I’m thinking about mining on Raspberry Pi 3, because 4x1.2GHz and 1gb RAM seems to be nice for it.
I tried to download and make the nheqminer on it, but it fails with error “Could NOT find Threads (missing: Threads_FOUND)”. Probably the reason is the miner use SSE2 / AVX1 / AVX2, and these are only in x86 architecture, not ARM.
What do you think guys, how much work to rewrite the neccessary parts in nheqminer to mine on ARM based processors?
Other question: whether is worth it?
I’m very curious about it, maybe I’ll try to code it.
sure.
considering the differences between CISC and RISC architectures I presume you’d need to loose all the optimizations that made current cpu zec miners so effective.
If we can manage to port a gpu opencl miner we can mine on it. I am currently porting to os x and maybe to ios. Opencl is nearly the same on all plattforms.
I don’t think it would do much.
I ported the old solver to an ARM/Android phone, and the result was about a quarter of one CPU core for all 8 cores on the phone.
RPi, i expect, would do about as badly, but i don’t have one recent enough to try it on.
Probably the SSE2/AVX is the secret of PC CPU mining, but it’s in only x86 architecture.
Have ARM cores some kind of hardware specialty which great for equihash?
EDIT:
NEON is the SIMD instruction set equivalent to SSE2 for ARM.
Looked into this, yet its takes 27x+ longer to compute than that of SSE2 Mandelbrot Set with SIMD Intrinsics, so srsly don’t waste your time.
Not sure about this but I’m sure the bus speeds and L2 and L3 cache sizes has a huge impact on hash rates and I just cant see a RPi besting a mid-level cpu. I mean a newer laptop with an AMD chip that has 4cpu and 6gu cores was eeking out 5 sol/s
Say you can get 5 on each and you set up a 10 unit grid…thats alot of setup for something a midlevel GPU can do by simply plugging into a basic PC setup. An interesting and fun project, but not worth it from an efficiency standpoint
That came out wrong.
Originally meant to say that it takes 27x+ longer to compute than that of plain SSE2, so no its unfortunately not worth while in long run.