5

I have a lot of identical nodes all connected to the SXMR pool. They are CPU mining on 40 real cores (hyperthreading is disabled in BIOS) but I am finding that some nodes are performing far slower than other ones.

Example of Node with Low Hashrate

My question here is, am I doing something incorrect? Should I not connect the ~100 nodes I have to the same pool? Could this simply be a performance impact from having all of these nodes making an upstream connection? Is this just a matter of hash availability?

I'm aware that I'm not getting the maximum I could out of these due to 1 in every 10 nodes not performing anywhere near as well as the others.

stuts
  • 51
  • 3

2 Answers2

3

The problem may be that CryptoNight is a memory-intensive PoW algorithm. For effective mining, you need 2MB of L3 cache per core.

Here's an interesting answer describing CryptoNight.

scoobybejesus
  • 5,495
  • 18
  • 42
  • 1
    Thank you for your response, I'm still a bit confused as to how this would effect only some of the nodes (seeing as they're all identical). If it is a matter of not having enough L3 cache per core then all of the nodes should be effected by this. – stuts Nov 24 '17 at 11:05
1

If you have hundreds of miners, you should have a look at xmr-node-proxy, it reduces load on the pool servers and allows for easier configuration of your miners.

Matthew Smith
  • 451
  • 3
  • 9