Topic: The next ACSI? (Read 5885 times)

You obviously don't know what you are talking about.  Did you even read the link you posted as 'proof'?  TSMC is a foundry and accepts orders to use their foundry.  Apple is even considering using TSMC for their A7 processor:
http://www.zdnet.com/apple-rumored-to-be-dumping-samsung-in-favor-of-tsmc-for-a7-processors-7000013834/

Lots of big companies are fabless (such as Apple).  Xilinx & Altera have even made their FPGAs at TSMC in the past.  Although I've lost track where the Virtex7 is made now:
http://www.eetimes.com/electronics-news/4087890/Xilinx-confirms-Samsung-TSMC-in-UMC-out-at-28-nm
http://www.eetimes.com/design/programmable-logic/4412103/Altera-turns-to-TSMC-s-embedded-flash-process

Anyone can buy foundry time.  You can even run small MPW lots using MOSIS.
https://www.mosis.com/db/pubf/fsched?ORG=TSMC

Thanks VM1990 and ihtpf, you beat me to the punch.  Just pitching in to say that these guys are right and fill in some detials. 

By definition, ASIC's (remember this just means App Specific IC's) cannot be beat except by newer / better ASIC fab techniques which incrementally come over time.  Reason: by designing operations at the die/transistor level, you don't have any unnecessary operations wasting die space, and you can finely tune the chip physically (rather than with software) so that the specific calculation you require is fully completed (ideally) in a single clock cycle rather than chunked across multiple beats which eats buffer space and leaves portions of the chip idle.  Then you can line these up in parallel to start to really rock and roll quick style.  There are various different kinds of ASIC platforms and a design team must pick one taking into account the many cost / performance tradeoffs while also going with a platform your engineers are familiar enough to work with.

Also worth noting:  It's not new.  It's an old chip design philosophy which was employed to varying degrees for many of the historically significant computer products from the 80's forward, and today silently drives many of the limited purpose mass produced electronics people use in day to day life.  It also powers certain niche high performance applications where people will pay premium prices to cover low volume - which is the case i think with BTC mining at the moment...  Normally, the development of an ASIC chip only happens when high volume purchasers exist to cover the design and fab costs, as they aren't usually affordable at low volume.  Markets below a certain size are usually forced to use general purpose off the shelf chips like CPU's and GPU's which while quite flexible - will always consume more power and be slower than a custom ASIC replacement designed for the problem at hand.

So it's a sign of maturity that BTC mining is a big enough market to finance fully custom ASIC chips, Huzzah!   AFAIK: ASIC's can only be beaten with better ASIC's. 

Only downside for ASIC's:  They are kind of useless outside of their design purpose unlike a general purpose chips which can surf porn.  Now there's the dream ASIC chip - when will the Porn-ASIC be developed??  (lol)

ok so the Avalon chips are made by TSMC which make chips but dont do custom orders they develop there own chips and you buy them off them...

proof they use TSMC chips
https://en.bitcoin.it/wiki/Avalon#Chip_Specification

and proof they build there own chips not to custom orders http://www.tsmc.com/english/dedicatedFoundry/technology/future_rd.htm

An ASIC is just a custom computer chip, designed for an intended application.
http://en.wikipedia.org/wiki/Application-specific_integrated_circuit

As more and more time goes by, butterfly labs is really starting to look more and more like a scam, that or, butterfly labs bit off more then they could chew by designing and mass producing an extremely delicate and complicated chip, therefore currently as much as they may want to release their asic minings rigs they're probably having serious reliability issues due to poor engineering from the ground up. All that being said, a worse case scenario, they may have already successfully built all mining rigs for batch/round 1 sales and they're simply mining with them until no longer very profitable todo so. Either way you look at BFL there's just something not right. IMHO

I don't think so -- unless you are arguing semantics.  What would be the part number for it be if its off the shelf?  And where could you buy it from besides Avalon?
http://store.avalon-asics.com/?page_id=9605

I also have considered the Parallela design.  Assuming C1D guesstimates about the Epiphany's hashrate are close enough, you could achieve 1Gh/s by putting together a bunch of Parallela boards (it would take 122 boards to achieve 1Gh/s, by my calculations, and I didn't take into account any kind of overhead required for processing the Bitcoin/getblocktemplate protocols).  122 x $100 = $12,200.  Each of those boards would require at least 5W, but probably more than that given the expected heavy load.  We'll be kind and leave it at 5W.  Compare Avalon #3: 63 Gh/s @ 600W at a cost of $5000-$6000 (depending on exchange rate, but this is close enough for now).  If you scaled up the Parallela board to produce a comparable hash rate, you'd be paying $769,000 for the hardware, and you'd need 31.5kW to run them (not to mention a place to put them and additional equipment and electricity to cool them).  This is also all assuming you can purchase or manufacture this many boards at a cost of $100 each.

Think you might be a bit wrong on some info the butterfly labs chips are the only ones designed from the ground up to be bitcoin mining chips the ASIC ones are just off the shelf ones (hence the reason theres so many on each board)

I see your point, I'm quite new to the world of ASIC and FPGA's and I so I thought this could compete. I will still be using this as a server and a second board for a custom PC. Also can ASIC's run applications or are they designed for mining?

Hi,
   This looks interesting, but I don't think it will be able to compete with the Avalon ASIC.  The 64 cores are certainly nice, but a single Avalon ASIC can do 275 Mhash/s. 
   Even if each Epiphany-IV core can produce one full hash every 640 clock cycles, it will only generate 80 Mhash/s.  I'm more of an FPGA/ASIC person, so I'm basing the 640 clock cycles on 10 clocks per individual SHA256, and 64 rounds per hash, which might even be low. 
   I think it is hard to compete with the Avalon ASIC custom chip, which was custom designed for mining.
 
Regards,
Pat

Hello,
Before I begin; parallel computing is the way in which you can run two or more programs at the same time with out having to dedicate a thread to each program.
I've been reading up about Parallella -a new parallel supercomputer- which is now priced at $99. This board uses Parallel computing to run and is amazingly fast.
It comes with a normal ARM CPU but its the Microprocessor that does it all. It comes with the Epiphany-IV 64-core 28nm Microprocessor (E64G401) (64 Cores!). The Epiphany-IV 64 runs at 100 GFLOPS and 800MHz Operating Frequency, you can see the rest of the specs here: http://www.adapteva.com/products/silicon-devices/e64g401/.
It comes with Ubuntu pre-installed and I think this could change Bitcoin Mining and beat ACSI! What do you think about, I'm on the waiting list once they're ready to sell these I'm defentily getting one!!
The main point that hits home is the fact that It will only run on 5 Watts and will be the size of a credit card!
Here is the Main Site: http://www.adapteva.com/.
Here is the Parrallella: http://www.adapteva.com/products/eval-kits/parallella/.