Re: [JCE] Ultrafast CN-Heavy/Tube/HVX miner, low power, Vega56 1750+ h/s
I won't comment about the speed, happy if you got it better, but about the settings: Cast, SRB, TeamRed and JCE are close source miners, with different code each. We obviously don't use the reference code (the one of XMRig/stak). So we have each our own best settings, and we don't even use the same units, SRB intensity setting (a number somewhere between 0 and ~100) is unlike my multi_hash which is often arount 500, 1000 or 2000, alpha and beta. For each miner, take time to tune, even between two revisions of same miner.
On GitHub I often get tickets from people who complain they cannot apply one-to-one their xmrstak config. I don't know how and where to say we're two unrelated software
0.33b6 done, now testing. Planned fixes:
On GitHub I often get tickets from people who complain they cannot apply one-to-one their xmrstak config. I don't know how and where to say we're two unrelated software
0.33b6 done, now testing. Planned fixes:
Quote
OpenCL O-2.10 bug solved
More stable hashrate
Little perf boost on all algos ~0.15%
Temperature Watchdog (AMD GPUs only)
More stable hashrate
Little perf boost on all algos ~0.15%
Temperature Watchdog (AMD GPUs only)
First of all, fantastic miner!
What I noticed very quickly was that JCE multihash figures are very similar to intensity figures shared by XMR-Stak and XMRig. When comparing SRB to them, I noticed that SRB intensity is 1/16th of what the open source miners are. By adapting SRB, I took the numbers I used and multiplied them by 16 and the miners seem to go ok. The number I couldn't use though was worksize 9. When I do that, JCE insists that multi_hash be divisible by 16 and by the worksize 9. If it could have been one or the other, I could experiment more. No worries if you want to leave it the way it is. Tuning is just a bit of a hobby of mine
