CL_OUT_OF_RESOURCES error when running on more than 3 GPUs

Question

CL_OUT_OF_RESOURCES error when running on more than 3 GPUs

Closed this issue 6 years ago · 21 comments

Hello Jonano614, I was able to successfully run the gpu miner on my mining rig using 3 GPUs and it worked fine. However, when I tried to run it on more than 3, I was getting the error. Following is the log output.
Note: In one command window, the executable is already running with 3 devices. I then ran the following in a new window. (Note that if I just run it with all the GPUs, for the 4th, 5th and all other GPUs, it shows 0 Mh/s.

Appreciate if you can look into it Thanks for your work!

DaggerGpuMiner.exe -G -a GdVbEseUZef0hM4C2DeYALflaTE1xtOH -p 77.55.222.85:443 -opencl-device 4
i 23:15:54|main Founded OpenCL devices:
i 23:15:57|main GeForce GTX 1080 Ti with 11811160064 bytes of memory
i 23:16:04|main GeForce GTX 1080 Ti with 11811160064 bytes of memory
i 23:16:10|main GeForce GTX 1080 Ti with 11811160064 bytes of memory
i 23:16:16|main GeForce GTX 1080 Ti with 11811160064 bytes of memory
i 23:16:22|main GeForce GTX 1080 Ti with 11811160064 bytes of memory
i 23:16:28|main GeForce GTX 1080 Ti with 11811160064 bytes of memory
i 23:16:36|main GeForce GTX 1080 Ti with 11811160064 bytes of memory
i 23:16:46|main GeForce GTX 1080 with 8589934592 bytes of memory
... 23:16:52|main Sent block info t=169a31b50d4 res=OK
d00fdf4f2013b2c187d3c6353169e5b7009837d802ce84f4e76594c7125bd519
cl 23:16:52|main Platform: NVIDIA CUDA
cl 23:16:55|main Device: GeForce GTX 1080 Ti / OpenCL 1.2 CUDA
X 23:17:15|main OpenCL init failed: clCreateContext: CL_OUT_OF_RESOURCES (-5)
... 23:17:21|main Failed to initialize mining
Failed to start mining

Answer 1 · 2018-01-29T07:59:37.000Z

Hi, I didn't face such error yet. Let me think what can be a reason.

Answer 2 · 2018-01-29T08:09:22.000Z

jonano614, a couple of more observations. When I increased the number of used devices (Ex, -d 4 or -d 8),

the CPU consumption gets extremely high (looks like the GPU miner is also doing CPU mining). Would it be possible for you to disable CPU mining in the GPU miner?)
the MH/s for each card is going down significantly. Fro ex, when I use only 2 GPUs (-d 2), I get about 1100MH/s for each card. And when I use 3 GPUs (-d 3), I get about 700-800 MH/s. Not sure if this is due to the high CPU utilization or due to something else.

Please let me know if you need any additional info. I noticed that there is no logging, or else I would have sent you the log file. Thanks!

Answer 3 · 2018-01-29T11:09:44.000Z

Hi.

No, GPU-miner does not use CPU-mining. There are some tasks for CPU, but they must not consume much CPU time. It is a bug.
Probably the problem is related to previous, GPU might be waiting when CPU completes some work.

I don't have such problem. Probably because I have only one GPU with fairly low performance (240 MH/s). So it can take some time to find out the reason of the issue.

Answer 4 · 2018-01-29T12:02:09.000Z

Yes， I use 6 gpus, so the miner take 100% cpu.

Answer 5 · 2018-01-29T15:39:37.000Z

jonano614, one other comment. On my 8 GPU mining rig, the CPU I have is an Intel pentium processo, which is less expensive and has low capacity. This may be one of the reason for hitting 99% CPU limit.

I have also one more laptop that has Intel HD Graphics 3000 which has Intel core i5, which is quite fast. In one of the posts I see that you have a miner for Intel. If you can share with me the miner for the Intel graphics card, I can test it on my laptop and see if I observe the same issue.

Thanks!

Answer 6 · 2018-01-29T22:50:07.000Z

Probably Intel HD Graphics 3000 is not supported. I've tested it only on Intel HD Graphics 4600. Try "-list-devices -G"
Regarding error "clCreateContext: CL_OUT_OF_RESOURCES": as I undestand you try to run different instances of miner. I don't think it is a good idea, but probably it can work. But "-d N" parameter always uses first N devices. Try "-opencl-device 0 1 2" and "-opencl-devices 3 4 5".

Answer 7 · 2018-02-02T00:27:38.000Z

Jonano, did you get a chance to look into the CPU resource issue. Any plans to release an updated miner?

Answer 8 · 2018-02-02T01:00:17.000Z

Hi. I tryed to solve the problem, but unfortunately I didn't manage to fix that bug yet.

Answer 9 · 2018-02-02T14:44:20.000Z

I also get the CL_OUT_OF_RESOURCES if I attempt to run the gpu miner on 3+ cards (interestingly, when I was at only 3 cards, I could run it on all 3, but now that I have 4 cards, it fails above 2). Currently, I work around it by having a separate xdag account and directory instance, and in each session, I mine against two of the cards (against separate accounts and storage directories to ensure no write conflicts with shares). However, I'm not sure why I get CL_OUT_OF_RESOURCES if I run against 3 or 4 devices in the same window (using -opencl-devices 0 1 2 or 0 1 2 3 (and I tested various permutations, but that had no effect)). I have an i7 5930k CPU and with two cards, its usage is around 40%, so I'm definitely not tapped out on the CPU. With the 4 cards in two instances, usage is around 50%.

Answer 10 · 2018-02-02T20:06:59.000Z

New version 0.2.0 resolves this for me. 👍

Answer 11 · 2018-02-02T20:09:55.000Z

I'm glad to hear that. Probably one improvement helped.

Answer 12 · 2018-02-12T07:23:26.000Z

Sykoboy. I've released v0.2.3 yesterday. Could you please look at it. Probably it can help you.

Answer 13 · 2018-02-12T17:48:00.000Z

Hi Jonano, thank you. I will switch to the new version and will let you know after I run it for a day.

Answer 14 · 2018-02-14T19:09:18.000Z

Jonano, I have been running the new version 0.2.3 on my rig that has 2 GPUs. I do see this error. I did not get this error on earlier version 0.2.0.

000000002b5782386925b8370ec394a6e60f9107fc3b8b53b208b6f6cf2a4d25
m 12:41:22|main Speed 1951.50 Mh/s u/0 1096.46 u/1 855.04
m 12:41:29|main Speed 1947.18 Mh/s u/0 1102.62 u/1 844.56
... 12:41:35|main Share t=16a120bffff res=OK
000000000d101f56b81c7556466d2a85cd0ec35768d7826b31a47524a27cdeef X 12:41:40|cl-1 OpenCL Error: clEnqueueReadBuffer: CL_OUT_OF_RESOURCES (-5)

X 12:41:40|cl-0 OpenCL Error: clEnqueueReadBuffer: CL_OUT_OF_RESOURCES (-5)
m 12:41:42|main Speed 1369.98 Mh/s u/0 981.93 u/1 388.05
m 12:41:46|main Speed 1369.98 Mh/s u/0 981.93 u/1 388.05
m 12:41:50|main Speed 1369.98 Mh/s u/0 981.93 u/1 388.05

Answer 15 · 2018-02-14T19:37:07.000Z

Ok, I will try to find information what can lead to that error. By the way, since main coin dev left the project it is hard to find time now for miner...

Answer 16 · 2018-02-14T22:26:05.000Z

Jonano, one update on that out of resources error. I think it has to do with loose cable on the 1080 card. I re-plugged all the cables and kicked off the miner and don't see the error. Mostly, the error may be due to loose cable. I will let you know in case I see it agian.

BTW, you have done some good work on this miner. I would like to send you some XDAG that I mined, but its not trading for much. I will send you little in ETH. Please post your ETH address.

Answer 17 · 2018-02-14T22:53:50.000Z

Hi. Thank you 0x2d1044fe7eaf1e1d19eae70878862509f54bba8d

Answer 18 · 2018-02-14T23:49:26.000Z

Jonano, just sent you little ETH. You should be able to see in your account. Thanks!

Answer 19 · 2018-02-15T01:41:56.000Z

Jonano, the .25 ETH I sent got confirmed. But, I noticed something odd. As soon as the funds came to your account, they got moved to some other address. I am not sure whats going on. Something does not look right.

https://etherscan.io/tx/0xc65b6e8ad7887ac9e227e66ee7943f22aa8fb6eb6db64acd2d0b94f87270e8db

Anyways, not sure if we should talk about this in this particular thread. I will ping you on telegram tomorrow.

Answer 20 · 2018-02-15T04:27:47.000Z

Thank you! It is very much!
The address is on exchange, so do not worry.

Answer 21 · 2018-02-15T04:31:20.000Z

Cool, good to hear. Thanks again and keep up the great work.