I have access to a 100 CPU diskless beowolf cluster. Does anyone know how compatibile zcash is with a setup like that?


#1

It's my understanding that zcash mines well on CPUs.

They are all 8 core diskless servers which run off ramdisks. It was build for genetic machine learning.

I'm curious if zcash benefits from shared mining or if they will work just as well all mining private zcash instances. It would be relatively easy to have all machines start mining zcash to a shared account.

It's my understanding that equihash is memory bandwidth constrained. I have a lot of memory bandwidth at my disposal. Has the algorithm been properly rebuilt for multicore? I had read that it wasn't yet properly parallelized.

Thanks for any help you can give. I have no doubt this will be the top google result for "zcash beowolf" since there aren't any now.


#2

Currently it isn't optimized for multicores but tests show roughly a x2.8 increase from 1 to 4 cores and then it drops off.

The biggest problem you will have is finding connections to the node/nodes if your only using one connection.


#3

hello you rented me :smiley:


#4

What do you mean by connections. They all have their own internet hookup.

I am planning to boot them from ramdisks running a core OS container pre setup to start running zcash at boot.

I don't see why the multi core problem can't be solved by running multiple clients, but I'm sure it is something more important than that. Not running on multiple cores makes it gpu resistant in a way, although I'm sure there's more to that as well.

Not sure if the miner is incomplete or if there is a real reason it is difficult to do multicore. Luckily I have a lot of machines.

Any news on pooling and if that is worth it to speed up hash rate?


#5

Ocminer is developing one and it currently works in testnet