbers as neither is divisible by 3? 4 populated could just be so that you have a spare in case of connection problems?

Rich
legendary
Activity: 4256
Merit: 8551
'The right to privacy matters'
June 02, 2016, 08:26:02 AM
#8
Link to this alleged S9 controller board?

(Edit)

 Nevermind, found it - only 4 connectors populated, but I do see the holes for the other 16. I DO wonder why they populated 4 connectors though if they only need 3 for the hashboards?

I would guess some in house ones may have all 20 populated.



Maybe leaving four is easy compared to leaving 3 due to assembly techniques.

legendary
Activity: 1498
Merit: 1030
June 02, 2016, 01:38:43 AM
#7
Link to this alleged S9 controller board?

(Edit)

 Nevermind, found it - only 4 connectors populated, but I do see the holes for the other 16. I DO wonder why they populated 4 connectors though if they only need 3 for the hashboards?
hero member
Activity: 572
Merit: 506
June 01, 2016, 12:19:57 AM
#6
The new s9 controller is insane: up to 20 hashboard?? With this Bitmain could build a +90Th/s miner!! It is possible Bitmain will make their version of the neversaw SP-50?
legendary
Activity: 1498
Merit: 1030
May 30, 2016, 03:19:45 AM
#5
Given the issues Intel has had getting 10nm to work at all, I'd bet AGAINST 10nm going mainstream in 2018, and wouldn't bet on it managing mass production in 2019.

Also, it's VERY doubtfull that 10nm would manage a 3:1 improvement in efficiency - and iffy if it could manage even a 2:1 ratio.
Quantum effects have gotten to be a serious issue and limitation as feature sizes keep getting smaller.


 My guess is "no S11 before 2020", but I wouldn't be shocked to be a year off.
legendary
Activity: 1582
Merit: 1006
beware of your keys.
May 29, 2016, 05:36:11 AM
#4
I saw the s-10 spec thread  and I know we will never have an s-10.

So I am asking that about the correctly named thread.


Also I will get to have the first s-11 thread .  


My guess is  it will not come out for a long time and that the s-9  is going to last for more then 20 months.

My guess is no s-10.  Since bitmain did not do an s-6
Depends how hard the block halving will hit the remaining manufacturers. If it boosts their profits from selling coin, then we could see more R&D leading to faster improvements in chips. If not we could potentially see bankruptcy... Who knows maybe the s-11 will be burned in over the winter?

I agree about the S10, i did not reply there, since all i would had to said about it is party pooping.

The S11... i really don't know, maybe a improved version of the 14nm chip? And ran at much lower voltage? I don't know but other than that i think it will take a while.

For a improved 14nm at lower voltage we might see 100% gain in efficiency, for the next NM step... probably triple the H/J gain, at the same voltage, in efficiency as usual.

14nm would probably be outdated, because the fabrication of 10nm is going mainstream in early 2018, IIRC.
even S10 is not possible to be real, it would be at least 1 year after S9 is released, the biggest reason, it is RAD problem.

S11 must be at least 20THs in order to qualify the difficulty in the future.
legendary
Activity: 1302
Merit: 1068
May 29, 2016, 04:11:19 AM
#3
I saw the s-10 spec thread  and I know we will never have an s-10.

So I am asking that about the correctly named thread.


Also I will get to have the first s-11 thread .  


My guess is  it will not come out for a long time and that the s-9  is going to last for more then 20 months.

My guess is no s-10.  Since bitmain did not do an s-6
Depends how hard the block halving will hit the remaining manufacturers. If it boosts their profits from selling coin, then we could see more R&D leading to faster improvements in chips. If not we could potentially see bankruptcy... Who knows maybe the s-11 will be burned in over the winter?

I agree about the S10, i did not reply there, since all i would had to said about it is party pooping.

The S11... i really don't know, maybe a improved version of the 14nm chip? And ran at much lower voltage? I don't know but other than that i think it will take a while.

For a improved 14nm at lower voltage we might see 100% gain in efficiency, for the next NM step... probably triple the H/J gain, at the same voltage, in efficiency as usual.
legendary
Activity: 2212
Merit: 1118
Lie down. Have a cookie
May 29, 2016, 01:57:07 AM
#2
I saw the s-10 spec thread  and I know we will never have an s-10.

So I am asking that about the correctly named thread.


Also I will get to have the first s-11 thread .  


My guess is  it will not come out for a long time and that the s-9  is going to last for more then 20 months.

My guess is no s-10.  Since bitmain did not do an s-6
Depends how hard the block halving will hit the remaining manufacturers. If it boosts their profits from selling coin, then we could see more R&D leading to faster improvements in chips. If not we could potentially see bankruptcy... Who knows maybe the s-11 will be burned in over the winter?
legendary
Activity: 4256
Merit: 8551
'The right to privacy matters'
May 28, 2016, 09:38:01 PM
#1
I saw the s-10 spec thread  and I know we will never have an s-10.

So I am asking that about the correctly named thread.


Also I will get to have the first s-11 thread .  


My guess is  it will not come out for a long time and that the s-9  is going to last for more then 20 months.

My guess is no s-10.  Since bitmain did not do an s-6
Jump to: