Hi Pallas, George and everyone here.
So I just got this free subscription on Azure. and I decided to whip up a instance with 4 K80 GPUs (2 per card).
It had 4992 cores, 5000MHz memspeed, and 480 GB/s bandwidth, 24GB video memory on each card.
I just wanted to determine what they will give.
But I am shocked I am getting less than 24MH/s. Each GPU die is more powerful than a 1070 (in all the numbers), so I should have at least 40MH/s (even the worst 1070 here gave 10MH/s)
How can I optimise that? What I am not getting right?
The card is Compute 3.7 and I am using djm34 CUDA miner on Windows.
Thanks for all your help.
EDIT: OR does it have anything to do with setting the vardiff on Suprnova
It is an old architecture, also probably lower core clock (memclock matters very little).
When they will update to pascal based cards it might be interesting to try cloud gpu mining.
Thank you for your reply.
core clock is 875MHz and I set it on max.