AnsweredAssumed Answered

OpenCL GPU device not detected in Ubuntu 12.04

Question asked by clipper on Oct 26, 2012
Latest reply on Nov 22, 2012 by clipper
andreas@mercurius:~/phoenix_miner/phoenix-1.60$ clinfo 
Number of platforms:                                         1
  Platform Profile:                                         FULL_PROFILE
  Platform Version:                                         OpenCL 1.2 AMD-APP (1016.4)
  Platform Name:                                         AMD Accelerated Parallel Processing
  Platform Vendor:                                         Advanced Micro Devices, Inc.
  Platform Extensions:                                         cl_khr_icd cl_amd_event_callback cl_amd_offline_devices




  Platform Name:                                         AMD Accelerated Parallel Processing
Number of devices:                                         3
  Device Type:                                                   CL_DEVICE_TYPE_GPU
  Device ID:                                                   4098
  Board name:                                                   ATI Radeon HD 5800 Series 
  Device Topology:                                         PCI[ B#3, D#0, F#0 ]
  Max compute units:                                         18
  Max work items dimensions:                               3
    Max work items[0]:                                         256
    Max work items[1]:                                         256
    Max work items[2]:                                         256
  Max work group size:                                         256
  Preferred vector width char:                               16
  Preferred vector width short:                               8
  Preferred vector width int:                               4
  Preferred vector width long:                               2
  Preferred vector width float:                               4
  Preferred vector width double:                     2
  Native vector width char:                               16
  Native vector width short:                               8
  Native vector width int:                               4
  Native vector width long:                               2
  Native vector width float:                               4
  Native vector width double:                               2
  Max clock frequency:                                         870Mhz
  Address bits:                                                   32
  Max memory allocation:                               134217728
  Image support:                                         Yes
  Max number of images read arguments:                     128
  Max number of images write arguments:                     8
  Max image 2D width:                                         8192
  Max image 2D height:                                         8192
  Max image 3D width:                                         2048
  Max image 3D height:                                         2048
  Max image 3D depth:                                         2048
  Max samplers within kernel:                               16
  Max size of kernel argument:                               1024
  Alignment (bits) of base address:                     2048
  Minimum alignment (bytes) for any datatype:           128
  Single precision floating point capability
    Denorms:                                                   No
    Quiet NaNs:                                                   Yes
    Round to nearest even:                               Yes
    Round to zero:                                         Yes
    Round to +ve and infinity:                               Yes
    IEEE754-2008 fused multiply-add:                     Yes
  Cache type:                                                   None
  Cache line size:                                         0
  Cache size:                                                   0
  Global memory size:                                         536870912
  Constant buffer size:                                         65536
  Max number of constant args:                               8
  Local memory type:                                         Scratchpad
  Local memory size:                                         32768
  Kernel Preferred work group size multiple:           64
  Error correction support:                               0
  Unified memory for Host and Device:                     0
  Profiling timer resolution:                               1
  Device endianess:                                         Little
  Available:                                                   Yes
  Compiler available:                                         Yes
  Execution capabilities:                                         
    Execute OpenCL kernels:                               Yes
    Execute native function:                               No
  Queue properties:                                         
    Out-of-Order:                                         No
    Profiling :                                                   Yes
  Platform ID:                                                   0xb73fcf20
  Name:                                                             Cypress
  Vendor:                                                   Advanced Micro Devices, Inc.
  Device OpenCL C version:                               OpenCL C 1.2 
  Driver version:                                         1016.4
  Profile:                                                   FULL_PROFILE
  Version:                                                   OpenCL 1.2 AMD-APP (1016.4)
  Extensions:                                                   cl_khr_fp64 cl_amd_fp64 cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_3d_image_writes cl_khr_byte_addressable_store cl_khr_gl_sharing cl_ext_atomic_counters_32 cl_amd_device_attribute_query cl_amd_vec3 cl_amd_printf cl_amd_media_ops cl_amd_popcnt 




  Device Type:                                                   CL_DEVICE_TYPE_GPU
  Device ID:                                                   4098
  Board name:                                                   ATI Radeon HD 5800 Series 
  Device Topology:                                         PCI[ B#3, D#0, F#0 ]
  Max compute units:                                         18
  Max work items dimensions:                               3
    Max work items[0]:                                         256
    Max work items[1]:                                         256
    Max work items[2]:                                         256
  Max work group size:                                         256
  Preferred vector width char:                               16
  Preferred vector width short:                               8
  Preferred vector width int:                               4
  Preferred vector width long:                               2
  Preferred vector width float:                               4
  Preferred vector width double:                     2
  Native vector width char:                               16
  Native vector width short:                               8
  Native vector width int:                               4
  Native vector width long:                               2
  Native vector width float:                               4
  Native vector width double:                               2
  Max clock frequency:                                         870Mhz
  Address bits:                                                   32
  Max memory allocation:                               134217728
  Image support:                                         Yes
  Max number of images read arguments:                     128
  Max number of images write arguments:                     8
  Max image 2D width:                                         8192
  Max image 2D height:                                         8192
  Max image 3D width:                                         2048
  Max image 3D height:                                         2048
  Max image 3D depth:                                         2048
  Max samplers within kernel:                               16
  Max size of kernel argument:                               1024
  Alignment (bits) of base address:                     2048
  Minimum alignment (bytes) for any datatype:           128
  Single precision floating point capability
    Denorms:                                                   No
    Quiet NaNs:                                                   Yes
    Round to nearest even:                               Yes
    Round to zero:                                         Yes
    Round to +ve and infinity:                               Yes
    IEEE754-2008 fused multiply-add:                     Yes
  Cache type:                                                   None
  Cache line size:                                         0
  Cache size:                                                   0
  Global memory size:                                         536870912
  Constant buffer size:                                         65536
  Max number of constant args:                               8
  Local memory type:                                         Scratchpad
  Local memory size:                                         32768
  Kernel Preferred work group size multiple:           64
  Error correction support:                               0
  Unified memory for Host and Device:                     0
  Profiling timer resolution:                               1
  Device endianess:                                         Little
  Available:                                                   Yes
  Compiler available:                                         Yes
  Execution capabilities:                                         
    Execute OpenCL kernels:                               Yes
    Execute native function:                               No
  Queue properties:                                         
    Out-of-Order:                                         No
    Profiling :                                                   Yes
  Platform ID:                                                   0xb73fcf20
  Name:                                                             Cypress
  Vendor:                                                   Advanced Micro Devices, Inc.
  Device OpenCL C version:                               OpenCL C 1.2 
  Driver version:                                         1016.4
  Profile:                                                   FULL_PROFILE
  Version:                                                   OpenCL 1.2 AMD-APP (1016.4)
  Extensions:                                                   cl_khr_fp64 cl_amd_fp64 cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_3d_image_writes cl_khr_byte_addressable_store cl_khr_gl_sharing cl_ext_atomic_counters_32 cl_amd_device_attribute_query cl_amd_vec3 cl_amd_printf cl_amd_media_ops cl_amd_popcnt 




  Device Type:                                                   CL_DEVICE_TYPE_CPU
  Device ID:                                                   4098
  Board name:                                                   
  Max compute units:                                         6
  Max work items dimensions:                               3
    Max work items[0]:                                         1024
    Max work items[1]:                                         1024
    Max work items[2]:                                         1024
  Max work group size:                                         1024
  Preferred vector width char:                               16
  Preferred vector width short:                               8
  Preferred vector width int:                               4
  Preferred vector width long:                               2
  Preferred vector width float:                               4
  Preferred vector width double:                     2
  Native vector width char:                               16
  Native vector width short:                               8
  Native vector width int:                               4
  Native vector width long:                               2
  Native vector width float:                               4
  Native vector width double:                               2
  Max clock frequency:                                         800Mhz
  Address bits:                                                   32
  Max memory allocation:                               1073741824
  Image support:                                         Yes
  Max number of images read arguments:                     128
  Max number of images write arguments:                     8
  Max image 2D width:                                         8192
  Max image 2D height:                                         8192
  Max image 3D width:                                         2048
  Max image 3D height:                                         2048
  Max image 3D depth:                                         2048
  Max samplers within kernel:                               16
  Max size of kernel argument:                               4096
  Alignment (bits) of base address:                     1024
  Minimum alignment (bytes) for any datatype:           128
  Single precision floating point capability
    Denorms:                                                   Yes
    Quiet NaNs:                                                   Yes
    Round to nearest even:                               Yes
    Round to zero:                                         Yes
    Round to +ve and infinity:                               Yes
    IEEE754-2008 fused multiply-add:                     Yes
  Cache type:                                                   Read/Write
  Cache line size:                                         64
  Cache size:                                                   65536
  Global memory size:                                         3758096384
  Constant buffer size:                                         65536
  Max number of constant args:                               8
  Local memory type:                                         Global
  Local memory size:                                         32768
  Kernel Preferred work group size multiple:           1
  Error correction support:                               0
  Unified memory for Host and Device:                     1
  Profiling timer resolution:                               1
  Device endianess:                                         Little
  Available:                                                   Yes
  Compiler available:                                         Yes
  Execution capabilities:                                         
    Execute OpenCL kernels:                               Yes
    Execute native function:                               Yes
  Queue properties:                                         
    Out-of-Order:                                         No
    Profiling :                                                   Yes
  Platform ID:                                                   0xb73fcf20
  Name:                                                             AMD Phenom(tm) II X6 1090T Processor
  Vendor:                                                   AuthenticAMD
  Device OpenCL C version:                               OpenCL C 1.2 
  Driver version:                                         1016.4 (sse2)
  Profile:                                                   FULL_PROFILE
  Version:                                                   OpenCL 1.2 AMD-APP (1016.4)
  Extensions:                                                   cl_khr_fp64 cl_amd_fp64 cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_3d_image_writes cl_khr_byte_addressable_store cl_khr_gl_sharing cl_ext_device_fission cl_amd_device_attribute_query cl_amd_vec3 cl_amd_printf cl_amd_media_ops cl_amd_popcnt 




andreas@mercurius:~/phoenix_miner/phoenix-1.60$ aticonfig --adapter=all --odgt --odgc
Adapter 0 - ATI Radeon HD 5800 Series 
            Sensor 0: Temperature - 54.00 C


Adapter 1 - ATI Radeon HD 5800 Series  
            Sensor 0: Temperature - 42.50 C


Adapter 0 - ATI Radeon HD 5800 Series 
                            Core (MHz)    Memory (MHz)
           Current Clocks :    157           300
             Current Peak :    870           300
  Configurable Peak Range : [550-1200]     [900-1400]
                 GPU load :    0%


Adapter 1 - ATI Radeon HD 5800 Series  
                            Core (MHz)    Memory (MHz)
           Current Clocks :    157           300
             Current Peak :    1000           300
  Configurable Peak Range : [600-875]     [900-1200]
                 GPU load :    0%
andreas@mercurius:~/phoenix_miner/phoenix-1.60$ aticonfig --adapter=all --lsa
* 0. 01:00.0 ATI Radeon HD 5800 Series 
  1. 03:00.0 ATI Radeon HD 5800 Series  


* - Default adapter
andreas@mercurius:~/phoenix_miner/phoenix-1.60$ 

 

device0 is a 5850, device1 is a 5830, using a fresh installl of Ubuntu 12.04, driver 12.10 and sdk 2.7.

As you can see above, line 36 (first device) and line 113 (second device) are the same:

Max clock frequency:


870Mhz

But according to "aticonfig --adapter=all --odgt --odgc", for the second device it should be 1000Mhz. When i'm assigning work to device1, it is forwarded to device0 and in case device0 is already in use, the processes are cannibalizing each other. No error message. Tried this http://devgurus.amd.com/thread/159566 which made the GPUs available through SSH, but did not solve the problem. The same configuration worked fine with Ubuntu 10.04. For SSH-GPU-access i didn't even have to make the changes proposed in http://devgurus.amd.com/thread/159566 .

Is it an issue of driver, sdk or Ubuntu or have i done s.th. wrong?

Outcomes