praxeology_guy [ARCHIVE] on Nostr: 📅 Original date posted:2017-04-07 📝 Original message:Daniele Pinna, Can you ...
📅 Original date posted:2017-04-07
📝 Original message:Daniele Pinna,
Can you please not forget to supply us more details on the claims made regarding the reverse engineering of the Asic chip?
gmaxwell told me that back even in S7 chips its possible to set the SHA256 midstate/IV instead of just resetting it to the standard SHA256 IV. This essentially allows you to re-use midstates, which is one of the key necessary features for the ASICBOOST optimization to work. From the chip's perspective there is not much difference between the covert and overt optimization methods, particularly given that the whole IV/midstate vector can be set.
The covert method just requires more work than the overt method:. overt you just permutate the version bits, vs the covert one requires you find partial hash collisions of the tx merkle root. The extra work to find the partial tx merkle root hash collisions could be done at different stages in the mining system... some speculate that it could be done in the miner's FPGA.
Not sure how exactly gmaxwell (or his friend) did it. I don't currently own any mining hardware nor the time to do it myself.
Cheers,
Praxeology Guy
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linuxfoundation.org/pipermail/bitcoin-dev/attachments/20170407/1a863b5f/attachment-0001.html>;
📝 Original message:Daniele Pinna,
Can you please not forget to supply us more details on the claims made regarding the reverse engineering of the Asic chip?
gmaxwell told me that back even in S7 chips its possible to set the SHA256 midstate/IV instead of just resetting it to the standard SHA256 IV. This essentially allows you to re-use midstates, which is one of the key necessary features for the ASICBOOST optimization to work. From the chip's perspective there is not much difference between the covert and overt optimization methods, particularly given that the whole IV/midstate vector can be set.
The covert method just requires more work than the overt method:. overt you just permutate the version bits, vs the covert one requires you find partial hash collisions of the tx merkle root. The extra work to find the partial tx merkle root hash collisions could be done at different stages in the mining system... some speculate that it could be done in the miner's FPGA.
Not sure how exactly gmaxwell (or his friend) did it. I don't currently own any mining hardware nor the time to do it myself.
Cheers,
Praxeology Guy
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linuxfoundation.org/pipermail/bitcoin-dev/attachments/20170407/1a863b5f/attachment-0001.html>;