-
Notifications
You must be signed in to change notification settings - Fork 1.8k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Same GPU, different hashrate #64
Comments
Do I have any chance to get an answer for this issue? |
It could be that the second gpu is overclocked with nvidia-smi or you have
heat issues.
If you look to the power usage of the first k20c you can see that the power
consumtion is lower.
Could you please run `nvidia-smi -q`
|
|
As you can see the cards are identical |
You set affinity to true. This can mean that the gpu process is pined to
the wrong cpu socket. If thw gpu is hosted on the second socket und the cpu
process is pinned to the first than the connection between gpu and cpu is
slower.
Try to set affinity to false
|
I tried all the 4 combinations of affinity: TT TF FT FF
|
I checked the code. true is equal to false :-) add a value like 0 or 1 or
12 to pin to the core. depending on the pci slot you need to select a
core.try to use a core which is not used by the cpu miner
|
I don't get it. |
add only one number. The number is the number of the cpu core.
|
For instance, assuming I don't use CPU mining, I have 2 cpus with 16 cores each, so I can use 0 to 31 :
|
|
|
|
As you can see:
No improvement |
set one gpu to affinity 0 and the other to 6. But this depends on your
system configuration. Please run the newest hwloc-ls or check the mainboard
manuel which pci express slot is connected to which cpu socket.
|
from your last two results I can see that one gpu is on socket 0 and one on
socket 1.
My last suggestion should solve it.
|
I tried also 0 and 6 no CPU mining, nothing. :-( |
|
please use 21 and 15 and post the results. sry 0 and 6 are used by the cpu
miner
|
try also 1 for fpu 0 and 14 for gpu 1
|
GPU Solo - 6 and 0
GPU Solo - 21 and 15
GPU Solo - 1 and 0
GPU+CPU - 21 and 15
GPU+CPU - 15 and 21
|
|
I need to check if we have issues with the thread pinning for the gpus.
your results are strange.
Thx for the full test documentation
|
Thanks to you for your support |
Hi,
I have a pc with dual GPU and dual CPU, the GPUs are the same, but they are performing very different.
Here the hashrate:
Here the nvidia.txt:
Here the nvidia-smi output:
Can you explain why?
Thanks
The text was updated successfully, but these errors were encountered: