3 ms·
Couldn't you defeat ASICs pretty definitively (at least for a long time) by making the algorithm dependent upon a gigantic completely random substitution table?
by api 9y ago
Couldn't you defeat ASICs pretty definitively (at least for a long time) by making the algorithm dependent upon a gigantic completely random substitution table? It would at the very least require ASICs to have giant amounts of ROM or RAM (with a memory controller etc) which would multiply ASIC cost.
AFIAK for today's ASICs a table larger than a few megabytes would be a strong deterrent. Go really big (e.g. 64mb) and you'd have some margin.
This would also be completely deadly to GPUs, especially if combined with branches and complex algorithms.
Edit: even worse: perturb the table using entropy from the block chain, making it impossible to burn into ROM and requiring tons of on-board RAM or a memory controller.
- tatersolid 9y agoThis is exactly what “memory-hard” hashing algorithms like scrypt, Argon2, and ballon hashing already do. You can tune their parameters to use GBs of memory if you wish.