Ninjastic
PostEdited versionsQuotes to this post
Post
Topic
Board Mining (Altcoins)
Re: [ANN]: cpuminer-opt v3.6.5, open source optimized multi-algo CPU miner
by
Elder III
on 14/07/2017, 21:17:10 UTC
Quote from: joblo on July 12, 2017, 09:35:34 PM
Quote from: Elder III on July 12, 2017, 08:07:44 PM
Quote from: joblo on June 29, 2017, 04:59:12 PM

I am still curious about AVX2 performance on Ryzen. I have read that their implementation isn't native but I haven't
seen any performance comparisons. If anyone has a Ryzen I would be interested in a performance comparison of
AVX vs AVX2 on algos that have AVX2 optimizations.


If you can list the algos you would like tested I can try it out on both Ryzen R7 1700X and R5 1600 CPUs. Possibly on Threadripper too towards the end of summer if I talk myself into getting one. Wink

Lyra2v2, Lyra2z and Deep have the most AVX2 and should see the biggest improvement. All the X algos, qubit, timetravel
and a few others also have varying amounts of AVX2 code.

Preliminary results:

Ryzen 1700X @ stock settings with 8 threads set in the batch file.

Lyra2v2 = 1001 Kh/s on avx2
             = 1016 Kh/s on avx
Timetravel = 799 Kh/s on avx2
                = 843 Kh/s on avx
X11          = 686 Kh/s on avx2
                = 711 Kh/s on avx
X17          = 294 Kh/s on avx2
                = 294 Kh/s on avx

X17 was the same for both options, but I was surprised to see avx beating avx2 for the other algorithms tested. My understanding was that it would likely be the other way around....