Is it normal that hashrate per card decrease with more GPUs attached to single MB ?
I have 3 systems for testing, each with 4/5/6 GPUs configuration running on identical HW, SW.
Each card is attached via powered riser (they are not all same, different batches) to PCIe x1 slot.
I can see that system with 4 cards performs best in the meaning of hashrate per card, 2nd is 5GPU and 3rd is 6GPU.
Difference on equihash algo with ccminer 2.2.0 on linux is ~10% (best/worst).
What could be a bottleneck ?
It is not really a problem but I would like to know why it performs like that.
I have 3 systems for testing, each with 4/5/6 GPUs configuration running on identical HW, SW.
Each card is attached via powered riser (they are not all same, different batches) to PCIe x1 slot.
I can see that system with 4 cards performs best in the meaning of hashrate per card, 2nd is 5GPU and 3rd is 6GPU.
Difference on equihash algo with ccminer 2.2.0 on linux is ~10% (best/worst).
What could be a bottleneck ?
It is not really a problem but I would like to know why it performs like that.
I also saw this on my rigs but didnt find out why... until 4 cards everything runs perfect, after adding more hashrate drops and seams a bit instable. Like 1070 34,1 Eth 5th card add only 32.2 otherwise crashes...
Or 1000MHZ Memory overclock after 5th card only 900Mhz...Maybe Mainboard (i use Asus) or because of the 16 pcie lanes? CPU i tried Celeron and Pentium all same behavior.
Anybody else?