Re: [ANN] dstm's ZCash / Equihash Nvidia Miner v0.5.8 (Linux / Windows)
I actually advocated EWBF earlyer, but sticking with DSTM now 
Test ended with average 3049 EWBF vs 3099 DSTM. 50 free hashes? Sounds good for me... --fee 0 didn't helped EWBF. And at least in my case i still see 2% less shares then reported by miner. And even some say they see 1% reduce it still does not save EFBW.
And this is GDDR5X cards witch presumably should have worked better on EFBW.
P.S. I am not related to DSTM or EFBW in any way, long time used both but recent discussion in DSTM thread pushed me to do a test myself. Believe me or not, this was exactly same machiene of 6x1080 (non ti) not tweaked in any way for one or another miner, and it was not even restarted during test period.
https://imageking.eu/images/dstmcompar.png large image
So for us miners it is actually bad, DSTM does not have competition and would get out even with 3% fee....
Test ended with average 3049 EWBF vs 3099 DSTM. 50 free hashes? Sounds good for me... --fee 0 didn't helped EWBF. And at least in my case i still see 2% less shares then reported by miner. And even some say they see 1% reduce it still does not save EFBW.And this is GDDR5X cards witch presumably should have worked better on EFBW.
P.S. I am not related to DSTM or EFBW in any way, long time used both but recent discussion in DSTM thread pushed me to do a test myself. Believe me or not, this was exactly same machiene of 6x1080 (non ti) not tweaked in any way for one or another miner, and it was not even restarted during test period.
https://imageking.eu/images/dstmcompar.png large image
So for us miners it is actually bad, DSTM does not have competition and would get out even with 3% fee....
Why do you say "GDDR5X cards witch presumably should have worked better on EWBF"? Based on what? I have 2 x MSI gtx1080 gaming which are GDDR5X and they do slightly better than my 1070's in DSTM, but much better in EWBF - but still very similar (low 500's sol when at powerlimit 70, core +100, memory +50) in both.
When I mine neoscrypt my 1070s actually do better (!) which is crazy but I realize there is some kind of memory latency issue with GDDR5X. Has anyone worked out any way around this? Is there any way we can set or modify memory timings or anything else to mitigate this? According to CUDA processing power the 1080 should really do around 600 sol but the stupid memory is getting in the way. Maybe I will start a thread to ask about any possible solutions.