Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Miner not using AMD 6990 GPU #472

Open
m03e5 opened this issue Dec 11, 2017 · 221 comments
Open

Miner not using AMD 6990 GPU #472

m03e5 opened this issue Dec 11, 2017 · 221 comments
Assignees

Comments

@m03e5
Copy link

m03e5 commented Dec 11, 2017

Please provide as much as possible information to reproduce the issue.

Basic information

  • Xeon X5660
  • AMD 6990
  • Win 10 Pro X64

Do you compiled the miner by our own?
-no

Stability issue

  • Is the CPU or GPU overclocked? / No overclocking
  • Is the Main memory of the CPU or GPU undervolted? / No

When initially executed miner creates 2 config files, one for cpu and one for gpu. However when miner is running only cpu is mining, no AMD gpu visible and no errors, acts just like its not there. Its a dual GPU on one physical video card and i know it works as it was mining under nicehash, both gpu's are listed in device manager and obviously they are seen by miner as it creates config file for it...
Here is config, generated during initial run of program.

/*

  • GPU configuration. You should play around with intensity and worksize as the fastest settings will vary.
  •  index    - GPU index number usually starts from 0
    
  • intensity - Number of parallel GPU threads (nothing to do with CPU threads)
  • worksize - Number of local GPU threads (nothing to do with CPU threads)
  • affine_to_cpu - This will affine the thread to a CPU. This can make a GPU miner play along nicer with a CPU miner.
  • strided_index - switch memory pattern used for the scratch pad memory
  •             true  = use 16byte contiguous memory per thread, the next memory block has offset of intensity blocks
    
  •             false = use a contiguous block of memory per thread
    
  • "gpu_threads_conf" :
  • [
  • { "index" : 0, "intensity" : 1000, "worksize" : 8, "affine_to_cpu" : false, "strided_index" : true },
  • ],
    */

"gpu_threads_conf" : [
// gpu: Cayman memory:384
// compute units: 24
{ "index" : 0,
"intensity" : 0, "worksize" : 8,
"affine_to_cpu" : false, "strided_index" : true
},
// gpu: Cayman memory:384
// compute units: 24
{ "index" : 1,
"intensity" : 0, "worksize" : 8,
"affine_to_cpu" : false, "strided_index" : true
},

],

/*

  • Platform index. This will be 0 unless you have different OpenCL platform - eg. AMD and Intel.
    */
    "platform_index" : 0,
    ===============================================================
    What am i doing wrong or not doing right?
@m03e5
Copy link
Author

m03e5 commented Dec 11, 2017

Off top question - is there a way to disable device(cpu) in mining, i have another computer with core 2 duo, that is used for mining when miner running and its just useless... i have 3 video cards on that board that do work well though. I want to use only cards and not use cpu at all. Tried deleting config and its being re-created, tried modding config to a point that it cant be used and miner says invalid config and just quits, so video cards cant mine, any suggestions?

@psychocrypt
Copy link
Collaborator

psychocrypt commented Dec 11, 2017 via email

@psychocrypt
Copy link
Collaborator

psychocrypt commented Dec 11, 2017 via email

@bovcan
Copy link

bovcan commented Dec 11, 2017

try to change "platform_index" : 0, to 1 or 2.

@Simaex
Copy link

Simaex commented Dec 11, 2017

What driver are you using for HD6990? I have many issues with the card before I roll back to 15.7 driver but I never think that XMR-STAK can work with such old driver. I'll try to run 2.1 version of XMR-STAK and report results.

@m03e5
Copy link
Author

m03e5 commented Dec 11, 2017

xmr amd driver screen


xmr-stak 2.1.0 30888a3

Brought to you by fireice_uk and psychocrypt under GPLv3.
Based on CPU mining code by wolf9466 (heavily optimized by fireice_uk).
Based on NVIDIA mining code by KlausT and psychocrypt.
Based on OpenCL mining code by wolf9466.

Configurable dev donation level is set to 2.0%

You can use following keys to display reports:
'h' - hashrate
'r' - results
'c' - connection

[2017-12-11 09:37:52] : Start mining: MONERO
[2017-12-11 09:37:52] : Starting NVIDIA GPU thread 0, no affinity.
[2017-12-11 09:37:52] : Compiling code and initializing GPUs. This will take a while...
[2017-12-11 09:37:54] : Device 0 work size 8 / 32.
[2017-12-11 09:37:54] : Error CL_INVALID_BUFFER_SIZE when calling clCreateBuffer to create hash scratchpads buffer.
[2017-12-11 09:37:54] : WARNING: AMD device not found
[2017-12-11 09:37:54] : WARNING: backend AMD disabled.
[2017-12-11 09:37:54] : Starting 1x thread, affinity: 0.
[2017-12-11 09:37:54] : hwloc: memory pinned
[2017-12-11 09:37:54] : Starting 1x thread, affinity: 2.
[2017-12-11 09:37:54] : hwloc: memory pinned
[2017-12-11 09:37:54] : Starting 1x thread, affinity: 4.
[2017-12-11 09:37:54] : hwloc: memory pinned
[2017-12-11 09:37:54] : Starting 1x thread, affinity: 6.
[2017-12-11 09:37:54] : hwloc: memory pinned
[2017-12-11 09:37:54] : Starting 1x thread, affinity: 8.
[2017-12-11 09:37:54] : hwloc: memory pinned
[2017-12-11 09:37:54] : Starting 1x thread, affinity: 10.
[2017-12-11 09:37:54] : hwloc: memory pinned
[2017-12-11 09:37:54] : Starting 1x thread, affinity: 12.
[2017-12-11 09:37:54] : hwloc: memory pinned
[2017-12-11 09:37:54] : Starting 1x thread, affinity: 14.
[2017-12-11 09:37:54] : hwloc: memory pinned
[2017-12-11 09:37:54] : Starting 1x thread, affinity: 16.
[2017-12-11 09:37:54] : hwloc: memory pinned
[2017-12-11 09:37:54] : Starting 1x thread, affinity: 18.
[2017-12-11 09:37:54] : hwloc: memory pinned
[2017-12-11 09:37:54] : Starting 1x thread, affinity: 20.
[2017-12-11 09:37:54] : hwloc: memory pinned
[2017-12-11 09:37:54] : Starting 1x thread, affinity: 22.
[2017-12-11 09:37:54] : hwloc: memory pinned
[2017-12-11 09:37:54] : Fast-connecting to ca.minexmr.com:3333 pool ...
[2017-12-11 09:37:54] : Pool ca.minexmr.com:3333 connected. Logging in...
[2017-12-11 09:37:54] : Difficulty changed. Now: 2000450.
[2017-12-11 09:37:54] : Pool logged in.
[2017-12-11 09:38:14] : Difficulty changed. Now: 2000450.
[2017-12-11 09:38:16] : New block detected.
HASHRATE REPORT - CPU
| ID | 10s | 60s | 15m | ID | 10s | 60s | 15m |
| 0 | 40.0 | 39.9 | (na) | 1 | 40.0 | 39.9 | (na) |
| 2 | 40.0 | 39.9 | (na) | 3 | 40.1 | 40.0 | (na) |
| 4 | 41.9 | 41.7 | (na) | 5 | 41.7 | 41.6 | (na) |
| 6 | 40.0 | 38.7 | (na) | 7 | 39.9 | 38.3 | (na) |
| 8 | 39.8 | 38.7 | (na) | 9 | 39.9 | 38.8 | (na) |
| 10 | 41.8 | 40.5 | (na) | 11 | 41.5 | 40.2 | (na) |

HASHRATE REPORT - NVIDIA
| ID | 10s | 60s | 15m |
| 0 | (na) | 25.5 | (na) |

Totals: (na) 503.8 (na) H/s
Highest: 0.0 H/s

6990 is very good card for xmr, it did 770hs which is same as 1080ti OC does on stock clocks.

@m03e5
Copy link
Author

m03e5 commented Dec 11, 2017

I have not yet tried other suggestions, due to latest win10 update messing up system, had to undo update...took a while. Will do it a bit later. AMD driver is on screenshot and it is indeed 15.7.1 WHQL

  • Rebooting system, deleting all configs and re-run of xmr stack caused creation of new config files that are identical to the original ones posted above. PC has 100+GB of ram...no other programs/tasks are running when xmr stack runs. Screen attached to GT730.
    Thanks

@Simaex
Copy link

Simaex commented Dec 11, 2017

I've managed to get xmr-stak running. I changed intensity to non-zero value and then try to set bigger values. Maximum value that I was able to reach was 250. Overall performance is 3 times lower then using old 9,7 Claymore miner. I beleive that problem is with incorrect memory detection. Only 384 M per GPU are detected when 2G are available. Low hashrate of GPU6 corresponds to performance in other miners. Please advise how to proceed with optimisation. Please note that we have some sort of bug if intensity is set to zero, we get error not stopped adapter.

> xmr-stak 2.1.0 30888a3

Brought to you by fireice_uk and psychocrypt under GPLv3.
Based on CPU mining code by wolf9466 (heavily optimized by fireice_uk).
Based on NVIDIA mining code by KlausT and psychocrypt.
Based on OpenCL mining code by wolf9466.

Configurable dev donation level is set to 2.0%

You can use following keys to display reports:
'h' - hashrate
'r' - results
'c' - connection
-------------------------------------------------------------------
[2017-12-11 21:04:43] : Start mining: MONERO
WARNING: NVIDIA Insufficient driver!
[2017-12-11 21:04:43] : WARNING: backend NVIDIA disabled.
[2017-12-11 21:04:43] : Compiling code and initializing GPUs. This will take a while...
[2017-12-11 21:04:45] : Device 0 work size 8 / 32.
[2017-12-11 21:04:49] : Device 1 work size 8 / 32.
[2017-12-11 21:04:54] : Device 2 work size 8 / 32.
[2017-12-11 21:04:58] : Device 3 work size 8 / 32.
[2017-12-11 21:05:02] : Device 4 work size 8 / 32.
[2017-12-11 21:05:07] : Device 5 work size 8 / 32.
[2017-12-11 21:05:11] : Device 6 work size 8 / 32.
[2017-12-11 21:05:16] : Device 7 work size 8 / 32.
[2017-12-11 21:05:20] : Starting AMD GPU thread 0, no affinity.
[2017-12-11 21:05:20] : Starting AMD GPU thread 1, no affinity.
[2017-12-11 21:05:20] : Starting AMD GPU thread 2, no affinity.
[2017-12-11 21:05:20] : Starting AMD GPU thread 3, no affinity.
[2017-12-11 21:05:20] : Starting AMD GPU thread 4, no affinity.
[2017-12-11 21:05:20] : Starting AMD GPU thread 5, no affinity.
[2017-12-11 21:05:20] : Starting AMD GPU thread 6, no affinity.
[2017-12-11 21:05:20] : Starting AMD GPU thread 7, no affinity.
[2017-12-11 21:05:20] : Starting 1x thread, affinity: 0.
[2017-12-11 21:05:20] : hwloc: memory pinned
[2017-12-11 21:05:20] : Fast-connecting to xmr-eu1.nanopool.org:14433 pool ...
[2017-12-11 21:05:21] : Fast-connecting to xmr-eu2.nanopool.org:14433 pool ...
[2017-12-11 21:05:21] : TLS fingerprint [xmr-eu1.nanopool.org:14433] SHA256:w4iG797lQuvZmAG3XHXTSY2Xl4u83sB8cnHLGXKeAU8=
[2017-12-11 21:05:21] : Pool xmr-eu1.nanopool.org:14433 connected. Logging in...
[2017-12-11 21:05:21] : TLS fingerprint [xmr-eu2.nanopool.org:14433] SHA256:w4iG797lQuvZmAG3XHXTSY2Xl4u83sB8cnHLGXKeAU8=
[2017-12-11 21:05:21] : Pool xmr-eu2.nanopool.org:14433 connected. Logging in...
[2017-12-11 21:05:23] : Difficulty changed. Now: 120001.
[2017-12-11 21:05:23] : Pool logged in.
[2017-12-11 21:05:23] : Difficulty changed. Now: 120001.
[2017-12-11 21:05:23] : New block detected.
HASHRATE REPORT - CPU
| ID |    10s |    60s |    15m |
|  0 |   (na) |   (na) |   (na) |
---------------------------
HASHRATE REPORT - AMD
| ID |    10s |    60s |    15m | ID |    10s |    60s |    15m |
|  0 |   (na) |   (na) |   (na) |  1 |   (na) |   (na) |   (na) |
|  2 |   (na) |   (na) |   (na) |  3 |   (na) |   (na) |   (na) |
|  4 |   (na) |   (na) |   (na) |  5 |   (na) |   (na) |   (na) |
|  6 |   (na) |   (na) |   (na) |  7 |   (na) |   (na) |   (na) |
-----------------------------------------------------
Totals:     (na)   (na)   (na) H/s
Highest:     0.0 H/s
HASHRATE REPORT - CPU
| ID |    10s |    60s |    15m |
|  0 |   45.1 |   (na) |   (na) |
---------------------------
HASHRATE REPORT - AMD
| ID |    10s |    60s |    15m | ID |    10s |    60s |    15m |
|  0 |  130.9 |   (na) |   (na) |  1 |  135.2 |   (na) |   (na) |
|  2 |  142.9 |   (na) |   (na) |  3 |  147.2 |   (na) |   (na) |
|  4 |  153.7 |   (na) |   (na) |  5 |  159.4 |   (na) |   (na) |
|  6 |   97.1 |   (na) |   (na) |  7 |  158.1 |   (na) |   (na) |
-----------------------------------------------------
Totals:   1169.6   (na)   (na) H/s
Highest:  1170.0 H/s
__________________________________________________________________________
AMD.TXT 

> 
"gpu_threads_conf" : [
  // gpu: Cayman memory:384
  // compute units: 24
  { "index" : 0,
    "intensity" : 200, "worksize" : 8,
    "affine_to_cpu" : false, "strided_index" : true
  },
  // gpu: Cayman memory:384
  // compute units: 24
  { "index" : 1,
    "intensity" : 210, "worksize" : 8,
    "affine_to_cpu" : false, "strided_index" : true
  },
  // gpu: Cayman memory:384
  // compute units: 24
  { "index" : 2,
    "intensity" : 220, "worksize" : 8,
    "affine_to_cpu" : false, "strided_index" : true
  },
  // gpu: Cayman memory:384
  // compute units: 24
  { "index" : 3,
    "intensity" : 230, "worksize" : 8,
    "affine_to_cpu" : false, "strided_index" : true
  },
  // gpu: Cayman memory:384
  // compute units: 24
  { "index" : 4,
    "intensity" : 240, "worksize" : 8,
    "affine_to_cpu" : false, "strided_index" : true
  },
  // gpu: Cayman memory:384
  // compute units: 24
  { "index" : 5,
    "intensity" : 250, "worksize" : 8,
    "affine_to_cpu" : false, "strided_index" : true
  },
  // gpu: Cayman memory:384
  // compute units: 24
  { "index" : 6,
    "intensity" : 250, "worksize" : 8,
    "affine_to_cpu" : false, "strided_index" : true
  },
  // gpu: Cayman memory:384
  // compute units: 24
  { "index" : 7,
    "intensity" : 250, "worksize" : 8,
    "affine_to_cpu" : false, "strided_index" : true
  },

],

/*
 * Platform index. This will be 0 unless you have different OpenCL platform - eg. AMD and Intel.
 */
"platform_index" : 1,

@m03e5
Copy link
Author

m03e5 commented Dec 11, 2017

So i tried various intensity and could not go over "250" anything beyond this renders miner not seeing AMD GU. With intensity value 250 card gives about half of what it should be able to do. Any other recommendations are welcome. Also, does anyone know what does that little switch do on 6990? (near CROSSFIRE )


xmr-stak 2.1.0 30888a3

Brought to you by fireice_uk and psychocrypt under GPLv3.
Based on CPU mining code by wolf9466 (heavily optimized by fireice_uk).
Based on NVIDIA mining code by KlausT and psychocrypt.
Based on OpenCL mining code by wolf9466.

Configurable dev donation level is set to 2.0%

You can use following keys to display reports:
'h' - hashrate
'r' - results
'c' - connection

[2017-12-11 15:48:57] : Start mining: MONERO
[2017-12-11 15:48:57] : Starting NVIDIA GPU thread 0, no affinity.
[2017-12-11 15:48:57] : Compiling code and initializing GPUs. This will take a while...
[2017-12-11 15:48:59] : Device 0 work size 8 / 32.
[2017-12-11 15:49:04] : Device 1 work size 8 / 32.
[2017-12-11 15:49:09] : Starting AMD GPU thread 0, no affinity.
[2017-12-11 15:49:09] : Starting AMD GPU thread 1, no affinity.
[2017-12-11 15:49:09] : Starting 1x thread, affinity: 0.
[2017-12-11 15:49:09] : hwloc: memory pinned
[2017-12-11 15:49:09] : Starting 1x thread, affinity: 2.
[2017-12-11 15:49:09] : hwloc: memory pinned
[2017-12-11 15:49:09] : Starting 1x thread, affinity: 4.
[2017-12-11 15:49:09] : hwloc: memory pinned
[2017-12-11 15:49:09] : Starting 1x thread, affinity: 6.
[2017-12-11 15:49:09] : hwloc: memory pinned
[2017-12-11 15:49:09] : Starting 1x thread, affinity: 8.
[2017-12-11 15:49:09] : hwloc: memory pinned
[2017-12-11 15:49:09] : Starting 1x thread, affinity: 10.
[2017-12-11 15:49:09] : hwloc: memory pinned
[2017-12-11 15:49:09] : Starting 1x thread, affinity: 12.
[2017-12-11 15:49:09] : hwloc: memory pinned
[2017-12-11 15:49:09] : Starting 1x thread, affinity: 14.
[2017-12-11 15:49:09] : hwloc: memory pinned
[2017-12-11 15:49:09] : Starting 1x thread, affinity: 16.
[2017-12-11 15:49:09] : hwloc: memory pinned
[2017-12-11 15:49:09] : Starting 1x thread, affinity: 18.
[2017-12-11 15:49:09] : hwloc: memory pinned
[2017-12-11 15:49:09] : Starting 1x thread, affinity: 20.
[2017-12-11 15:49:09] : hwloc: memory pinned
[2017-12-11 15:49:09] : Starting 1x thread, affinity: 22.
[2017-12-11 15:49:09] : hwloc: memory pinned
[2017-12-11 15:49:09] : Fast-connecting to ca.minexmr.com:3333 pool ...
[2017-12-11 15:49:09] : Pool ca.minexmr.com:3333 connected. Logging in...
[2017-12-11 15:49:10] : Difficulty changed. Now: 2000450.
[2017-12-11 15:49:10] : Pool logged in.
[2017-12-11 15:50:33] : New block detected.
HASHRATE REPORT - CPU
| ID | 10s | 60s | 15m | ID | 10s | 60s | 15m |
| 0 | 31.3 | 31.8 | (na) | 1 | 37.3 | 37.1 | (na) |
| 2 | 37.4 | 37.0 | (na) | 3 | 37.7 | 37.5 | (na) |
| 4 | 39.5 | 39.2 | (na) | 5 | 36.0 | 36.3 | (na) |
| 6 | 39.8 | 39.3 | (na) | 7 | 39.5 | 39.4 | (na) |
| 8 | 39.8 | 39.6 | (na) | 9 | 39.5 | 39.5 | (na) |
| 10 | 41.3 | 41.2 | (na) | 11 | 41.2 | 41.2 | (na) |

HASHRATE REPORT - AMD
| ID | 10s | 60s | 15m | ID | 10s | 60s | 15m |
| 0 | 150.2 | 150.1 | (na) | 1 | 150.4 | 150.5 | (na) |

HASHRATE REPORT - NVIDIA
| ID | 10s | 60s | 15m |
| 0 | (na) | 25.5 | (na) |

Totals: (na) 785.2 (na) H/s
Highest: 0.0 H/s

@m03e5
Copy link
Author

m03e5 commented Dec 12, 2017

I got switch figured:
amd 6990 switch

@m03e5
Copy link
Author

m03e5 commented Dec 12, 2017

while its very convenient to have a cpu+ ati+ nvda miners in one, having cards underperform significantly, it was decided to switch 6990 to claymore 9.7, now card does what it should hash - 766hs. I would be really happy to use xmr stack all in one if this problem would of have been fixed. Greets!

        Claymore CryptoNote GPU Miner  v9.7 Beta            ║

╚════════════════════════════════════════════════════════════════╝
XMR: 1 pool is specified
Main pool is ca.minexmr.com:3333

Cards available: 2
GPU #0: name: Cayman, 2048 MB available, 24 units
GPU #1: name: Cayman, 2048 MB available, 24 units
Total cards: 2

Catalyst 15.x is REQUIRED for best performance and compatibility
At least 16 GB of Virtual Memory is required for multi-GPU systems
Make sure you defined GPU_MAX_ALLOC_PERCENT 100
Be careful with overclocking, use default clocks for first tests
Initializing...

GPU #0: HashCnt = 768
GPU #1: HashCnt = 768
POOL version
GPU #0: -a 3 mode selected
GPU #1: -a 3 mode selected
1 pool specified.
Remote management (READ-ONLY MODE) is enabled on port 3333
Press "m" key for tune mode, "s" for current statistics, "0".."9" to turn on/off cards, "r" to reload pools
Stratum - connecting to 'ca.minexmr.com' <158.69.xxx..xxx> port 3333 (unsecure)
DevFee: Stratum - connecting to 'xmr.suprnova.cc' <145.239.xxx.xxx> port 5222 (unsecure)
Stratum - Connected (unsecure)
Stratum - Connected (unsecure)
Pool Diff 200008
DevFee: Pool Diff 10000
DevFee: 12/12/17-10:39:15 - SHARE FOUND (target 5000) - (GPU 1 of 2)
Speed: 766 h/s, Total Shares: 0, Rejected: 0, Mining time: 00:01
12/12/17-10:39:31 - SHARE FOUND (target 200007) - (GPU 1 of 2)
Share accepted (140 ms)!
12/12/17-10:39:38 - New job received from ca.minexmr.com:3333
Speed: 766 h/s, Total Shares: 1, Rejected: 0, Mining time: 00:01

@psychocrypt
Copy link
Collaborator

psychocrypt commented Dec 12, 2017 via email

@Simaex
Copy link

Simaex commented Dec 12, 2017

I remove amd.txt several times, no change. No other apps use video adapters, this is a dedicated rig. Fresh reboot do not change anything. Claymore 9.7 is running on the system 385 H/s per GPU, 3040 H/s for 4 cards (8 GPU). Should I manually set different memory size in config? What size should I try (each GPU have 2G dedicated RAM) if it is reasonable to do so?

@m03e5
Copy link
Author

m03e5 commented Dec 12, 2017

@Simaex - thought you have claymore 9.7 doing normally.... as mine does 766 stable(OC switch pos 1) with it.
P.S. i ahve looked at the logs from claymore9.7 from nicehash and here some settings from it:

15:59:21:863 169c ÉÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍ»
15:59:21:863 169c º Claymore CryptoNote GPU Miner v9.7 Beta º
15:59:21:878 169c ÈÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍÍͼ
15:59:22:097 169c XMR: 1 pool is specified
15:59:22:097 169c Main pool is cryptonight.usa.nicehash.com:33355
Cards available: 2
15:59:22:972 169c GPU #1: name: Cayman, 2048 MB available, 24 units
15:59:22:972 169c Total cards: 1
15:59:22:988 169c Initializing...
15:59:22:988 169c card 0, m1 2048, m2 512 w 192
15:59:22:988 169c 0 m = 2
15:59:22:988 169c GPU #0: HashCnt = 768
15:59:22:988 169c POOL version
15:59:23:003 169c b155
15:59:24:066 169c GPU #0: -a 3 mode selected
15:59:37:629 169c 1 pool specified.

@Simaex
Copy link

Simaex commented Dec 12, 2017

@m03e5 I have CCNM 9.7 running, 785-807 H/s per card but it did not help much in tuning XMR-STAK :-)

@psychocrypt
Copy link
Collaborator

@Simaex Could you please run clinfo that I have a overview of your gpus and the memory which is reported by opencl.

@psychocrypt
Copy link
Collaborator

Which app sdk do you have installed. I compiled the binary with app sdk 3 maybe there is something incompatible with app sdk 2.X

@m03e5
Copy link
Author

m03e5 commented Dec 12, 2017

@psychocrypt while it could matter how miner is compiled on @Simaex system, if he did it himself, but i am using already compiled (by Dev's) version for win. I would assume original compilation would be properly assembled to work on any system. looking forward to see the outcome. Thx

p.s. it seems that we all agree on one - it is VRAM related problem.

@Simaex
Copy link

Simaex commented Dec 13, 2017

I try to use pre-compiled windows version. Please find clinfo output. I run it when Claymore 9.7 was working.
HD6990.txt

@Simaex
Copy link

Simaex commented Dec 22, 2017

No chances to get HD6990 working I assume :-(

@m03e5
Copy link
Author

m03e5 commented Dec 22, 2017 via email

@elizeche
Copy link

Hi friends, I am new here ,but have almost same problem. Using Clarymore, and it keeps showing 0 h/s, no matter what I do. Maybe some step wrong, since I am new at mining world. Could someone help?Thanks!

error gpu

@dougvj
Copy link
Contributor

dougvj commented Dec 30, 2017

I have the same problem on Arch Linux running the proprietary driver with failing to detect memory of the GPU (HD 6990). Setting intensity manually solves the problem (although I am not getting what the hashrate should be for this card, perhaps an unrelated issue as this card has just been repaired)

Attached is a clinfo dump and the autogenned amd.txt
amd.txt
clinfo.txt

@Simaex
Copy link

Simaex commented Dec 30, 2017

@elizeche your problem is that you are trying to use miner that do not support your card. Use 9.7 version and please look for appropriate thread to ask your questions.
@dougvj what hashrate do you get for HD6990?

@dougvj
Copy link
Contributor

dougvj commented Dec 30, 2017

@Simaex I get about 260H/s per GPU so about 520H/s for the entire card. Benchmarks indicate that this card can exceed 700H/s. Could merely be an optimization problem since none of the benchmarks I see use xmr-stak or it could be a problem with my particular card. I will spend more time investigating it this weekend trying different miners looking into throttling and so on, I just wanted to report that I see the same behavior of xmr-stak autosetting an intensity of 0 here.

@Simaex
Copy link

Simaex commented Dec 30, 2017

@dougvj I have four HD6990 - the same for each, 770 H/s for card with Claymore 9.7 but not more then 500 with xmr-stak when manually setting intensity.

dougvj added a commit to dougvj/xmr-stak that referenced this issue Dec 31, 2017
Occassionally the auto adjust doesn't find enough memory and the
intensity is detected too low and aligned to 0 with the compute units.
This patch fixes this situation by issuing a warning with a suggestion
to set environment vars and then ignoring the alignment to 0

Per several issues:

Principally:
fireice-uk#81

Related:
fireice-uk#490
fireice-uk#472
dougvj added a commit to dougvj/xmr-stak that referenced this issue Dec 31, 2017
Occassionally the auto adjust doesn't find enough memory and the
intensity is detected too low and aligned to 0 with the compute units.
This patch fixes this situation by issuing a warning with a suggestion
to set environment vars and then ignoring the alignment to 0

Per several issues:

Principally:
fireice-uk#81

Related:
fireice-uk#490
fireice-uk#472
@m03e5
Copy link
Author

m03e5 commented Apr 25, 2018

@Russia102 @Simaex
I was able to get my 6990 to 768hs only by getting it to work with this:

"gpu_threads_conf" : [
// gpu: Cayman memory:1663
// compute units: 24
{ "index" : 0,
"intensity" : 880, "worksize" : 16,
"affine_to_cpu" : false, "strided_index" : 2, "mem_chunk" : 2,
"comp_mode" : true
},
// gpu: Cayman memory:1663
// compute units: 24
{ "index" : 1,
"intensity" : 880, "worksize" : 16,
"affine_to_cpu" : false, "strided_index" : 2, "mem_chunk" : 2,
"comp_mode" : true
},

Any intensity above would cause problems any intensity below would cause hash rate to drop, changing worksize to 8 - lowers hashrate, slightly differs if intensity changed along but overall still drops. Memchunk setting below 2 or above, up to 16, drops hash rate. GPU switch in position 1 (slightly OC) So my sweet spot is with this config (768hs - just like it did before fork). Runs for few hours so far, no errors. Wondering how you guys achieved 800hs...
14.4_x64 driver, 2017 c++ redist_x64, win10 home_x64, 32gb swap/104gb ram, Miner - 2.4.3/Monero7

@Simaex
Copy link

Simaex commented Apr 25, 2018

@m03e5 If you care to read the thread you will easily find that we used different worksizes from 17 to 21.
For me usually 19 work the best, but if you can get it stable 21 may give better hashrate.
Win HOME edition often develop problems for mining, so if you can switch to Pro version.
You really do not need so much RAM, usually 4Gb works even for maximum of 8 GPU (4xHD6990)

@cjp0008
Copy link

cjp0008 commented Apr 28, 2018

Friends, now I'm trying to install the 14.4 driver in Windows 10 Pro. I used AMD's "clean uninstall" utility to remove any previous AMD drivers, and downloaded the 14.4 drivers nominally for Windows 8.1 from AMD's site. When I run the installer on Express mode, it finishes with errors, saying it was unable to install the driver because "Driver Install: the driver package does not specify a hardware identifier". Can anyone help? I know this is off topic but Google didn't turn up anything....

@Simaex
Copy link

Simaex commented Apr 28, 2018

@cjp0008 Do not use AMD utility, use DDU
http://www.guru3d.com/files-details/display-driver-uninstaller-download.html

@GrigoryMel
Copy link

Я немного убрал частоту ядра и получилось 792,8 х/с, иногда проскакивает 800х/с надо пробовать варианты.

Архив ZIP - WinRAR.zip

@leou0
Copy link

leou0 commented May 6, 2018

5-ый день не могу победить!
win7 x64, driver 14.4, 6xGPU_mod_1_0_0_1, 4хHD6990
после переустановки драйверов с 15.7.1 на 14.4 (все как полагается с DDU) на 2-ой день перестал работать MSI Afterburner (пробовал разные версии 3, 4, 4.5), но это пол беды, при майнинге отваливаются GPU по очереди (без ошибок, без синих эранов и т.п.) майнер через какое то время зависает, спасает перезапуск майнера. С настройками самого майнера и с разгоном карт игрался (не спасает)
"gpu_threads_conf" : [
// gpu: Cayman memory:1663
// compute units: 24
{ "index" : 0,
"intensity" : 880, "worksize" : 16,
"affine_to_cpu" : false, "strided_index" : 2, "mem_chunk" : 2,
"comp_mode" : true
},
// gpu: Cayman memory:1663
// compute units: 24
{ "index" : 1,
"intensity" : 880, "worksize" : 16,
"affine_to_cpu" : false, "strided_index" : 2, "mem_chunk" : 2,
"comp_mode" : true
},
Пробывал Win 10, тоже самое... проблема не в железе (проверено)

@Russia102
Copy link

leou0 Пиши на почту помогу. berqaga83@gmail.com

@Simaex
Copy link

Simaex commented May 6, 2018

@leou0 do not use Win7 for mining. Try to fix your problems using Win10. Do fresh install Win10 Pro, 14.4 drivers, switch off updates. Do not install AB and give it a try. You should get around 765 H\s per HD6990 rock stable provided you have normal ambient temperatures. If you want to try performance settingth you can use Claymore 9.7 miner to set your core and mem clocks and core voltage. GPU will retain the values until next reboot. Try to set your default display to on-board grafics if available else you have to fine-tune GPU that is used for video output. It may cause major performance drop for it or stability issues.

@leou0
Copy link

leou0 commented May 6, 2018

Встроенная графика отсутствует в мат. плате, на win 10 AB работает вполне корректно, 1 из 8 ядер не майнит (т.к. нет встроенной графики).... но основная проблема не в этом, происходит хаотичный отвал графических ядер и далее зависание майнера как в win7 так и в win 10

@Simaex
Copy link

Simaex commented May 7, 2018

@leou0 just get rid of AB and lower your intensity to the level that ALL GPU are mining. If you will still get stability issues it is you MB PCIe controller, try other MB.

@stefanbuerger
Copy link

What is meant with AB? Please explain.

@Simaex
Copy link

Simaex commented May 7, 2018

@stefanbuerger I mean that when trying to troubleshoot an issue you should get rid of everything that may cause the problem exept miner itself. If it helps it is not a miner problem. BTW I explained how to OC the card without using AB (AfterBurner) at all. I have issues with AB and HD6990 cards some time ago.

@leou0
Copy link

leou0 commented May 7, 2018

С AB разобрался, проблема оказалась в мониторе (при отключенном мониторе и подключении через RDP, AB не работает. Через VNC все ОК)... Но отвалы остались, повторяю, проблема не в железе, ПРОВЕРЕНО! На Claymore v.9.7 подобных проблем не наблюдается. Разгона карт нет: 880 по ядру и 1250 по памяти. Перегрева нет, максимум 74 градуса.

// gpu: Cayman memory:1663
// compute units: 24
{ "index" : 0,
"intensity" : 800, "worksize" : 19,
"affine_to_cpu" : false, "strided_index" : 1, "mem_chunk" : 2,
"comp_mode" : true

При таких настройках - 350 H/s на ядро (700 h/s с карты)
Интенсивность снижал, не помогает...
Если кто может помочь, буду очень благодарен. mamonovdo@outlook.com

@leou0
Copy link

leou0 commented May 8, 2018

Есть предположение что проблема в 6xGPU mod, есть ли еще какие либо способы исправить ошибку 43?

@Milierkovic
Copy link

I have a very similar problem with my AMD Sapphire HD 6970 2GB Graphics Card. It only sees i have 384 Cayman memory?

This is a part of my amd.txt file:

"gpu_threads_conf" : [
// gpu: Cayman memory:384
// compute units: 24
{ "index" : 0,
"intensity" : 191, "worksize" : 8,
"affine_to_cpu" : false, "strided_index" : 1, "mem_chunk" : 2,
"comp_mode" : true
},

I tried many different settinsg and versions of XMR STak and ipbc-miner-win64-cuda8-1.1.4 which partialy has the same code, but i cant get it to use much more memory and have a higher H/s.

The max intensity i can use is about 256.

Please help or advice me, i think i should be able to have much higher hashrates.

@psychocrypt
Copy link
Collaborator

psychocrypt commented May 12, 2018 via email

@Milierkovic
Copy link

I started with xmr-stak i tried different versions then i tried different versions ipbc miner because my goal is to mine ipbc but somehow the miner does not see my 2GB video memory.

image

image

@xmronadaily
Copy link

xmronadaily commented May 13, 2018 via email

@Simaex
Copy link

Simaex commented May 13, 2018

@Milierkovic You are usung 15.2 driver. It is known problem discussed above. Deinstall 15.2 using DDU and Install 14.4 driver to access all of GPU video memory and try to use worksize from 17 to 22. Usually 19 work best. You can expect 400+ H/s for monero7 algo. If you are using the HD6970 as a primary display card you can have additional problems that affect performance.

@leou0
Copy link

leou0 commented May 13, 2018

Нужен рабочий скрипт (bat) на перезапуск xmr-stack при отвале драйвера... на других площадках годного не нашел...

@Milierkovic
Copy link

Milierkovic commented May 13, 2018

@Simaex thanks somehow i did not understand this, i am trying to fix it as we speak!

@leou0 please share the (bat) script with me when you found it

@Milierkovic
Copy link

Milierkovic commented May 14, 2018

@Simaex Problem solved!

"gpu_threads_conf" : [
// gpu: Cayman memory:1663
// compute units: 24
{ "index" : 0,
"intensity" : 960, "worksize" : 19,
"affine_to_cpu" : false, "strided_index" : 1, "mem_chunk" : 2,
"comp_mode" : true
},

I have a very good hashrate now:

HASHRATE REPORT - AMD
| ID | 10s | 60s | 15m |
| 0 | 768.4 | 763.9 | 764.2 |
Totals (AMD): 768.4 763.9 764.2 H/s

I am using ipbc-miner-1.1-win64 now. Thank you very much for your help

EDIT: only invalid results :(

@Russia102
Copy link

Ктонибудь пробовал на 69 серии запускать TRTL? Алгоритм Lite_v7 похоже вообще не работает на 69xx) хотя на монеро идет отлично. Пробовал разные майнеры ни на одном не пошел этот алгоритм.

@GrigoryMel
Copy link

Всем привет! Помогите решить проблему, майнер (xmrig-amd) выдает ошибки, а именно:

2018-05-14 23:01:48] speed 10s/60s/15m n/a n/a n/a H/s max: 750.0 H/s
[2018-05-14 23:01:55] [pool.supportxmr.com:7777] error: "Unauthenticated", code:
-1
[2018-05-14 23:01:55] no active pools, stop mining
[2018-05-14 23:02:01] use pool pool.supportxmr.com:7777 212.232.25.157
[2018-05-14 23:02:01] new job from pool.supportxmr.com:7777 diff 30000
[2018-05-14 23:02:48] speed 10s/60s/15m n/a n/a n/a H/s max: 750.0 H/s
[2018-05-14 23:03:11] new job from pool.supportxmr.com:7777 diff 30000

Это происходит, непонятно по какой закономерности, если кто сталкивался поделитесь. Заранее спасибо!

@Spudz76
Copy link
Contributor

Spudz76 commented Jun 4, 2018

Downgraded my Ubuntu 14 to kernel 3.2 using old precise repo and a preferences file to pin just the kernel and junk, installed fglrx 13.152 (aka amd-catalyst-13.9-linux-x86.x86_64) and getting personal best ever 173.4H/s on this Sapphire HD5570 1GB (stock clocks, stock bios) using

  // gpu: Redwood memory:896
  // compute units: 5
  { "index" : 0,
    "intensity" : 510, "worksize" : 5,
    "affine_to_cpu" : false, "strided_index" : 1, "mem_chunk" : 2,
    "comp_mode" : true
  },

Any more intensity ran out of memory. This driver (referenced just above) and kernel combo works nice.
Also runs the Tahiti LE cards in same rig just same as all other versions (too bad Terascales couldn't have been so workable in all versions)

@whatifitlives
Copy link

Does anyone know of a complete step by step procedure for this?

I am currently testing with a HD 5450 and 5570s. I tried using drivers 14.4 and I just tried with 13.12 and I have had no luck with my GPUs doing anything. The CPU miner works just fine. Did the latest release update kill this trick (2.4.5)? I am just looking for guidance on what people did. I have a fresh install of windows 10, pulled the c++ 2015 libraries, and have the gpus above.

Thoughts? I would like to get this working as I have a few 5900, 5800, and 6900 cards.

Thanks for any help.

@Simaex
Copy link

Simaex commented Jul 9, 2018

Please start from reading the thread. All necessary info is already mentioned. If still no luck please be more specific in your info and include all info like autogenerated config files, clinfo output, windows version. As 69XX make much more sense start with single 69XX card. Later you will have chance to practice with 5XXX cards.

@Spudz76
Copy link
Contributor

Spudz76 commented Jul 16, 2018

13.xx are the only ones that work.
If you didn't use DDU to scrub system after 14.x before installing 13.x that may be the issue, but I don't know for sure as I don't use AMD on Windows due to requirement of dummy plugs/actual screens hooked up.

Speaking of which, none of the AMD's that aren't connected to a "screen" (or dummy plug aka screen without a screen) under Windows will show up for OpenCL usage at all. So if you don't have any of them on a LCD or whatever they won't show up (other than in Device Manager) regardless of having the right drivers. Thanks AMD, I guess you think a GPU is only for displaying graphics (idiots!)

Run Linux if you want actual headless AMD mining. Same 13.xx driver limit applies, which means fglrx, which means nothing newer than Ubuntu 14.x (you need the antique kernel/Xorg versions, or fglrx doesn't work).

Or buy three dummy plugs. Test by hooking up some old LCD to one of them first, it should magically show up for work then.

Also if you have Intel iGPU drivers installed, you have to kill the OpenCL part of that (or uninstall the driver / use Standard VGA for it) else it will crash the opencl backend during probe/startup. Thanks Intel...

Latest release (and dev branch too) all work great on my HD5570, just recompiled last night.

xmr-stak 2.4.5 03a036c

HASHRATE REPORT - CPU
| ID |    10s |    60s |    15m | ID |    10s |    60s |    15m |
|  0 |   40.4 |   40.4 |   40.4 |  1 |   39.8 |   39.9 |   39.9 |
|  2 |   37.9 |   37.9 |   37.9 |  3 |   35.1 |   35.2 |   35.1 |
|  4 |   40.6 |   40.7 |   40.7 |  5 |   39.6 |   39.7 |   39.7 |
|  6 |   34.7 |   34.7 |   34.7 |
Totals (CPU):   268.1  268.3  268.3 H/s
-----------------------------------------------------------------
HASHRATE REPORT - AMD
| ID |    10s |    60s |    15m | ID |    10s |    60s |    15m |
|  0 |  173.2 |  173.5 |  173.5 |  1 |  329.3 |  329.6 |  329.5 |
|  2 |  328.1 |  329.4 |  329.5 |
Totals (AMD):   830.6  832.4  832.5 H/s
-----------------------------------------------------------------
HASHRATE REPORT - NVIDIA
| ID |    10s |    60s |    15m | ID |    10s |    60s |    15m |
|  0 |  219.9 |  219.9 |  219.9 |  1 |  179.2 |  179.2 |  179.2 |
|  2 |  188.1 |  188.1 |  188.1 |
Totals (NVIDIA):   587.2  587.2  587.2 H/s
-----------------------------------------------------------------
Totals (ALL):   1685.9 1688.0 1688.1 H/s
Highest:  1689.8 H/s
-----------------------------------------------------------------

The AMD thread 0 doing ~173 is the HD5570, the other two are Tahiti (HD68xx or HD69xx or such) which also work nicely on the old drivers.
CUDA 8.0 also works fine under ancient Ubuntu 14 with downgraded kernel 3.2.x (only situation where fglrx still works) those are some GTX550/GTX465 old things running on the same rig.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests