4 ms·

ASIC inevitability is a falsehood. Software can be modified very quickly, whereas hardware requires much longer development cycles. Software can always outpace

by wes-exp 13y ago

ASIC inevitability is a falsehood. Software can be modified very quickly, whereas hardware requires much longer development cycles. Software can always outpace hardware, if the community is willing to commit to ASIC resistance by changing the software.

raverbashing 13y ago
That's why FPGAs exist And I'm sure existing ASICS can be adapted to new algorithms, they're developed in VHDL/Verilog, surely, not as easily as changing an FPGA, still, it won't be a work from scratch.
- Bootvis 13y ago
  They have to be build though, which is expensive. If the hashing algorithm is parameterized by the block number in some smart way it will be harder to deploy them. For example, one could use 10 different algorithms and choose a new one every block. The ASIC will only work 10% of the time but this can be implemented on the GPU. I agree that FPGA resistance is much harder to achieve.
- wes-exp 13y ago
  Give me a break. FPGAs aren't even useful for Scrypt (due to memory bandwidth requirements, as I understand it). ASICs have to go through a manufacturing cycle. Software doesn't. It is plainly obvious that software can move faster than hardware.
- oakwhiz 13y ago
  The only way to exclude FPGAs from mining a cryptocurrency is to create a problem that requires the average FPGA to completely (not partially) self-reconfigure faster than the average FPGA's configuration time in order to be competitive. Either that or somehow create a problem that requires more LUTs and logic cells than an FPGA is capable of supplying. But then you could just make multiple FPGAs work together. So you'd need some kind of latency based problem to make it difficult for multiple FPGAs to work together. Has anybody tried looking at making GPUs do what they are designed for? Namely sorting and rasterizing triangles. If a cryptocurrency was made to use a very large number of triangles as the problem, perhaps GPUs would be the most efficient device to solve it with.
  - sliverstorm 13y ago
    FPGA's rule themselves out. You can use them for prototyping your ASIC, but FPGA's are extremely costly per silicon area, and generally less power efficient than anything else. Oh, and nobody is using triangles because hashing lends itself so well to Proof-of-Work. The "work" must be difficult to perform, but very easy to verify. Like a zip file that takes 20 minutes to compress, but only 5 seconds to unzip. To verify a triangle you would have to re-render it.

by wes-exp 13y ago

raverbashing 13y ago

That's why FPGAs exist And I'm sure existing ASICS can be adapted to new algorithms, they're developed in VHDL/Verilog, surely, not as easily as changing an FPGA, still, it won't be a work from scratch.

Bootvis 13y ago
They have to be build though, which is expensive. If the hashing algorithm is parameterized by the block number in some smart way it will be harder to deploy them. For example, one could use 10 different algorithms and choose a new one every block. The ASIC will only work 10% of the time but this can be implemented on the GPU. I agree that FPGA resistance is much harder to achieve.
wes-exp 13y ago
Give me a break. FPGAs aren't even useful for Scrypt (due to memory bandwidth requirements, as I understand it). ASICs have to go through a manufacturing cycle. Software doesn't. It is plainly obvious that software can move faster than hardware.
oakwhiz 13y ago
The only way to exclude FPGAs from mining a cryptocurrency is to create a problem that requires the average FPGA to completely (not partially) self-reconfigure faster than the average FPGA's configuration time in order to be competitive. Either that or somehow create a problem that requires more LUTs and logic cells than an FPGA is capable of supplying. But then you could just make multiple FPGAs work together. So you'd need some kind of latency based problem to make it difficult for multiple FPGAs to work together. Has anybody tried looking at making GPUs do what they are designed for? Namely sorting and rasterizing triangles. If a cryptocurrency was made to use a very large number of triangles as the problem, perhaps GPUs would be the most efficient device to solve it with.
- sliverstorm 13y ago
  FPGA's rule themselves out. You can use them for prototyping your ASIC, but FPGA's are extremely costly per silicon area, and generally less power efficient than anything else. Oh, and nobody is using triangles because hashing lends itself so well to Proof-of-Work. The "work" must be difficult to perform, but very easy to verify. Like a zip file that takes 20 minutes to compress, but only 5 seconds to unzip. To verify a triangle you would have to re-render it.