The only thing I really know about TSMC is that Apple chose them as an alternative to Samsung a couple of years ago, and then TSMC was unable to get their fab operating at a level that would allow Apple to source major product CPUs from them. Last I heard Apple is still working with them, but not getting any volume production.
This tells me that if a small third party operator tried to go to them for fab work, the n00b would be way over their head. Maybe their blueprint would get fabbed at an acceptable level and maybe not. It wouldn’t be easy or quick. The likely chip reject percentage at first would be very high.
The Bitcoin Brothers have a “bright idea” but they sound like n00bs in the whole semiconductor fab biz. Going to a new, smaller grid sounds simple, but the smaller you go the harder you fall.
Personally, I’d be surprised if TSMC can produce anything like the BB chips at volume this year. And if they do succeed in proving out the fab, the first customer in line for volume production won’t be Bitcoin Brothers.
The “believable” chips looming on the near-term horizon are from Bitfury and Bitmain. When those hit all our existing miners will take a hit. S5’s will still mine ok for people like me, with very cheap power.
Side note: forget upgrading anything Bitmain made or makes. That would cause the company 10x more problems than it’s worth. None of the Bitmain boxes are elegant in any way, in fact quite the opposite. S5’s are a cobbled up hack. I love my S5 but it’s a single-purpose tool that has a limited life span.
Having said all that, I could be wrong. Usually I’m not when it comes to hardware manufacturing, but this time I could be. The bitcoin miner dev world has the potential to spring major surprises, but I don’t believe this is one of them.
Don´t underestimate TSMC…They held 58.7% of foundry market last year and reaching revenue near 15 billion USD. But, they are worried of Samsung now, which is much smaller, but have 14nm process in mass production now and growing customer base quickly.
They are exclusive maker for NVIDIA, most of the Qualcomm etc…They are definitely capable to manufacture those chips, if they tune up their FAB right.
And, there is also Samsung and Global foundries ready to accept orders…
Another fact is, that 14Nm wafers are twice more expensive than 28Nm atm…
Final price for chips could be too high now
But you are probably right with the early yield…they could have major problems getting acceptable levels of working chips volume and there is still question, whether they will have some priority in manufacturing before lot and lot of regular customers.
But is weird, if i look on one of the BB newsletter, there is this:
"We build our supercomputers, ASICs, and complimenting hardware with a well-established, successful global production line (of over 16 years) from design to full products, used by most German automakers for complex CPUs in their high-tech cars, by global players in the medical field (such as Siemens and GE), and others, producing over 275 million ASICs per year (i.e. over 5 million every week). The millions of our ASICs just add on top of that (256,000 per each MSEM). The same applies to the supporting hardware and components to make our supercomputers."
Yet, there is no reference to company, which they are talking about…
275 million of Asics for cars etc…Who, which company…??
Not sure but Thomas Ackermann (CTO) has a very serious and solid background. I’ve been looking into BB and the people involved since early December. Mark Welle the COO is apparently the one with chip design experience but I couldn’t find much background info on him. His LinkedIn is rather boring compared to Ackermann’s.
I agree and Q1 has officially passed and no EH super-computers been seen. I don’t see their claims becoming a reality in the “near” future. The available mining hardware is still sold in TH combined to PH on pools. I can’t see a new chip making the jump from TH to EH.
Some other hardware rumoured to hit the markets by July are a lot more realistic in their specifications than ExaHash super-computers. Then again I guess it’s a questions of how many chips you’re packing into your super-computers to put one ExaHash online.