loot loot_x4 14577.78 15988.83 37 9/24/2017 8:49:29 PM <- Ryzen 1700x 1.0.4.0 64bit after 2 hours+ mining (take a look at the CPU load)
Just bumped the genproclimit to 32, now im getting 32-35% CPU load again (and probably 20-25k HPS2) just check the leaderboard in a few hours.
YES, I did some tests with lower genproclimits too, but the hashrate/shares/CPU load drops by 30-40% when using anything lower than 32.
Waiting for Oliwer's genproclimit 11 test on the 32bit wallet
EDIT: 30 minutes after switching back to genproclimit 32-> HPS1: 24189.32 HPS2: 29913.39
Now testing 1.0.4.0 on genproclimit 11...
1.0.4.0 64bit with genproclimit 32 - HPS 24k and HPS2 25k-ish with CPU load 32-37%. My HPS2 drops to 15k with SMT disabled btw. Did you disable it?
Thanks for the testing. So in a nutshell you are saying the Ryzen will only use 35% of the processor no matter what proclimit is set, and both the 32 bit and 64 bit builds are essentially the same.
Ok, that sounds like a different problem than I had on the windows xeon 2 proc. The xeon on 64 bit only used 50% of the proc, but on 32 bit I got it to use 90% of the proc and on the pool it solves roughly what the linux equiv box solves.
Additionally, non use of resources is a different problem than was discussed after f7000 went live. The problem we looked at before was the hash rate drop on the servers compared to the laptops and the shares solved per HPS- while assuming the whole processor was being utilized. This is more of an issue where the processor is not being utilized.
I think I will have time to take a look at making a test miner that holds a reference to two KJV bibles.
In the mean time, the other thing that no one has reported is what happens when you run two copies of BBP against two different datadirectories on the Ryzen? If anyone can test that out and report back that would be immensely helpful also. I think inblue was going to try that on a server at one point.