Wu Announces BM 7NM pending Miner release. 42 J/TH

[BenRickert]

So today in Georgia Jihan made his long awaited announcement. ( See Bitmain Web page ) The new 7NM ASIC from BM is coming. When? How Much? most importantly " How much better. Claiming 42 J/TH for the 7NM. Thoughts?

NEW MINING MACHINE WILL BE AVAILABLE SOON
New Chip Uses the 7nm Semiconductor Manufacturing Process and Will Greatly Improve Power Efficiency

BEIJING, P.R. China, September 21, 2018 – Jihan Wu, CEO and co-founder of Bitmain Technologies Ltd. announced Bitmain’s next-generation ASIC chip BM1391 for mining cryptocurrencies using the SHA256 algorithm at the World Digital Mining Summit in Georgia today. Mr. Wu also revealed that Bitmain will soon launch a new generation of mining machines equipped with the BM1391 chip.

The new ASIC SHA256 algorithm acceleration chip carefully built by Bitmain uses one of the world’s most advanced semiconductor manufacturing technologies, 7nm FinFET. The ASIC chip integrates more than a billion transistors and is optimized for maximum efficiency. Thanks to its unique circuit structure and low power technology, it is stable and more efficient. Mr. Wu said tests have shown that the chip can achieve a ratio of energy consumption to the mining capacity that is as low as 42J/TH.

At his keynote speech at the summit, Wu Jihan said mass production of the 7nm chip will begin and be applied to the next generation of Bitmain’s Antminer machines, improving the new miners in various performance indicators, in terms of power and power consumption, and enabling outstanding performance.

[1714465740]

1714465740

[1714465740]

1714465740

[1714465740]

1714465740

[1714465740]

1714465740

[1714465740]

1714465740

[HagssFIN]

They published a press release here: https://blog.bitmain.com/en/bitmain-announces-next-generation-7nm-asic-chip/

[BenRickert]

Thanks for the Link Hags. I'm a bit sleepy. Sorry.

They published a press release here: https://blog.bitmain.com/en/bitmain-announces-next-generation-7nm-asic-chip/

Hey I'm curious as to your take. I'm throwing a wild guess at possible efficiency gain. My assumptions; 28-32 TH with a 1700-1800 watt power draw. So 3-4 times efficiency gain over the S9 13.5th. Yours??

[mgoz]

My assumptions; 28-32 TH with a 1700-1800 watt power draw.

That would be 56-60 J/TH. At same power draw it would be 40-43 TH/s if it were 42 J/TH. I'd like to see them do sub 1k power draw model. Something like 18 TH/s @ 750W.

[HagssFIN]

It will be more like 2 - 2.5 times power efficiency gain vs. the Ant S9.

[BenRickert]

It will be more like 2 - 2.5 times power efficiency gain vs. the Ant S9.

Well.....depending on price......We'll see how worth it it will be to sit up all night hitting the "refresh button" on the bitmain order page.

[philipma1957]

It will be more like 2 - 2.5 times power efficiency gain vs. the Ant S9.

if it comes and works as well as they say.

I could see a 48 to 52 watt per th unit.

but so far no hard evidence of it doing 42 watts per th


Bitmain  has not made the best chip  for any algo for more then a year.

That is in terms of efficiency.

I have heard a lot of negative buzz on the 7nm chip wafer they produced.

But as always if they come out with a 42 watt per th   miner at the lowest price per th

They will remain the king.

A simple 25 th  at 42 watts a th = 1050 watt foot print.

I wonder if they go the M10 route or the T2Turbo route  and they  come with  built in psu.

[Zand_]

I haven't been keeping up with the latest ASIC releases but I'm guessing (if specs are accurate) this would be the most effecient SHA-256 miner on the market?

[VRobb]

Math is off: 42% of the power is used for same hash, or a bit more than 2x as efficient which is still quite impressive!

[hashestohashes]

I figured they would announce something at that thingy thats been headlining their front page for....months it feels like.

[Zand_]

Math is off: 42% of the power is used for same hash, or a bit more than 2x as efficient which is still quite impressive!

Can you elaborate? Not sure I follow

[sarcheer]

Based on the reliability of the S9, Bitmains failures with their test wafers, and IPO problems I would wait at least 6 months after these are in the hands of people before touching them.

Just feels like this machine is going to have some problems.

Edit: not to mention KYC requirements. No way in hell am I sending my sensitive ID documentation to hack-prone Bitmain.

[HagssFIN]

Can you elaborate? Not sure I follow

You need to learn some more math.

If it would be four times as efficient, it would mean 25 J per Th/s.

[Zand_]

You need to learn some more math.

If it would be four times as efficient, it would mean 25 J per Th/s.

Not disagreeing with you, joules were never covered in math class. Let me see if I got this right...

.1J/GH (or .1W/GH) comes out to 100J/TH. So 42J/TH would make it roughly 138% more efficient?

[VRobb]

Yes, it's 2.38 times as efficient, so it has 138% more hash for the power.  A bit convoluted, but on!  

[Gws24]

The stated efficiency is only for the chip and not the miner (and only in a lab so best case) so I would expect something like 55 J/TH rather than 42 J/TH but that is not really that impressive since Whatsminer is doing that on 16nm.

nice tweetstorm from Samson Mow
https://twitter.com/Excellion/status/1043181897319247872

and yes, I know Samson Mow and Jihan Wu aren't exactly friends but he raises some interesting questions.

[philipma1957]

The stated efficiency is only for the chip and not the miner (and only in a lab so best case) so I would expect something like 55 J/TH rather than 42 J/TH but that is not really that impressive since Whatsminer is doing that on 16nm.

nice tweetstorm from Samson Mow
https://twitter.com/Excellion/status/1043181897319247872

and yes, I know Samson Mow and Jihan Wu aren't exactly friends but he raises some interesting questions.

Yes he does doesn’t he.

I will add that bitmain has not produced a top efficiency miner of any kind in quite a while.

I wish I had a few more m10

I think the newer m10s will come out soon and be the miner to get.

[NotFuzzyWarm]

Not disagreeing with you, joules were never covered in math class. Let me see if I got this right...

.1J/GH (or .1W/GH) comes out to 100J/TH. So 42J/TH would make it roughly 138% more efficient?

As a point of reference, the Joule is a measure of total energy per-second so 1 Joule = a 1 Watt load averaged over one second.
A Watt is a measure of energy at any given instant in time.

A fun example: One of our lasers produces an average of 100W of power firing 100J pulses that are 1ns wide (0.000,000,001sec) at a rate of 1 shot per-second: peak power of those 1ns wide pulses = 100,000,000,000W or 100Giga-watts.

[Foofighter]

We were also there and wrote some follow up story. Jihan basically shared his Vision for the next 10 years and at last a sneak peak of a new chip.
There was also alot about Ai and Quantum computing stuff in his speech.

https://www.cryptouniverse.at/world-digitial-mining-summit-kicks-off-with-an-amazing-speech-and-new-7nm-miners/

[Thetaj]

Well something is definitely going on since Bitmain is now out of 16nm chips (from my sources). From what I hear they are full-steaming for production of this new 7nm wafer. Bad news for anyone with an S9 since there will be no support in 3 months (but it'll be trash by then anyways so...)