1 Reply Latest reply on Jun 16, 2011 10:52 AM by himanshu.gautam

    does opencl give wrong info about the devices?

    guruvyasa

      I ran the ranlux random number generator posted in amd forums..on the GPU the information it is giving is:

       

      Num platforms:       1

      Platform Vendor:     Advanced Micro Devices, Inc.

      Platform profile:    FULL_PROFILE

      Platform version:    OpenCL 1.1 ATI-Stream-v2.2 (302)

      Platform name:       ATI Stream

      Platform extensions: cl_khr_icd cl_amd_event_callback

      Device name:         Cedar

      Device type:         GPU 

      Device vendor ID:    4098

      Max compute units:   2

      Max work item dim:   3

      Max work group size: 128

      Max mem alloc size:  128 MiB

      Global memory size:  512 MiB

      Global memory cache: None

      Local memory type:   Dedicated

      Prof Timer Res:      1 ns

      OpenCL Driver ver:   CAL 1.4.838

      Building OpenCL program: Done

      the max compute units is only 2 and global memory is 512MB.. My GPU is Radeon HD 6370M with 1GB memory..so is this correct?
      On the other hand for CPU execution the CPU details are:
      Num platforms:       1
      Platform Vendor:     Advanced Micro Devices, Inc.
      Platform profile:    FULL_PROFILE
      Platform version:    OpenCL 1.1 ATI-Stream-v2.2 (302)
      Platform name:       ATI Stream
      Platform extensions: cl_khr_icd cl_amd_event_callback
      Device name:         Intel(R) Core(TM) i7 CPU       Q 740  @ 1.73GHz
      Device type:         CPU 
      Device vendor ID:    4098
      Max compute units:   8
      Max work item dim:   3
      Max work group size: 1024
      Max mem alloc size:  512 MiB
      Global memory size:  1024 MiB
      Global memory cache: Read/Write cache
      Local memory type:   Global memory
      Prof Timer Res:      1 ns
      OpenCL Driver ver:   2.0
      Building OpenCL program: Done
      however the speed of number generation on CPU is:
      Generation speed:        44.20 million numbers per second
      and that on GPU is :
      Generation speed:        138.19 million numbers per second
      So what exactly is happening here? the device info shows that CPU has more number of cores and memory but results show that GPU is better..please help me..i did not find the specs for AMD HD 6370M on the net..