Antminer S6 speculation

[RichBC]

Looks good on paper. 20nm ASIC's, 6THS, $3299, 0.26J/W and can be underclocked for greater efficiency.

Whether it exists or not I do not know? and this is the S6 speculation thread.....


Rich

[QuintLeo]

Given that their "Ares Scrypt miner" is obviously an Alcheminer at a ridiculously high price, I have some doubts about this Uranus thing - though to be fair, they do specify somewhere on their page

In production we use our US partner developed ASIC chips and motherboards produced by us

 The BIG sticking point is that "20 nm" part of their spec - I've not seen ANYONE from an ASIC company claiming to be working with a 20nm process.

 They also say the CHIP efficiency is .24w/GH, the MACHINE efficiency is .26w/GH, yet the machine is "designed to be underclocked" - voltage converters for typical low-voltage applications like SHA256 ASIC do NOT achieve a well over 90% efficiency, and there seems to be ZERO allowance in those specs for the power consumed by the controler board and other parts needed for the ASIC to talk to the outside world.


 Definitely a "I want to at least see a review from a few REPUTABLE sources" item before I will believe it's real.

[cogabonito]

S6 has been mentioned (I believe by Dogie, but might have been by BitMain directly) that it's not going to happen, will be "skipped".

 It's the S7 we're all waiting on now....

Is it true or are you trolling us? Can you show me source of this? I can't believe they skip a generation for whatever reason.

[philipma1957]

S6 has been mentioned (I believe by Dogie, but might have been by BitMain directly) that it's not going to happen, will be "skipped".

 It's the S7 we're all waiting on now....

Is it true or are you trolling us? Can you show me source of this? I can't believe they skip a generation for whatever reason.

what is true is dogie wrote:

" the s-4 +  was kind of the s-6  and the s-7 comes next." 

I quote this from memory so it is not necessarily a perfect quote.

[VirosaGITS]

S6 has been mentioned (I believe by Dogie, but might have been by BitMain directly) that it's not going to happen, will be "skipped".

 It's the S7 we're all waiting on now....

Is it true or are you trolling us? Can you show me source of this? I can't believe they skip a generation for whatever reason.

They're not skipping a generation, the even number generation is a Rack mount unit, the uneven number is a small form factor. The next unit coming will be s7 as reported by several users of reply from Bitmain.

[QuintLeo]

They're skipping a RACK MOUNT generation, to be picky.

 9-)

[VirosaGITS]

They're skipping a RACK MOUNT generation, to be picky.

 9-)

Dunno, they didn't say they would not release a rack mount, they just said they were releasing the smaller factor next. It was supposed to be last month but now they said they would announce at some point within a month or two.

[QuintLeo]

I don't see them making a rack-mount miner from the chips to be used in the S7 and calling it an S6.

[VirosaGITS]

I don't see them making a rack-mount miner from the chips to be used in the S7 and calling it an S6.

You're not seeing a S7 either. They did not announce either yet. Only that the next thing to be released will be the S7.

[QuintLeo]

I don't see them making a rack-mount miner from the chips to be used in the S7 and calling it an S6.

You're not seeing a S7 either. They did not announce either yet. Only that the next thing to be released will be the S7.

 Nope, not yet. "soon". I'm guessing a month or so, however long it takes them to move out most-to-all of their existing S5 units in their farm.

[VirosaGITS]

This turned up in the S5 thread:

S5+

432 chips, 7722GH/s, 3436 Wt








Have anyone heard about it?

Looks interesting.

I would like to see some source. The member who brought it up did not mention any, he just threw some insane specs in and left.

[QuintLeo]

Saw it over there, but just now got to thinking about heat dissipation.

 No bloody WAY those tiny heat sinks are going to keep a BM1384 running at .75 volts (3 chains of 16 per board is the ONLY viable option that fits the board setup, 2 chains of 24 is below the BM1384 minimum voltage spec, 4 chains of 12 is WAY over) cool, expecially with that many clustered that close together.

 The "3 PCI-E connectors per board" part of the picture fits in with that scenario, which comes down to LESS heat sink area total than on the S5, trying to dissipate about 40% more power.



 Not quite a Nuclear Meltdown scenario, or an Intel Pentium original generation that couldn't run 75 Mhz due to heat issues, but.....

[MarkAz]

I doubt those are all the headsinks - with the S5, they actually experimented with those very same heatsinks.  They go on the opposite side of the main one, so I imagine that's the same with this, just a bit of extra heat dissipation on the opposite side of the board.  More than likely inside there you have a big headsink similar to the one on the S5 bank to back to back with each board.  In terms of the S5, the density is basically 2x, since each one of these blades are the same hashing power of one S5...

[VirosaGITS]

I doubt those are all the headsinks - with the S5, they actually experimented with those very same heatsinks.  They go on the opposite side of the main one, so I imagine that's the same with this, just a bit of extra heat dissipation on the opposite side of the board.  More than likely inside there you have a big headsink similar to the one on the S5 bank to back to back with each board.  In terms of the S5, the density is basically 2x, since each one of these blades are the same hashing power of one S5...

Ah its using, allegedly, mini aluminium headsink and a high air pressure is blown through. Dissipating the heat is quite possible since it would obviously use something like 6 of those fans S5 uses. The chips would run hot but thats probably okay.

Anyways on the bright side, the diff won't go up TOO much.

[philipma1957]

I doubt those are all the headsinks - with the S5, they actually experimented with those very same heatsinks.  They go on the opposite side of the main one, so I imagine that's the same with this, just a bit of extra heat dissipation on the opposite side of the board.  More than likely inside there you have a big headsink similar to the one on the S5 bank to back to back with each board.  In terms of the S5, the density is basically 2x, since each one of these blades are the same hashing power of one S5...

Ah its using, allegedly, mini aluminium headsink and a high air pressure is blown through. Dissipating the heat is quite possible since it would obviously use something like 6 of those fans S5 uses. The chips would run hot but thats probably okay.

Anyways on the bright side, the diff won't go up TOO much.

Bingo you got it. If the power improvement is .51 down  to .44 diff will not move much

[RichBC]

So just thinking a bit more about the S5+, as others have said the name under calls things a bit S5+3 or S5+++ might have been more appropriate?

Got me thinking that perhaps the development started as the item that philipma1957 wants? A single bay S5+ with controller. Original brief might have been to push the hash rate to the max in the form factor and slightly improve the J/GH?

Then someone realised that there might be some economies to be made, common frame & hash board, by tacking 3 bays together rather than designing a completely new BM1384 based S6?

Rich

[usenet]

Price for s5 + will be maintained probably. But the price for the s5 has fallen already . Only 352 USD ( 1.323 BTC) .
It seems that the weight of the parcel jumped to 4.2kg . They can now sell a version with small heatsinks or I missed something ?
Hashnest also got a new s5 delivery . Replacing the equipment continues.
Now we have a very interesting situation , because it looks like the S5 will get a good deal on the market.

[QuintLeo]

The S5+ price also dropped, quoted at 17200 yuan now instead of the original 19000 - still almost $2750.
http://shop.bitmain.cn/goods.php?id=47
(though I don't understand the chinese on most of that page the price is obvious).

 Seems Bitmain is finally responding to the recent bitcoin small price dump.


 I also finally just noticed the apparent "intended for rack mount" tabs - dunno why they didn't call this the S6, abet the tabs don't seem to be spaced right for a standard rack mount.

[VirosaGITS]

The S5+ price also dropped, quoted at 17200 yuan now instead of the original 19000 - still almost $2750.
http://shop.bitmain.cn/goods.php?id=47
(though I don't understand the chinese on most of that page the price is obvious).

 Seems Bitmain is finally responding to the recent bitcoin small price dump.


 I also finally just noticed the apparent "intended for rack mount" tabs - dunno why they didn't call this the S6, abet the tabs don't seem to be spaced right for a standard rack mount.

Typically i would stack rackmount units, not put them in a rack. This doesn't particularly look like something i'd be able to stack unto one another, so hmm...

[usenet]

The S5+ price also dropped, quoted at 17200 yuan now instead of the original 19000 - still almost $2750.
http://shop.bitmain.cn/goods.php?id=47
(though I don't understand the chinese on most of that page the price is obvious).
Seems Bitmain is finally responding to the recent bitcoin small price dump.

Who knows? They are close to the target $2500 . Perhaps we will see the price soon.