Ok, smooth.
But I can assure you that 23 threads on an 8-core Xeon is slower than 8 threads.
But I can assure you that 23 threads on an 8-core Xeon is slower than 8 threads.
Of course. I was talking about a 32 core instance. The post I quoted stated minerd at 3100%, which indicates 32 active threads. If you reduce threads below cores you will usually get better performance.
The GPU instances are even a bit more weird, but I don't have time to get into that.
I used wolf's cpuminer without specifying the number of threads which automatically launches nproc - 1.
The only thing that's not variable is nr_hugepages which I left at 21.
For ccminer i went with 12x64. That gave me the best results.
Edit:
Smooth, just saw your sugestion above. It should also for for setting the nr_hugepages with sysctl.
With that I think it will have good performance in most instance types.
5w00p, you're right. I went with the 8 cores myself. The max I got from a 32 core was 850h/s.
The 8 core gives you about 100h/s more for less bucks. And there is less competition too.
