2.5nm transistors produced: Thoughts on them & miners?

philipma1957

legendary

Activity: 4102

Merit: 7765

'The right to privacy matters'

Quote from: Artemis3 on January 28, 2019, 10:30:50 PM

And at 16nm the MicroBT M10 is far more efficient than the S9 also at 16nm, so what gives? This was the same guy that used to work with Bitmain so why such difference? If MicroBT can switch to a smaller node, it should take over the lead in my opinion. Unless Braiins OS can break thru the efficiency of the S15 as well as it did to the S9...

Also if the 7nm asic Bitmain uses is unreliable, what is being used in the S15 that some people here have bought and are using at this very moment?

it could be etching layout is much better on the m10 chip then the s9 chip.

it could also be that asicboost which is on the s9 and not on the m10 takes up room to etch it in.

thus the m10 can do more in the same space.

All trade secrets to some extent

Artemis3

legendary

Activity: 1988

Merit: 1561

CLEAN non GPL infringing code made in Rust lang

Quote from: yianni on January 27, 2019, 08:01:49 PM

Actually both E11+ from Ebang and T3 from Innosilicon at 10nm both outperform the Bitmain S15 in efficiency, 48W/TH and 45W/TH. Bitmain is around 50-58W/TH I believe?
And reading some reviews it is a very unstable and unreliable machine.
So far only Bitmain and GMO have attempted 7nm tech. GMO was a total failure, as efficiency was close to 16nm. Only Bitmain came close but with a very unreliable chip.
For now I'm still running my S9's. It's not worth the investment risk yet to upgrade with what is available in the market.
Hopefully they will improve but in these market conditions I seriously doubt it.

And at 16nm the MicroBT M10 is far more efficient than the S9 also at 16nm, so what gives? This was the same guy that used to work with Bitmain so why such difference? If MicroBT can switch to a smaller node, it should take over the lead in my opinion. Unless Braiins OS can break thru the efficiency of the S15 as well as it did to the S9...

Also if the 7nm asic Bitmain uses is unreliable, what is being used in the S15 that some people here have bought and are using at this very moment?

philipma1957

legendary

Activity: 4102

Merit: 7765

'The right to privacy matters'

Quote from: NotFuzzyWarm on January 28, 2019, 10:10:05 AM

Hmm, I take it someone forgot about Canaan and their A921 which came out last October?
Very stable BUT - unfortunately its efficiency is just barely better than the A841.

actually i can downclock an a841 and do 900 watts and 10th which is 90 watts a th.
just a tiny bit better the the a921 as it still has firmware issues.

I have yet to see that bitmain has a bad chip I suspect the psu has been its issue.

and the 10nm innosilicon that says it does 48watts a th is yet to have full shipping and testing.
the ebit e11++ claims 1980 watts at 44th and I can't get one to test or demo

but 1980/44000 = 45watts a th for a 10nm chip

if someone shows it I will believe it.

So for now while I really am not a bitmain fan the s15 at low speed is 900/18000 = 50 watts a th as demo's by a few people .

mine will come by friday I pruchased a used one to confirm it does 50 watts a th .

if I can get an e11++ to test and demo i will do so
if I can get an innosilicon terminator 3 to test and demo.

NotFuzzyWarm

legendary

Activity: 3612

Merit: 2506

Evil beware: We have waffles!

Quote from: yianni on January 27, 2019, 08:01:49 PM

...
So far only Bitmain and GMO have attempted 7nm tech. GMO was a total failure, as efficiency was close to 16nm. Only Bitmain came close but with a very unreliable chip.
For now I'm still running my S9's. It's not worth the investment risk yet to upgrade with what is available in the market.
Hopefully they will improve but in these market conditions I seriously doubt it.

Hmm, I take it someone forgot about Canaan and their A921 which came out last October?
Very stable BUT - unfortunately its efficiency is just barely better than the A841.

philipma1957

legendary

Activity: 4102

Merit: 7765

'The right to privacy matters'

The ebang e11++ claims to do 44th at 1980 watts. So yeah if it is doing this and is shown to be doing this by real miners with real proof of product videos it would be the most efficient miner.

As of today I have yet to see proof of the claim. It costs 2500 but it is not about price it is about efficiency.

There have been real posts or videos on the s15 showing it does 18th 900 watts so it is a true 50 watts a th chip.
The chip is good but some psus have been bad.

As for the inno t2. I have a T2Turbo24th and it is 70 watts a th.
I have not seen real evidence that the new model does 48 watts a th
I have yet to see posts blogs or videos that show the new model doing under

yianni

member

Activity: 141

Merit: 18

BScEE, BScCS, MScCS

Actually both E11+ from Ebang and T3 from Innosilicon at 10nm both outperform the Bitmain S15 in efficiency, 48W/TH and 45W/TH. Bitmain is around 50-58W/TH I believe?
And reading some reviews it is a very unstable and unreliable machine.
So far only Bitmain and GMO have attempted 7nm tech. GMO was a total failure, as efficiency was close to 16nm. Only Bitmain came close but with a very unreliable chip.
For now I'm still running my S9's. It's not worth the investment risk yet to upgrade with what is available in the market.
Hopefully they will improve but in these market conditions I seriously doubt it.

mikeywith

legendary

Activity: 2170

Merit: 6279

be constructive or S.T.F.U

one thing we know for sure is that market conditions are essential to any major improvements in the industry, as you mentioned sales of current 7nm miners are not that promising, at this point 16nm miners (s9) are getting more of the over sales than the 7nm based miners. simply due to cost per TH.

also from a technical stand point , those 2.5 nm have a very long way to get to mass production level, i think from now till end of 2020 we will only see minor to medium improvements on the 7nm.

philipma1957

legendary

Activity: 4102

Merit: 7765

'The right to privacy matters'

Quote from: yianni on January 26, 2019, 10:25:00 AM

As we saw with 7nm chips making them smaller doesn't necessarily mean better efficiency. Right now most 10nm chips are way more efficient than any 7nm out there. Also with smaller chips you run into issues like overheating, too many propagation errors which end up making them less efficient and slower.
I think with 7nm we are already pushing the limits of silicon chips.

The chips were not using silicon I think they use gallium or I read the article wrong.

and the most efficient gear right now is a 7nm the s15.

But I think there is a lot of improvement on the 7nm chip before we move on to a 3 or 2.5 nm chip.

If the m10 whatsminer does 60 watts a th on the 16nm chip on low speed high efficiency

and the bitmain s9 does 80 watts a th on the 16nm chip on low speed higher efficiency

which is true. the 16nm is maxed or close to maxed for efficiency .

the bitmain s15 is doing 50 watts a th on low speed on a 7nm chip

a 25% improvement on that is 37.5 watts a chip.

since the m3 was 150 watts a th vs s7 250 watts a th 28nm chip

the m10 was 60 watts a th vs s9 80 watts a th 16nm chip

the m? should so 37 watts a th vs 50 watts a th for the s15. 7nm chip

so if we do get a good 3nm to work I would guess 5 years from today and 25 watts a th

yianni

member

Activity: 141

Merit: 18

BScEE, BScCS, MScCS

As we saw with 7nm chips making them smaller doesn't necessarily mean better efficiency. Right now most 10nm chips are way more efficient than any 7nm out there. Also with smaller chips you run into issues like overheating, too many propagation errors which end up making them less efficient and slower.
I think with 7nm we are already pushing the limits of silicon chips.

clgrissom3

legendary

Activity: 1722

Merit: 1032

Carl, aka Sonny :)

Better efficiency is the only future path miners can take and atomic level manipulation appears to be the way to proceed.

Interesting article.

NotFuzzyWarm

legendary

Activity: 3612

Merit: 2506

Evil beware: We have waffles!

Well researchers have produced transistors as small as 2.5nm using techniques that look compatible with current bleeding-edge 7nm processes. Smallest 3d transistors yet

Now given the current downturn in the market I do not see these showing up in miners anytime soon, hell, 7nm miners are just out and sales are - meh, but still... Thoughts?

Topic: 2.5nm transistors produced: Thoughts on them & miners? (Read 399 times)