"Maximum number of clients reached" with > 6 GPUs (well, > 13 instances) about silentarmy

[jtremblaytdst]

Hi, i have the same problem. If i run 1 instance per gpu per terminal i can get 14 client for 7 gpu but if i run more gpu per command per terminal i can only run 13.

Someone have a trick to have more instance by the way :D

Is it the driver maximum or the software maximum ?

Thanks !

Ubuntu 14.04
Radon Crimson 15.12 Drivers
Zogminer or SilentArmy miner.

[mbevand]

If someone can provide me with ssh access to a system with 7 GPUs I will fix the problem in silentarmy.

[jtremblaytdst]

I can provide you a ssh, but do you thing we can have more than 14 instance ? Cause with 4 instance my gpu get 50-60S/s. Can we have 28 instances ? or 14 is the maximum ?

Thanks

[mbevand]

I don't even know at this point what is causing the limit, so I can't answer. Some people claim it's a maximum number of Xorg clients, but I was unable to verify. The silentarmy software itself should work with an unlimited number of instances.

[jtremblaytdst]

Can you send me a private message ? We need to chat on some platform to do that ? Telegram ?

[mbevand]

Email? [email protected]

[jtremblaytdst]

Thanks

[padrino121]

It's not silentarmy itself but something related to Xorg most likely... I say that because I have a number of 8 GPU systems where silentarmy runs 2 instances per card for all 8 cards

[padrino121]

On those systems they are using amdgpu without X running

[cesarreynab]

@padrino121 so the problem could be the radeon drivers 15.12...

[jtremblaytdst]

@padrino121 are you able to have more than 2 instance per gpu ? maybe the maximum number of client is 2 per gpu ?

[outragedhuman]

@padrino121 Is this maybe a ulimit (or similar) issue, where too many processes are running, or too many files are open? I would verify how many processes you're allowed to run (using ulimit) and then check the number of processes currently running as your user to make sure you're not even close.

Edit: Should have pointed out that these checks should be done while you are at full capacity (13 instances)

[jtremblaytdst]

@thekashkings i got this 👍

core file size (blocks, -c) 0
data seg size (kbytes, -d) unlimited
scheduling priority (-e) 0
file size (blocks, -f) unlimited
pending signals (-i) 47784
max locked memory (kbytes, -l) 64
max memory size (kbytes, -m) unlimited
open files (-n) 1024
pipe size (512 bytes, -p) 8
POSIX message queues (bytes, -q) 819200
real-time priority (-r) 0
stack size (kbytes, -s) 8192
cpu time (seconds, -t) unlimited
max user processes (-u) 47784
virtual memory (kbytes, -v) unlimited
file locks (-x) unlimited

What parameter can i change to have more instances.

I have 237 process running right now and the maximum is 47K.

Thanks

[padrino121]

I suspect it's limited to fglrx and Xorg, I cannot say for sure since I don't have anymore of my 3XX GPUs on hand but I do recall some limits on the maximum number of OpenCL programs running simultaneously with fglrx. I pulled up a note on the AMD forums on the topic and found this:

This can be fixed by starting X with the -noreset command. One of our developers starts X with the following command to fix the issue:
nohup /usr/bin/X11/X -ac -noreset &

This is only a solution if X is only being used for running OpenCL apps and not if you are using X for other uses.

[ddobreff]

By default kernel supports maximum of 16 GPUs...you can recompile changing CONFIG_VGA_ARB_MAX_GPUS=16+

[duxr]

ddobreff the kernel of fglrx drivers or opencl you recompile with this arg?

[ddobreff]

kernel...sorry but I don't use fglrx driver. I'm on 16.04...

[duxr]

Ok