rumors of 3.5Th antminer S7 "lite"

[VirosaGITS]

Look like RichBC's guess did not come to be.

Tough game this guessing.  Then again who would have guessed that we would have 4 different Default Frequencies / TH/s and 2 different chip count S7 in such a short period of time?

Still they are running out of permutations, unless we start seeing different Default Frequency 135 Chip versions. 


Rich

I would really like to see a two board S7 with a high chip permutation, with bigger heatsink, should make the thing very quiet during winter and make it cheaper to buy it single unit by single unit.

Back to wishing for a 3.5TH~ "lite" for Christmas.

Yes I guess we have several games... Guessing what they might do, guessing what it is while we wait for someone to get one, and Wishing for what we actually want. 

On the Wishing front yes I still think that a 2 Board S7 would be a great product. I am actually now torn between wishing it would be based on 54 or 45 chips / board? 45 chip I feel would be a very solid product with a lot of scope for over & under clocking and volting. Whereas 54 chips / board would push up the hash rate and I think respond well to overvolting which is much easier to do with a Server PSU than undervolting. Unfortunately I don't expect to see either...

Am still intrigued to find out why they have found it necessary to have so many variants? I can only assume that they are getting a very broad spread of BM1395 chips and this is there way of coping with this? As has been said already it's going to be a nightmare in the 2nd hand market to know what you are buying.

Rich

Maybe they are dealing with different fabricant that can do one but not the other? Its nagging me, as to why they are doing this, as i can't come up with a reason. Sure they have lots of BM1385's but that doesn't really justify doing random things with them?

[1713580909]

1713580909

[1713580909]

1713580909

[1713580909]

1713580909

[1713580909]

1713580909

[1713580909]

1713580909

[notlist3d]

Look like RichBC's guess did not come to be.

Tough game this guessing.  Then again who would have guessed that we would have 4 different Default Frequencies / TH/s and 2 different chip count S7 in such a short period of time?

Still they are running out of permutations, unless we start seeing different Default Frequency 135 Chip versions. 


Rich

I would really like to see a two board S7 with a high chip permutation, with bigger heatsink, should make the thing very quiet during winter and make it cheaper to buy it single unit by single unit.

Back to wishing for a 3.5TH~ "lite" for Christmas.

Yes I guess we have several games... Guessing what they might do, guessing what it is while we wait for someone to get one, and Wishing for what we actually want. 

On the Wishing front yes I still think that a 2 Board S7 would be a great product. I am actually now torn between wishing it would be based on 54 or 45 chips / board? 45 chip I feel would be a very solid product with a lot of scope for over & under clocking and volting. Whereas 54 chips / board would push up the hash rate and I think respond well to overvolting which is much easier to do with a Server PSU than undervolting. Unfortunately I don't expect to see either...

Am still intrigued to find out why they have found it necessary to have so many variants? I can only assume that they are getting a very broad spread of BM1395 chips and this is there way of coping with this? As has been said already it's going to be a nightmare in the 2nd hand market to know what you are buying.

Rich

Maybe they are dealing with different fabricant that can do one but not the other? Its nagging me, as to why they are doing this, as i can't come up with a reason. Sure they have lots of BM1385's but that doesn't really justify doing random things with them?

I think it makes more sense the 5th vs the almost 4th on the 3 blade design.  I just think 10 PCIe for 4TH is to much, and I know some are going to say you can use less... and i'm sure some do but I have been following bitmain's warranty on last few miners.

I cannot for the life of me figure out the up vs down.   But I think they will end up closer to  5th on future batches maybe they got quality up or something, or they had better yields.  There has to be something behind it, and I'm guessing it's tied to money.

[Tmdz]

Maybe they buffed out some of the issues with the earlier batches and did a 4 th unit to just be on the safe side of things.  After some testing they said ok everything is good lets drop a 5 th unit and start pushing out more of these chips.  It is interesting to note that shipping times for batch 6 & 7 are exactly the same, so that would lead me to believe they had both these machines designed at the same time.

[carlosmnk]

i think they are rising the default frequency because is winter and because they are acceleratin the roi for this products; but maybe they changed the headsinks design for this batch...

[RichBC]

I have this feeling that we will never get to the bottom of the S7 variants. I have seen mention of at least 3 board revisions, but because it's a closed unit with the Boards covered in mini heatsinks we may not get any good pictures to see if the changes are cosmetic or just a BM1385 selection?

Rich

[RichBC]

Look like RichBC's guess did not come to be.

Tough game this guessing.   Then again who would have guessed that we would have 4 different Default Frequencies / TH/s and 2 different chip count S7 in such a short period of time?

Still they are running out of permutations, unless we start seeing different Default Frequency 135 Chip versions.  


Rich

Wow got a guess right.   Batch 8 is a 135 Chip Variant.

S7 Specifications:

1. Hash Rate: 4.73 TH/s ±5%

2. Power Consumption: 1293W + 10% (at the wall, with APW3, 93% efficiency, 25C ambient temp)

3. Power Efficiency: 0.25 J/GH + 10% (at the wall, with APW3, 93% efficiency, 25°C ambient temp)

4. Rated Voltage: 11.60 ~13.00V

5. Chip quantity per unit: 135x BM1385

6. Dimensions: 301mm(L)*123mm(W)*155mm(H)

7. Cooling: 2x 12038 fan

8. Operating Temperature: 0 °C to 40 °C

9. Network Connection: Ethernet

10. Default Frequency: 700M

Rich

[carlosmnk]

700M !!!!!!!!!!!!

[alh]

It would sure be interesting to know if the 135-chip miner is actually a de-populated 162-chip miner with "dummy jumpers" in place of the missing chips. I hope that somebody that gets a Batch #6 or Batch #8 opens it up enough to count.

[notlist3d]

It would sure be interesting to know if the 135-chip miner is actually a de-populated 162-chip miner with "dummy jumpers" in place of the missing chips. I hope that somebody that gets a Batch #6 or Batch #8 opens it up enough to count.

The problem with this is now if you did that it will void your bitmain warranty... which kinda sucks.  You basically have to have permission from them to do anything at this point.

I agree it would be interesting to see some of the chip counts.  They keep jumping up and down it's hard to keep track.  I think the S7 will be remembered as the miner that never really kept one speed.  It is very odd for them to keep changing it each batch it seems at this point.  I do admit I like they went up and not down on last batch put up on hashing speed.

[RichBC]

So following the publication of the picture of the Batch 8 Hash Board on the Bitnain Product page this is my current thinking.



As we knew they have removed 3 triples, 9 ASIC's, from the chain. However rather than that being it, which never completely added up because the resultant increase in the Core Voltage from 0.6V to 0.8V took the chips way off the data Sheet and would have ruined the efficiency. So they have removed the chips from the bottom of the chain and my best guess is that the have added a Buck converter to the board, in the vacated space.

We can see the large inductor & capacitors and I am assuming that the Buck Converter chip is on the other side, which they are choosing to keep from us at the moment.   So assuming that is correct I am further assuming that the 45 chips are still arranged as a 15 Chip triple node string. The buck converter presumably reduces the 12V to around 10V, giving us a hybrid design. In fact an integrated version of what we have been playing with on the S5 with a 12V PSU and an external Buck Converter.

The big unknown here is have they made this voltage software adjustable to give say 0.55V to 0.75V / Chip, or is it a fixed voltage? If they have done this perhaps it was done as a reaction to the Avalon6 which has an automatic frequency / undervolt algorithm?

Either way although there will be a small loss in efficiency with the Buck Converter but a gain in flexibility and overall be it software adjustable or good old pencil mod I think it represents an improvement over the straight string design.

It is also possible that I have completely misunderstood what the components are and the change that they have made is for something completely different.......


Rich

[carlosmnk]

So following the publication of the picture of the Batch 8 Hash Board on the Bitnain Product page this is my current thinking.



As we knew they have removed 3 triples, 9 ASIC's, from the chain. However rather than that being it, which never completely added up because the resultant increase in the Core Voltage from 0.6V to 0.8V took the chips way off the data Sheet and would have ruined the efficiency. So they have removed the chips from the bottom of the chain and my best guess is that the have added a Buck converter to the board, in the vacated space.

We can see the large inductor & capacitors and I am assuming that the Buck Converter chip is on the other side, which they are choosing to keep from us at the moment.   So assuming that is correct I am further assuming that the 45 chips are still arranged as a 15 Chip triple node string. The buck converter presumably reduces the 12V to around 10V, giving us a hybrid design. In fact an integrated version of what we have been playing with on the S5 with a 12V PSU and an external Buck Converter.

The big unknown here is have they made this voltage software adjustable to give say 0.55V to 0.75V / Chip, or is it a fixed voltage? If they have done this perhaps it was done as a reaction to the Avalon6 which has an automatic frequency / undervolt algorithm?

Either way although there will be a small loss in efficiency with the Buck Converter but a gain in flexibility and overall be it software adjustable or good old pencil mod I think it represents an improvement over the straight string design.

It is also possible that I have completely misunderstood what the components are and the change that they have made is for something completely different.......


Rich

Very interesting. Before i saw this picture, i did not want to but this batch, but now... I don't think so, perphaps it could be a good idea...

Thanks Rich!!