Open iamubt opened 6 years ago
hello, sure, add parameter -t 51 (this is 80% of 64) and it will use 51 threads. Of course I havent tested such a big configuration, my biggest CPU is a ryzen, but it should work.
However the best performance is not always with the max CPU usage: too many threads will flood your cache and lead to worse performance. Cryptonight is an algo that need more cache than compute power.
thanks for giving time.
But when i add -t 51 parameter it start give 800 H/s which is less than when using default configs.
@iamubt I am using it on a dual xeon setup (old E5649) and I just took the time to setup the config file for the cores I wanted to use. Works like a dream and I am able to pull 1.5kh/s on cn-lite with 12 cores. I get around 800h/s on CNv7.
i warned that more thread does not lead to more speed, because you're flooding the cache. try to add threads one by one, or use the manual config file to enable some no-cache threads.
i give example in the GitHub page, that's for ryzen but the principle applies to Xeon too.
Your goal is to use most of the cache and all of the physical cores, not 100% CPU usage. With only 64MB L3 cache that's probably 32 2MB CNv7 threads. Of course your OS does stuff too, so you probably will get better speeds with 31.
@PsychoSterope can you please share the config which gives you 1.5kh/s. It will be really helpful. I am finding difficulty in setup up. Any help will be appreciated.
@iamubt My xeons are 6 core 12 thread each, I am using the odd cores (although I don't think it matters), but I get up to 1.5kh/s on CN-Lite. Average last couple of days has been 1.35kh/s.
"cpu_threads_conf" : [
{ "cpu_architecture" : "auto", "affine_to_cpu" : 1, "use_cache" : true, "multi_hash":2 },
{ "cpu_architecture" : "auto", "affine_to_cpu" : 3, "use_cache" : true, "multi_hash":2 },
{ "cpu_architecture" : "auto", "affine_to_cpu" : 5, "use_cache" : true, "multi_hash":2 },
{ "cpu_architecture" : "auto", "affine_to_cpu" : 7, "use_cache" : true, "multi_hash":2 },
{ "cpu_architecture" : "auto", "affine_to_cpu" : 9, "use_cache" : true, "multi_hash":1 },
{ "cpu_architecture" : "auto", "affine_to_cpu" : 11, "use_cache" : true, "multi_hash":1 },
{ "cpu_architecture" : "auto", "affine_to_cpu" : 13, "use_cache" : true, "multi_hash":2 },
{ "cpu_architecture" : "auto", "affine_to_cpu" : 15, "use_cache" : true, "multi_hash":2 },
{ "cpu_architecture" : "auto", "affine_to_cpu" : 17, "use_cache" : true, "multi_hash":2 },
{ "cpu_architecture" : "auto", "affine_to_cpu" : 19, "use_cache" : true, "multi_hash":2 },
{ "cpu_architecture" : "auto", "affine_to_cpu" : 21, "use_cache" : true, "multi_hash":1 },
{ "cpu_architecture" : "auto", "affine_to_cpu" : 23, "use_cache" : true, "multi_hash":1 },
]
Hi,
I have a XEON processor, which have 32 cores and 64 logical processors. But when I run the miner with default config it only utilizes 31 threads (57% cpu consumption) and giving only 1000H/s. I want to utilize 80% of the cpu. Can you provide me a way for this? and please give me a sample config.![cpu](https://user-images.githubusercontent.com/40821007/42263733-519e64be-7f88-11e8-87b0-302ae006ce56.PNG)
Thanks