Will this work with multi-gpu?

[astronyu]

I have an old mining rig with 6 gpu that I want to repurpose. Can this plugin work with multi gpu?

[dennix85]

easiest would be to use docker and assign each gpu to a node container, not many plugins are optimized to do multi gpu.

[astronyu]

easiest would be to use docker and assign each gpu to a node container, not many plugins are optimized to do multi gpu.

Thanks. I've reinstalled using docker and created individual nodes for each gpu. I guess I need to find a new machine maybe an used optiplex since tdarr seems to use lots of cpu as well eventhough cpu transcoding not enabled.

[dennix85]

Are u sure u are using a gpu Plugin?

Also note all stuff except video will be handled by CPU
They all will be single threaded.
And if u overload the job amount where all jobs start doing cpu stuff it can bottleneck if ur cpu is not fast enough.
And if u are using pcie risers those will be limited to x1 bandwidth and probably severely hurt your performance btw.. Unless U use some fancy plug the gpu's directly on the mobo mining board.
Try doing 1 job per node and see how that goes..
And u are sure it's cpu load and noting like iowait or similar?

[astronyu]

Are u sure u are using a gpu Plugin?

Also note all stuff except video will be handled by CPU They all will be single threaded. And if u overload the job amount where all jobs start doing cpu stuff it can bottleneck if ur cpu is not fast enough. And if u are using pcie risers those will be limited to x1 bandwidth and probably severely hurt your performance btw.. Unless U use some fancy plug the gpu's directly on the mobo mining board. Try doing 1 job per node and see how that goes.. And u are sure it's cpu load and noting like iowait or similar?

Thanks. I'm using an old mining rig with 6 gpu on a riser. I did set 1 transcode per node for 3 gpus and 1 healthcheck per node for the other 3. Does that means I'm better off unplugging those gpus and only use 1 to get better performance in transcoding?

[astronyu]

Are u sure u are using a gpu Plugin?

Also note all stuff except video will be handled by CPU They all will be single threaded. And if u overload the job amount where all jobs start doing cpu stuff it can bottleneck if ur cpu is not fast enough. And if u are using pcie risers those will be limited to x1 bandwidth and probably severely hurt your performance btw.. Unless U use some fancy plug the gpu's directly on the mobo mining board. Try doing 1 job per node and see how that goes.. And u are sure it's cpu load and noting like iowait or similar?

Weird enough, I'm using gpu plugin and I also assign individual gpu to each node container. However, during transcoding, out of the 5 gpu, only 1 is showing encoding activity while the others has no encoding activities.

[dennix85]

Do you mind hopping on to the official discord? I guess that would make it more easier to diagnose.
As many mails get unread or lost among those other hundreds 😊

[astronyu]

Do you mind hopping on to the official discord? I guess that would make it more easier to diagnose. As many mails get unread or lost among those other hundreds 😊

Thanks, I've posted in help-forum on discord.
https://discord.com/channels/623392507828371476/1346241839044493492