|
Title: How much further performance are asics capable of producing? Post by: mackminer on January 11, 2014, 09:15:42 PM I suppose it's going to mean that eventually the technology will mature with current silicone and the limits of binary....?
Title: Re: How much further performance are asics capable of producing? Post by: Gator-hex on January 11, 2014, 09:25:52 PM The limit for a consumer product is the household mains ring.
KnC will breach the limits of standard US power sockets with their next batch and ship US under-clocked/cores disabled. They said this may be their last consumer unit. Title: Re: How much further performance are asics capable of producing? Post by: iglasses on January 12, 2014, 06:39:30 AM Well if you ask that guy Moore...
Title: Re: How much further performance are asics capable of producing? Post by: Pentium100 on January 12, 2014, 09:29:50 PM The limit for a consumer product is the household mains ring. There is a different limit to the performance. Of course, you can always have multiple chips or multiple devices and get very high hashrates, but the more interesting question is what performance per chip and per watt we will get when the SHA-256 ASICs catch up to the CPUs and slow down the increase in speed. The ASICs went from 330MH/s/chip to 400GH/s/chip (x1200 increase) or whatever the fastest chip currently is in the span of a bit more than a year. I don't think we are going to get 480PH/s chips next year.As this race reaches the capabilities of current technology, the latest ASICs should be able to earn a profit for a longer time than current devices do because difficulty won't be increasing as fast. Title: Re: How much further performance are asics capable of producing? Post by: mackminer on January 12, 2014, 11:16:19 PM The limit for a consumer product is the household mains ring. There is a different limit to the performance. Of course, you can always have multiple chips or multiple devices and get very high hashrates, but the more interesting question is what performance per chip and per watt we will get when the SHA-256 ASICs catch up to the CPUs and slow down the increase in speed. The ASICs went from 330MH/s/chip to 400GH/s/chip (x1200 increase) or whatever the fastest chip currently is in the span of a bit more than a year. I don't think we are going to get 480PH/s chips next year.As this race reaches the capabilities of current technology, the latest ASICs should be able to earn a profit for a longer time than current devices do because difficulty won't be increasing as fast. Apparently, 14nm is v. expensive and Kncminer plan to release this Neptune as there final product - due to the limitation on power apparently. I don't know much about processors but surely it's reached it's max nearly. The CPU's for example are gone multicore (of which seemingly will make little difference to SHA256 mining) - what other devices use ASIC's where there is a lot of money involved? Please bear in mind that my opinions expressed in this thread are v. uneducated and just based on general assumptions about the technology. Aren't CPU's the same as ASIC's apart from the fact that they just have different uses where the CPU would benefit from newer technology based on market size? What I'm saying is ASIC's are just as much a processing unit as a "CPU"....I'm probably being v. vague here but I've had one or two drinks. It's 11 at night here in W. Europe.... Title: Re: How much further performance are asics capable of producing? Post by: Ytterbium on January 13, 2014, 02:02:23 AM The limit for a consumer product is the household mains ring. KnC will breach the limits of standard US power sockets with their next batch and ship US under-clocked/cores disabled. They said this may be their last consumer unit. Well, if the device is modular, you can power different modules off of different circuits in your house. If you're talking about a single chip, you're also limited in how much heat that chip can dissipate. So a single chip sucking down 2KW seems pretty unlikely. However, you could still theoretically power it by using multiple circuits in a single house. You could also power it off of a 70 amp dryer plug or something like that. Title: Re: How much further performance are asics capable of producing? Post by: dogie on January 13, 2014, 02:03:29 AM The CPU's for example are gone multicore (of which seemingly will make little difference to SHA256 mining) - what other devices use ASIC's where there is a lot of money involved? Sha256 chips are extremely multicore, thats how you make 'fast' chips. Extremely parallel. Title: Re: How much further performance are asics capable of producing? Post by: Ytterbium on January 13, 2014, 02:05:54 AM Apparently, 14nm is v. expensive and Kncminer plan to release this Neptune as there final product - due to the limitation on power apparently. I don't know much about processors but surely it's reached it's max nearly. The CPU's for example are gone multicore (of which seemingly will make little difference to SHA256 mining) - what other devices use ASIC's where there is a lot of money involved? Please bear in mind that my opinions expressed in this thread are v. uneducated and just based on general assumptions about the technology. Aren't CPU's the same as ASIC's apart from the fact that they just have different uses where the CPU would benefit from newer technology based on market size? What I'm saying is ASIC's are just as much a processing unit as a "CPU"....I'm probably being v. vague here but I've had one or two drinks. It's 11 at night here in W. Europe.... bitcoin hashers are already massively multicore. The chips have tens or hundreds of individual hashing cores. Title: Re: How much further performance are asics capable of producing? Post by: miaviator on January 13, 2014, 02:13:56 AM I think as we move on to bio-electric hybrid chips with some mythological greek Zues power you will see a few PH per device easy.
https://bitcointalk.org/index.php?topic=412586.0 I'll delete this post upon request. Title: Re: How much further performance are asics capable of producing? Post by: miaviator on January 13, 2014, 02:16:39 AM Seriously though intel is testing out 13nm? chips
There is a physical limit to the standard chip size. Once that is reached or breached we should see some leveling. Also a lot of the market is driven by consumers and large mining farms. Consumers are driving up the price as non-knowledgeable peeps pay huge premiums for miners. Large corporate mines are buying up massive hashing power with massive non-disclosures and financial backing. I think individual mining will see a slow fall off to cloud and corporate mining just like no one ever intended. Title: Re: How much further performance are asics capable of producing? Post by: Bogart on January 13, 2014, 02:53:02 AM With chips already at 28nm/20nm, there's not much room left for gains by reducing the feature size. Future gains in efficiency will have to come from elsewhere.
I understand that the 55nm BitFury chips are already more efficient than the 28nm KnC chips. I'm definitely curious to see what BitFury comes up with for their next generation chips. Title: Re: How much further performance are asics capable of producing? Post by: minternj on January 13, 2014, 02:57:16 AM The problem really is that the cost savings with going smaller is just not there. Like bogart said, innovation comes from the asic design. Need to go fully custom to achieve that like bitfury did.
Title: Re: How much further performance are asics capable of producing? Post by: Ytterbium on January 13, 2014, 03:26:10 AM Seriously though intel is testing out 13nm? chips There is a physical limit to the standard chip size. Once that is reached or breached we should see some leveling. Also a lot of the market is driven by consumers and large mining farms. Consumers are driving up the price as non-knowledgeable peeps pay huge premiums for miners. Large corporate mines are buying up massive hashing power with massive non-disclosures and financial backing. I think individual mining will see a slow fall off to cloud and corporate mining just like no one ever intended. People have build transistors out of a single phosphorus atom. And Graphine transistors could potentially use far, far less power then silicon ones. There's a long way to go, in terms of absolute scaling before we hit a wall, although on an exponential axis it's probably not that much farther (so we can only 'double every 18 months' for so much longer) |