Hi guys,
I really would like to have other algorithm support on the D3 as well, so here is an update from what I've experienced so far:
I tried to isolate a single algorithm / other algorithms, but not yet succeeded. There are some parameters to tweak with, which seem to change something in the ASICS, but the problem is that even if it switches to another algorithm (which I am not 100% sure), you can't know which one... It's also not that straightforward to change the linux part to support another algorithm, so it's easy to make a small mistake there as well.
If I had some inside information that for example groestl is 100% for sure supported then it would be easier to go 100% for that. Now there are too many variables to play with.
I still have some ideas left to try: I want to build an automated matrix test that tries all parameter combinations and tries to verify it against all known algorithms. It will be quite an effort to build this, but I think it's the only possibility to find out if it's possible.
I really would like to have other algorithm support on the D3 as well, so here is an update from what I've experienced so far:
I tried to isolate a single algorithm / other algorithms, but not yet succeeded. There are some parameters to tweak with, which seem to change something in the ASICS, but the problem is that even if it switches to another algorithm (which I am not 100% sure), you can't know which one... It's also not that straightforward to change the linux part to support another algorithm, so it's easy to make a small mistake there as well.
If I had some inside information that for example groestl is 100% for sure supported then it would be easier to go 100% for that. Now there are too many variables to play with.
I still have some ideas left to try: I want to build an automated matrix test that tries all parameter combinations and tries to verify it against all known algorithms. It will be quite an effort to build this, but I think it's the only possibility to find out if it's possible.
Blissz, so you can be 100% sure which algorithms are supported by x11. Here is a list.
blake
bmw
groestl
jh
keccak
skein
luffa
cubehash
shavite
simd
echo
X12, X13, X14, X15 and X17 are all exactly the same but with additional algorithms added in. Out of all the X11 algorithms listed above groestl should currently be the most profitable even at a severely reduced hash rate. If possible, I would try to focus on groestl first and then skein as a 2nd alternative. Cracking this problem of isolating a single algo is something that I believe many of us would be willing to support financially in a big way as we would all benefit in a big way.
Support!