My server configuration is:
System: Windows 10 x64
CPU: dual E5-2699 V4, 44 core 88 threads
Memory: 256GB
Hard drive: 11 * 7.68T U.2 Nvme high-performance solid-state drive
Network card: Intel X710 10G network card
Broadband: can provide over 10Gbps
The default startup and operation methods have a relatively low utilization rate for hardware and broadband. How can we optimize and improve efficiency by adjusting the startup parameters of Node and Farm?
1 Like
This is quite old CPU, modern 13900K/14900K with just 24 physical cores will likely be much faster than E5-2699 V4 and certainly MUCH more efficient, this is what software is primarily optimized for.
The default startup and operation methods have a relatively low utilization rate for hardware and broadband
Network usage will be fully utilized by your config, if it doesn’t use much bandwidth, it means it doesn’t need it.
Same for RAM, farmer is designed to use reasonable amount of RAM. It doesn’t minimize usage to compromise performance, but it also tries to not waste it for no reason.
As to CPU, you might get different results depending on number of NUMA nodes your CPU has and there might be some tweaks to do there. There were some discussions about this, but it wasn’t really a high priority for us right now. Also memory latency is important for plotting.
It would help to get any details at about about the utilization you are seeing though, so far that is a very generic description that can mean almost anything.
--farming-thread-pool-size 999 --plotting-thread-pool-size 999 --replotting-thread-pool-size 999
You can try this, this can make your CPU usage reach 999%
That makes no sense unless you have 999 cores. It might appear to use more CPU, but it was always slower from my testing at least.