Bitmain Antminer S6 and S7 Speculation

[philipma1957]

I have a real love/hate feeling about Bitmain's keep quiet till it's ready to ship way of doing things.  I have to respect the fact that they don't do pre orders, at least no ones that are more then a couple weeks from shipping and they are always at least very close to shipping on time. 

On the other hand I want to know what those guys are up to...   The suspense is killer

well sfards has floated 1129 so they intentionally priced high and will self mine for a while. much like avalon did.

once again    bitmaintech still has very little incentive to sell us s-6's or s-7's.

They will slowly replace all their s-5's with s-6's and s-7's and keep selling s-5's at a price just a bit too high too buy.

Why it is good business for them to do so.

Very slow diff growth works for them if they are upgrading s-5's to s-6 or s-7 gear.

Just follow what they have done since the s-3 came out.

[mavericklm]

it would be interesting if miners producers have/will have a deal to sell at higher price, not like spondoolies vs bitmain, sp20 vs s5
they will upgrade, we will upgrade slower...
diff will go up slowly and they will have higher % of the network, and ~ constant % between themselves....

[philipma1957]

it would be interesting if miners producers have/will have a deal to sell at higher price, not like spondoolies vs bitmain, sp20 vs s5
they will upgrade, we will upgrade slower...
diff will go up slowly and they will have higher % of the network, and ~ constant % between themselves....

I think this is what they will do for the time being. It also is safer for them with the ½  ing   coming up.

[philipma1957]

There has been more buzz on the s-7  so maybe it comes out in late July.

[sergio]

Any ETA on S6? At least speculative?

I'm still thinking mid to late April.

Well, with the current BTC price, I m really not sure how profitable for them is to develop new miners. The profit margin s really, really low.

Not really. In fact when the S5 was first released it was sold for about $300 USD, but shortly after Bitmain got news that spoondolies was out of the hardware bussiness they jacked up the price to $340, well as of now the S5 costs $360, growing difficultly and a growin price.

I am sure at $300 they were making a  profit, so to say the profit magin is really low it is not accurate, doing basic math you can easily conclude that the profit margin is over 20%.

profit margin is really really low for indepent miners, not for the hardware vendors.

.

[philipma1957]

Any ETA on S6? At least speculative?

I'm still thinking mid to late April.

Well, with the current BTC price, I m really not sure how profitable for them is to develop new miners. The profit margin s really, really low.

Not really. In fact when the S5 was first released it was sold for about $300 USD, but shortly after Bitmain got news that spoondolies was out of the hardware bussiness they jacked up the price to $340, well as of now the S5 costs $360, growing difficultly and a growin price.

I am sure at $300 they were making a  profit, so to say the profit magin is really low it is not accurate, doing basic math you can easily conclude that the profit margin is over 20%.

profit margin is really really low for indepent miners, not for the hardware vendors.

.

 I agree with this part in bold.

I would argue that hardware builders have tighter margins then you think.
I would also argue that bitmaintech keeping price at 360 or more is not the issue.
I actually think they showed longer term thinking bty doing that.
If they sold gear dead cheap at 250 or so I think it would have lowered price of coins.
When they did the price war with sp-tech coins dropped under 200 usd around 180 in January 2015.

Our problem is exchange failures and lack of demand for btc by the general public.

[alh]

While it would be nice to actually hear something concrete on the S7 from Bitmain, I wonder what will actually motivate them to start selling it to the public? Do they need the cash?  Could they do just as well by replacing some set of their in-house S5 gear and continue to sell S5's? I think every one has accepted "new" means it has a warranty, not that it actually means "unused".

Is there actually any competitive threat to them at this time?

[jonnybravo0311]

Is there actually any competitive threat to them at this time?

Isn't BitFury deploying / about to deploy their new chips?  From a consumer sales standpoint, no, Bitmain doesn't have any serious competitive threats.

[Amph]

While it would be nice to actually hear something concrete on the S7 from Bitmain, I wonder what will actually motivate them to start selling it to the public? Do they need the cash?  Could they do just as well by replacing some set of their in-house S5 gear and continue to sell S5's? I think every one has accepted "new" means it has a warranty, not that it actually means "unused".

Is there actually any competitive threat to them at this time?

if the old trick of preorder to have money upfront isn't working anymore, and their revenue would be greater in selling those new miners, they will go for that way

remember that efficiency is not that great on those new miners, they are still working with 28nm(at least the last rumors said so...), so i doubt there is much to gain if all the s5 were to be replaced with s6-s7

[philipma1957]

selling the s-7 rather then mining with it is not really a good idea.

the new pattern for asic builders is replace your older gear slowly with the newer gear.  all the while slowly selling the older gear at the highest price the market will allow.

Both avalon and bitmaintech slowly sell the avalon 4.1 at high price and the s-5 at high price while mining it,

they then replace the 4.1 or the s-5 with more efficient gear.  they have no need to rush.

the new avalon mini does .4 watts vs .53 watts.  the new s-7 does > watts vs .52 watts.


the trick is to allow over-clocked gear die in the field   see my sp20 stat thread.

what does all of this do.  it lowers the builders risk it ends the gear-up wars  and allows for safer profit for the building-self mining company.

SO when the s-7 comes out look for a .35 to .4 watt item priced to break even in about 180-210 days if your power cost is 8 cents.

[jonnybravo0311]

selling the s-7 rather then mining with it is not really a good idea.

the new pattern for asic builders is replace your older gear slowly with the newer gear.  all the while slowly selling the older gear at the highest price the market will allow.

Both avalon and bitmaintech slowly sell the avalon 4.1 at high price and the s-5 at high price while mining it,

they then replace the 4.1 or the s-5 with more efficient gear.  they have no need to rush.

the new avalon mini does .4 watts vs .53 watts.  the new s-7 does > watts vs .52 watts.


the trick is to allow over-clocked gear die in the field   see my sp20 stat thread.

what does all of this do.  it lowers the builders risk it ends the gear-up wars  and allows for safer profit for the building-self mining company.

SO when the s-7 comes out look for a .35 to .4 watt item priced to break even in about 180-210 days if your power cost is 8 cents.

You and I are on pretty much exactly the same page regarding the S7's performance, Phil.  Earlier in the thread I stated that the S7 should likely get around 1.5TH/s for the same 590W the S5 demands.  Your numbers put the range between 1.475TH/s and 1.685TH/s.  Hopefully they will make the thing a bit quieter.  My S3s are virtually silent compared to the noise the SP20 and S5 make.

[-droid-]

selling the s-7 rather then mining with it is not really a good idea.

the new pattern for asic builders is replace your older gear slowly with the newer gear.  all the while slowly selling the older gear at the highest price the market will allow.

Both avalon and bitmaintech slowly sell the avalon 4.1 at high price and the s-5 at high price while mining it,

they then replace the 4.1 or the s-5 with more efficient gear.  they have no need to rush.

the new avalon mini does .4 watts vs .53 watts.  the new s-7 does > watts vs .52 watts.


the trick is to allow over-clocked gear die in the field   see my sp20 stat thread.

what does all of this do.  it lowers the builders risk it ends the gear-up wars  and allows for safer profit for the building-self mining company.

SO when the s-7 comes out look for a .35 to .4 watt item priced to break even in about 180-210 days if your power cost is 8 cents.

You and I are on pretty much exactly the same page regarding the S7's performance, Phil.  Earlier in the thread I stated that the S7 should likely get around 1.5TH/s for the same 590W the S5 demands.  Your numbers put the range between 1.475TH/s and 1.685TH/s.  Hopefully they will make the thing a bit quieter.  My S3s are virtually silent compared to the noise the SP20 and S5 make.

Yeah the noise was one of my wishes for the upcoming batch, hopefully they go the Avalon 4.1 route

[sidehack]

If they could justify the chip cost, Bitmain could have put out a 500W 1.4TH miner on the S[odd] chassis six months ago; it just would have had 51 chips per blade instead of 30. That's a practical maximum for chip density on a board that size, and gets there with the same 3x17 string they're using in the S4+ (which, by the by, if it was made with BM1384 the same PSU could source six modules instead of four and we'd have a 4.2TH miner drawing 1600W wall, which would have been one heck of a nice S6). Given the S5 could do 1.2TH off 600W for about half the investment and was still the best thing anyone could buy, there was no need. To get a significant and also cost-effective improvement would require a new chip, so I don't begrudge them the design at all. Even a 2x16 string, a practical stock clock would have been 100GH less (about 1050GH, albeit at under 500W wall) but that kind of efficiency wasn't really necessary. The only real competition at the time was the SP20, and they already had it beat on price by a long way.


ASICMiner's BE300 proved there's still a lot of room left in 28nm, as its specs were about 25% better than BM1384 across the board (0.19 vs 0.25 bottom clock, 0.35 vs 0.45 top clock). I'd expect to see at least a 30% efficiency gain from Bitmain's S7 chip. The BM1380 top clock was something like 6W per chip, BM1384 about 10. I'm going to assume the new chip will have a similar top clock power dissipation of 10W, with a 30% better efficiency than the BM1384. I think a slightly higher chip density than 60 is likely, as the core voltage will probably be reduced - say, 750mV stock vs 800mV stock gives a 16-chip string instead of 15 for 64 chips per machine instead of 60. If that's the case, and we're running at about 85% practical maximum clock (fairly typical neighborhood given S1/S3/S5 stock setpoints), we'll see probably 8W per chip for 530W DC consumption (including fan and controller, still under 600W at the wall) and approximately 1800GH.

As much as it would be super nice, I kinda doubt Bitmain will put up the extra $10 for quieter fans. They tested it out a bit with the S3, full shroud and dual fans for super durability and silent running, but the S5 they really didn't care - single super-loud fan, cheap plastic side panels, designed more for bulk shelving customers than regular folks. As long as 90% of their sales go to people that aren't concerned with how loud the fans are (admittedly, I'm among them though I'm by no means a bulk buyer - I just don't care about fan noise until it approaches SP10 annoyance/pain levels), Bitmain's not going to do jack to make miners quieter if it means adding expense.

That's my guesses anyways.

[jonnybravo0311]

Great points sidehack.  The BE300 did indeed beat the BM1384 and show that there's still plenty of wiggle room with the 28nm tech to get some nice efficiency gains.  Do I think Bitmain's going to throw all of their cards on the table with this next S6/S7 generation and offer up the best they can get from their chips?  Honestly, no.  Even though they could very easily put out an S7 running 1.8TH/s at under 600W, I just don't see the incentive for them to do so.  This will likely be a "hey it's good enough" generation while they transition to a smaller process node to compete with KnC's 16nm tech (which promises some absurdly low power draw).

[-droid-]

Great points sidehack.  The BE300 did indeed beat the BM1384 and show that there's still plenty of wiggle room with the 28nm tech to get some nice efficiency gains.  Do I think Bitmain's going to throw all of their cards on the table with this next S6/S7 generation and offer up the best they can get from their chips?  Honestly, no.  Even though they could very easily put out an S7 running 1.8TH/s at under 600W, I just don't see the incentive for them to do so.  This will likely be a "hey it's good enough" generation while they transition to a smaller process node to compete with KnC's 16nm tech (which promises some absurdly low power draw).

When does KnC plan to introduce this 16nm chip?

[TheRealSteve]

When does KnC plan to introduce this 16nm chip?

Your guess is as good as any.  They announced various stages of it in the past.
2014-11-18 - 16nm plan announcement - http://www.kncminer.com/blog/newsarchive#announcing-solar-a-step-change-in-processing-capability
2015-02-03 - tapeout complete - http://www.kncminer.com/blog/newsarchive#accel-investing-in-kncminer-16nm-tape-out-completed
2015-06-03 - deployed - http://www.kncminer.com/blog/newsarchive#changing-the-game-again

2015-07-05 - must be so efficient they can stand to bleed relative hash rate - https://i.imgur.com/HMvyTPa.png

[sidehack]

Just as soon as they figure out another surefire way to steal money from everyone and still convince important people that they're a respected business. I'm still confused how investors trusted them enough to pay for 16nm dev after they spent the whole last year ripping people off with nonexistent and/or poorly-made products.

[philipma1957]

My KNC confusion is pretty simple if they built this chip  and it 2 or 3x better then any chip ever built  why don't they really show proof of it. 

They could get a shit ton of investment money and then rob them too.

So far I have not seen clear evidence of a .1 watt mining machine mining as of today.

[TheRealSteve]

So far I have not seen clear evidence of a .1 watt mining machine mining as of today.

Remember when Spondoolies-Tech was planning one?

[mavericklm]

they don't have it!!!!
there is no .1w machine!