GPU 6: out of memory

[Romario0812]

Hello I have computer with: motherboard: Asus Prime Z270-P GPU: 8 х Nvidia 1080Ti at start of the "xmr-stak-nvidia-notls.exe" program an error message is given:

---

GPU 6: out of memory C:/Users/MAIN/Documents/Visual Studio 2015/Projects/xmr-stak-nvidia/xmr-stak-nvidia/nvcc_code/cuda_extra.cu line 193

---

number GPU constantly different (GPU3, GPU6, GPU7...)

config.txt "gpu_threads_conf" : [ { "index" : 0, "threads" : 32, "blocks" : 84, "bfactor" : 6, "bsleep" : 25, "affine_to_cpu" : false, }, { "index" : 1, "threads" : 32, "blocks" : 84, "bfactor" : 6, "bsleep" : 25, "affine_to_cpu" : false, }, { "index" : 2, "threads" : 32, "blocks" : 84, "bfactor" : 6, "bsleep" : 25, "affine_to_cpu" : false, }, { "index" : 3, "threads" : 32, "blocks" : 84, "bfactor" : 6, "bsleep" : 25, "affine_to_cpu" : false, }, { "index" : 4, "threads" : 32, "blocks" : 84, "bfactor" : 6, "bsleep" : 25, "affine_to_cpu" : false, }, { "index" : 5, "threads" : 32, "blocks" : 84, "bfactor" : 6, "bsleep" : 25, "affine_to_cpu" : false, }, { "index" : 6, "threads" : 32, "blocks" : 84, "bfactor" : 6, "bsleep" : 25, "affine_to_cpu" : false, }, { "index" : 7, "threads" : 32, "blocks" : 84, "bfactor" : 6, "bsleep" : 25, "affine_to_cpu" : false, }, ],

tell me please, in what the reason of this problem

[Jordonhalpin]

Lower "blocks" to 28 for each card.

[Romario0812]

I have replaced as you have specified: "gpu_threads_conf" : [ { "index" : 0, "threads" : 32, "blocks" : 28, "bfactor" : 6, "bsleep" : 25, "affine_to_cpu" : false, }, { "index" : 1, "threads" : 32, "blocks" : 28, "bfactor" : 6, "bsleep" : 25, "affine_to_cpu" : false, }, { "index" : 2, "threads" : 32, "blocks" : 28, "bfactor" : 6, "bsleep" : 25, "affine_to_cpu" : false, }, { "index" : 3, "threads" : 32, "blocks" : 28, "bfactor" : 6, "bsleep" : 25, "affine_to_cpu" : false, }, { "index" : 4, "threads" : 32, "blocks" : 28, "bfactor" : 6, "bsleep" : 25, "affine_to_cpu" : false, }, { "index" : 5, "threads" : 32, "blocks" : 28, "bfactor" : 6, "bsleep" : 25, "affine_to_cpu" : false, }, { "index" : 6, "threads" : 32, "blocks" : 28, "bfactor" : 6, "bsleep" : 25, "affine_to_cpu" : false, }, { "index" : 7, "threads" : 32, "blocks" : 28, "bfactor" : 6, "bsleep" : 25, "affine_to_cpu" : false, }, ],

The problem also remained: GPU 1: out of memory C:/Users/MAIN/Documents/Visual Studio 2015/Projects/xmr-stak-nvidia/xmr-stak-nvidia/nvcc_code/cuda_extra.cu line 193

GPU 7: out of memory GPU 5: out of memory

[Romario0812]

In addition I report characteristics of the computer: memory 4Gb Windows 10 swap file in Windows: automatically Windows is chosen

[Romario0812]

Windows 10 x64

[Romario0812]

In addition I report characteristics of the computer videocard: 8 х Gigabyte GeForce® GTX 1080 Ti Gaming OC 11G 11 Gb memory

[Jordonhalpin]

I had the same error and so reduced block number until error went away. But I only have 7 cards. Perhaps 8 cards required a slightly different setting. Try pulling one card to see if it helps. On Sep 4, 2017 4:26 AM, "Romario0812" notifications@github.com wrote:

In addition I report characteristics of the computer videocard: 8 х Gigabyte GeForce® GTX 1080 Ti Gaming OC 11G 11 Gb memory

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/fireice-uk/xmr-stak-nvidia/issues/119#issuecomment-326927309, or mute the thread https://github.com/notifications/unsubscribe-auth/AeAdB0DMQKHItX0luSD3dL9mhX0TerHdks5se9BggaJpZM4PLas7 .

[Jordonhalpin]

Have you disabled the IGP?

[Romario0812]

what is IGP where can he be disabled?

[Romario0812]

what is IGP

sorry undestand Integrated Graphics is disconnected from the very beginning.

[psychocrypt]

I am not shure but could you add more main memory to your PC. 4GB with 8 GPUs can be to less. You could also try to reduce the thread count for the first GPU, maybe windows is using to much memory.

[Romario0812]

thanks for the hint I have select swap file windows on manual in the size 20Gb and everything has earned ("blocks" : 28,)

I have tried to deliver to "blocks": 84, - a mistake "GPU 7: out of memory" has again appeared I have now started on "blocks": 28, - everything works Tell me please, reduction of the "blocks" parameter to 28, has strongly reduced work productivity? Thanks

[Jordonhalpin]

Fwiw I have used 84 and 28 blocks and it doesn't make any difference to hash rate for my 1080ti. Using 28 makes it far more stable though. On Sep 4, 2017 7:03 PM, "Romario0812" notifications@github.com wrote:

thanks for the hint I have select swap file windows on manual in the size 20Gb and everything has earned ("blocks" : 28,)

I have tried to deliver to "blocks": 84, - a mistake "GPU 7: out of memory" has again appeared I have now started on "blocks": 28, - everything works Tell me please, reduction of the "blocks" parameter to 28, has strongly reduced work productivity? Thanks

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/fireice-uk/xmr-stak-nvidia/issues/119#issuecomment-327045703, or mute the thread https://github.com/notifications/unsubscribe-auth/AeAdB9L9t1EeiehclcxGvmOGbR0lMr90ks5sfJ3lgaJpZM4PLas7 .

[epb5119]

Jordonhalpin what virtual memory did you set to maintain the 7 cards.

[Jordonhalpin]

Default settings, win10, for virtual memory. The memory I believe I was running out of when it errored out was video memory although each card has 6GB. I suspect it wasnt so much the amount but how it was being addressed. Playing with number of blocks and threads in the config sorted that out for me.

On Dec 7, 2017 9:08 PM, "epb5119" notifications@github.com wrote:

Jordonhalpin what virtual memory did you set to maintain the 7 cards.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/fireice-uk/xmr-stak-nvidia/issues/119#issuecomment-350167052, or mute the thread https://github.com/notifications/unsubscribe-auth/AeAdB_mQdd-FRHQuS-kU67HIPvvsJWvQks5s-LZHgaJpZM4PLas7 .