Ethminer: 6x RX580's with clEnqueueNDRangeKernel(-4) errors

Created on 27 Dec 2017  路  11Comments  路  Source: ethereum-mining/ethminer

Hi,

I've been mining successfully with 4x MSI RX580 Gaming X's and 1x Gigabyte Aorus RX580 with no isues.

Today I upgraded rig to 6x MSI RX580 Gaming X's (removing the Gigabyte card). Now when I run ethminer, I get clEnqueueNDRangeKernel(-4) errors, and ethminer hangs whilst trying to generate DAG's.

Here's my script:
ethminer -F pool/address/miner/email -G -t 6

Here's my output with errors:

Genoil's ethminer 0.9.41-genoil-1.1.9nr
=====================================================================
Forked from github.com/ethereum/cpp-ethereum
CUDA kernel ported from Tim Hughes' OpenCL kernel
With contributions from nicehash, nerdralph, RoBiK and sp_

Please consider a donation to:
ETH: 0xeb9310b185455f863f526dab3d245809f6854b4d

Found suitable OpenCL device [Ellesmere ] with 8589934592 bytes of GPU memory
  m  15:07:40|main  Getting work package...
  m  15:07:41|main  Got work package: #5832dce9
  i  15:07:41|GPU0  Initialising miner...
  i  15:07:41|GPU1  Initialising miner...
  i  15:07:41|GPU2  Initialising miner...
  i  15:07:41|GPU3  Initialising miner...
  i  15:07:41|GPU4  Initialising miner...
  i  15:07:41|GPU5  Initialising miner...
  m  15:07:41|main  Mining on PoWhash #5832dce9 : 0+0+0+0+0+0=0MH/s A0+0:R0+0
  m  15:07:42|main  Got work package: #20bcdd93
Using platform: AMD Accelerated Parallel Processing
Using platform: AMD Accelerated Parallel Processing
Using platform: AMD Accelerated Parallel Processing
Using device: Ellesmere(OpenCL 2.0 AMD-APP (2442.12))
Using platform: AMD Accelerated Parallel Processing
Using device: Ellesmere(OpenCL 2.0 AMD-APP (2442.12))
Using device: Ellesmere(OpenCL 2.0 AMD-APP (2442.12))
Printing program log
Using platform: AMD Accelerated Parallel Processing
Using platform: AMD Accelerated Parallel Processing
Printing program log

Using device: Ellesmere(OpenCL 2.0 AMD-APP (2442.12))
Creating cache buffer
Printing program log
Creating DAG buffer
Loading kernels
Writing cache buffer

Creating cache buffer
Creating DAG buffer

Using device: Ellesmere(OpenCL 2.0 AMD-APP (2442.12))
Creating buffer for header.
Using device: Ellesmere(OpenCL 2.0 AMD-APP (2442.12))
Loading kernels
Creating mining buffer 0
Printing program log
Creating cache buffer
Creating mining buffer 1

Creating DAG buffer
Printing program log
Creating cache buffer

  i  15:12:59|GPU0  2.25GB of DAG data generated in 16439 ms.Writing cache buffer

Printing program log
Creating cache buffer
Creating buffer for header.
Creating mining buffer 0
Creating DAG buffer
Creating DAG buffer
Creating mining buffer 1

Loading kernels
Loading kernels
Writing cache buffer
Creating cache buffer
Creating buffer for header.
Creating DAG buffer
Writing cache buffer
Creating mining buffer 0
Loading kernels
Loading kernels
Creating mining buffer 1
Creating buffer for header.
Writing cache buffer
Writing cache buffer
Creating mining buffer 0
Creating mining buffer 1
Creating buffer for header.
Creating mining buffer 0
Creating mining buffer 1
Creating buffer for header.
Creating mining buffer 0
Creating mining buffer 1
clEnqueueNDRangeKernel(-4)
clEnqueueNDRangeKernel(-4)
  i  15:14:35|GPU2  2.25GB of DAG data generated in 35632 ms.
  i  15:14:43|GPU1  2.25GB of DAG data generated in 13826 ms.
  i  15:14:45|GPU3  2.25GB of DAG data generated in 13585 ms.
  i  15:14:45|GPU5  2.25GB of DAG data generated in 13553 ms.

My GPU's have 8gig each, and I have 4gig of system RAM. Could the 4gig system RAM be the issue, going from 5 cards to 6?

Any assistance would be appreciated

Regards

picture lowlanda

Most helpful comment

Yeah, that should work.
Also try running: ethminer --help if you want to see what other params are available.
If I were you, I'd at least experiment with OpenCL parameters, e.g.
--cl-global-work 16384 --cl-local-work 256 --cl-parallel-hash 2

But you're using Genoil's fork, so not all parameters might work / might be differences in syntax.

picture androidi on 27 Dec 2017
👍2

All 11 comments

If you are using Windows10 and modded bios, try to patch your driver again, reboot the system and re-run ethminer.
I know I have to patch my driver every time I added new card.

efrasyad picture efrasyad on 27 Dec 2017
👍1

If you are using Windows10 and modded bios, try to patch your driver again, reboot the system and re-run ethminer.
I know I have to patch my driver every time I added new card.>

Thanks for your reply efrasyad. Yes I'm using Win10 + modded BIOS. When you say "patch" driver, do you mean reinstall it? I'm using AMD's blockchain driver fyi. Cheers

lowlanda picture lowlanda on 27 Dec 2017

What I meant was run atikmdag-patcher for AMD driver.
I'm also using AMD blockchain Aug driver.

efrasyad picture efrasyad on 27 Dec 2017
👍1

Ok cool, I've never heard of atikmdag-patcher before, I'll try it out. Thanks

lowlanda picture lowlanda on 27 Dec 2017

Before I do, could you explain what this does, in relation to the clEnqueueNDRangeKernel(-4) issue I'm having?

I've never heard or/experienced this signature check issue. Is that what I'm experiencing now?

Sorry for the additional questions, I'm just trying to understand the issue and solution. Thx again

lowlanda picture lowlanda on 27 Dec 2017

Have you tried generating the buffer on just one of the cards, and then copying it to others? There's a switch for that: -L sequential
The description says: "load DAG on GPUs one after another. Use this when the miner crashes during DAG generation"

androidi picture androidi on 27 Dec 2017
👍1

Have you tried generating the buffer on just one of the cards, and then copying it to others? There's a switch for that: -L sequential
The description says: "load DAG on GPUs one after another. Use this when the miner crashes during DAG generation"

No I haven't. So I just put "-L sequential" into my script?

Thanks

lowlanda picture lowlanda on 27 Dec 2017

Yeah, that should work.
Also try running: ethminer --help if you want to see what other params are available.
If I were you, I'd at least experiment with OpenCL parameters, e.g.
--cl-global-work 16384 --cl-local-work 256 --cl-parallel-hash 2

But you're using Genoil's fork, so not all parameters might work / might be differences in syntax.

androidi picture androidi on 27 Dec 2017
👍2

Increase virtual memory as stated in claymore miner, ethminer needs it too.

For multi-GPU systems, set Virtual Memory size in Windows at least 16 GB:
"Computer Properties / Advanced System Settings / Performance / Advanced / Virtual Memory".

ddobreff picture ddobreff on 27 Dec 2017
🎉1

Thanks for everyones input, much appreciated

Update:
I tried the following switches with exact same result. (will try some other things too)

-L sequential
-L single 0
-L single 4 (just to using different GPU)

From --help on -L:

 -L, --dag-load-mode <mode> DAG generation mode.
        parallel    - load DAG on all GPUs at the same time (default)
        sequential  - load DAG on GPUs one after another. Use this when the miner crashes during DAG generation
        single <n>  - generate DAG on device n, then copy to other devices

-L single still produces this:

Genoil's ethminer 0.9.41-genoil-1.1.9nr
=====================================================================
Forked from github.com/ethereum/cpp-ethereum
CUDA kernel ported from Tim Hughes' OpenCL kernel
With contributions from nicehash, nerdralph, RoBiK and sp_

Please consider a donation to:
ETH: 0xeb9310b185455f863f526dab3d245809f6854b4d

Found suitable OpenCL device [Ellesmere ] with 8589934592 bytes of GPU memory
  m  00:09:39|main  Getting work package...
  m  00:09:40|main  Got work package: #3269b5cf
  i  00:09:40|GPU0  Initialising miner...
  i  00:09:40|GPU5  Initialising miner...Using platform: AMD Accelerated Parallel Processing

  i  00:09:40|GPU4  Initialising miner...
Using device: Ellesmere(OpenCL 2.0 AMD-APP (2442.12))
  i  00:09:40|GPU2  Initialising miner...
Using platform: AMD Accelerated Parallel Processing
Using platform: AMD Accelerated Parallel Processing
  i  00:09:40|GPU3  Initialising miner...
Using device: Ellesmere(OpenCL 2.0 AMD-APP (2442.12))
  m  00:09:41|main  Mining on PoWhash #3269b5cf : 0+0+0+0+0+0=0MH/s A0+0:R0+0
Using platform: AMD Accelerated Parallel Processing  i  00:09:40|GPU1  Initialising miner...

Using device: Ellesmere(OpenCL 2.0 AMD-APP (2442.12))  m  00:10:41|main  Got work package: #a807ea3c

Using device: Ellesmere(OpenCL 2.0 AMD-APP (2442.12))
Printing program log
Printing program log
Printing program log

Using platform: AMD Accelerated Parallel Processing

Using platform: AMD Accelerated Parallel Processing
Creating cache buffer
Using device: Ellesmere(OpenCL 2.0 AMD-APP (2442.12))
Using device: Ellesmere(OpenCL 2.0 AMD-APP (2442.12))
Creating DAG buffer
Loading kernels
Writing cache buffer

Printing program log
Printing program log
Printing program log
Creating cache buffer

Creating buffer for header.

Creating cache buffer

Creating cache buffer
Creating cache buffer
Creating DAG buffer
Creating mining buffer 0
Creating cache buffer
Creating DAG buffer
Loading kernels
Creating mining buffer 1
Writing cache buffer
Creating DAG buffer
Loading kernels
Loading kernels
Creating DAG buffer
Loading kernels
Writing cache buffer  i  00:13:04|GPU4  2.25GB of DAG data generated in 16439 ms.

Creating buffer for header.
Writing cache buffer
Creating mining buffer 0
Creating buffer for header.
Writing cache buffer
Creating mining buffer 1
Creating DAG buffer
Creating buffer for header.
Loading kernels
Creating buffer for header.
Writing cache buffer
Creating mining buffer 0
Creating mining buffer 0
Creating mining buffer 0
Creating mining buffer 1
  i  00:13:48|GPU2  2.25GB of DAG data generated in 14907 ms.
Creating buffer for header.
Creating mining buffer 1
Creating mining buffer 1
Creating mining buffer 0
Creating mining buffer 1
clEnqueueNDRangeKernel(-4)
clEnqueueNDRangeKernel(-4)
  i  00:14:03|GPU3  2.25GB of DAG data generated in 13739 ms.
  i  00:14:06|GPU0  2.25GB of DAG data generated in 13581 ms.
  i  00:14:06|GPU5  2.25GB of DAG data generated in 13565 ms.
lowlanda picture lowlanda on 27 Dec 2017

Increase virtual memory as stated in claymore miner, ethminer needs it too.

For multi-GPU systems, set Virtual Memory size in Windows at least 16 GB:
"Computer Properties / Advanced System Settings / Performance / Advanced / Virtual Memory".

Winner, winner, chicken dinner! Virtual Memory was the solution in this case. big thank you ddobreff.

Also thanks to androidi and efrasyad, all very interesting and useful information.

Have a great new year everyone.

lowlanda picture lowlanda on 27 Dec 2017
Was this page helpful?
0 / 5 - 0 ratings

Related issues

nguylb73 picture nguylb73  路  3Comments
rawsh picture rawsh  路  5Comments
RJBetancor picture RJBetancor  路  5Comments
skynet picture skynet  路  4Comments
phalexo picture phalexo  路  5Comments