Open irungentootoo opened 10 years ago
I've also been curious and just been looking into whether FPGA or ASICs could be at all used for vanity-gen - they dont use the same logic so ASICs would be hardest to make work with it, if at all (in case you're not that technical, ASICs are app-specific as the acronym indicates, and impossible to make significant changes to their logic).
I was wondering if any of the ASIC designs (ideally the cheaper ones but unlikely) would be modular enough to work around/only call specific parts of it, but I strongly suspect not (writing as an experienced C/C++/Python software dev not a hardware techy though I've done some basics at university many years ago): probably the ASICs are just too different from whats needed, even if you could make it believe the "block" to be hashed was your nonce-incremented address based on the partial public key; setting the difficulty could be set to the highest possible to match more exactly but I think its matching the wrong thing and couldn't be made to match the right one. Disclaimer: I may slightly misunderstand the algorithm, having only read it a couple of times :)
FPGA's have more potential, and some of them might even support OpenCL so need less porting effort - or MPI or other relatively high level languages/systems to program them in. I only own a lower level GPU card - still 20x better Hashes/sec than my AMD dual-core CPU, but I want faster without spending well into the hundreds or more and not sure I've got enough spare 4x/16x/.. PCIe slots :)
I did have more of a read about the (Kickstarter'd) Adapteva Epiphany - basically a couple(iirc) of ARM cores, Xilinx chips and some FPGAs. Comparing their CoreMark score to my CPU/badly-scaled-up hypothetical GPU score (tried to get one, needs work) didn't seem enough value for money, plus they've been slow/uncommunicative at times and its too late to pre-order on their site and hope to get one soon.
Shouldn't we move this to bitcointalk.org maybe (there's an existing vg. thread there too).
On 13 January 2014 07:16, irungentootoo notifications@github.com wrote:
How about butterflylabs devices, it posible to use in vanity?
— Reply to this email directly or view it on GitHubhttps://github.com/samr7/vanitygen/issues/44 .
Mining ASICs are useless for vanitygen. Mining uses sha256 and address generation uses sha256 and ripemd160.
You could program an FPGA, but I don't think anyone has.
The forums would be a better place for this.
I'm already aware ASICs are unlikely to allow you to farm out just the sha-256 portion and report that back to try and speed it up; they'd have to be more modular to be usable for vanity-generation at all, which would mean less efficient at blockchain mining (much more data throughput). Which manufacturer's going to do that when that market is so big and so competitive that every MHash (or GH now) matters per watt? About none of them. Vanity-address market is pretty small in comparison so far.
An fpga miner for it exists but its only a proof of concept though it works. Allegedly; seen one other (than the author) report it working. I'm hoping to be able to improve it to get notably better than GPU performance out of it, as currently it seems almost less efficient and a lot slower than any typical OpenCL-supporting GPU of last couple of years.
I posted to bitcointalk in the existing vanitygen thread, I agree lets keep replies there from now unless its to do with this code.
On 26 January 2014 22:28, Bryan Stitt notifications@github.com wrote:
Mining ASICs are useless for vanitygen. Mining uses sha256 and address generation uses sha256 and ripemd160.
You could program an FPGA, but I don't think anyone has.
The forums would be a better place for this.
— Reply to this email directly or view it on GitHubhttps://github.com/samr7/vanitygen/issues/44#issuecomment-33332744 .
How about butterflylabs devices, it posible to use in vanity?