Benchmark for AMD FirePro W5100

BeO

• 6. Januar 2016 21:23

Does somebody have a Capture One OpenCL benchmark figure from imgcore.log for this gfx card?
Noise level and temperature complaints?

Can you also pls. list your CPU type? (ix Core or Xeon)

Thx in advance
BeO

0

• 

  Christian Gruner

  • 7. Januar 2016 08:15

  2016-01-05 11:12:16.105> OpenCL Device : Bonaire
  2016-01-05 11:12:16.105> OpenCL Driver Version : 1800.11 (VM)
  2016-01-05 11:12:16.105> OpenCL Compute Units : 12
  2016-01-05 11:12:16.183> OpenCL : Building kernels
  2016-01-05 11:38:54.708> OpenCL : Benchmarking
  2016-01-05 11:38:54.969> OpenCL : Initialization completed
  2016-01-05 11:38:54.969> OpenCL benchMark : 0.237481
  
  Running a 5k display. Not particularly fast for CO, especially not for the amount of money it costs.
  CPU is AMD is this particular machine.

  0
• 

  BeO

  • 7. Januar 2016 11:02

  

  Thanks Christian.
  
  Do you know when you will release more benchmark information as Lionel indicated in the capture one blog comment about OpenCL?
  
  Is the benchmark figure linear, meaning 0.12 will be twice as fast as 0.24?
  
  Is there a fix threshhold when the gfx card will be used instead of the processor (e.g. 1.0) or is it relative to the CPU speed (e.g. 0.3 with slow CPU, 0.1 with fast CPU)?
  
  FirePro W5100 will support OpenCL 2.0. Does that have any meaning for C1?
  
  THanks
  BeO

  0
• 

  Christian Gruner

  • 7. Januar 2016 12:21

  [quote="BeO" wrote:
  Thanks Christian.
  
  Do you know when you will release more benchmark information as Lionel indicated in the capture one blog comment about OpenCL?
  
  Is the benchmark figure linear, meaning 0.12 will be twice as fast as 0.24?
  
  Is there a fix threshhold when the gfx card will be used instead of the processor (e.g. 1.0) or is it relative to the CPU speed (e.g. 0.3 with slow CPU, 0.1 with fast CPU)?
  
  FirePro W5100 will support OpenCL 2.0. Does that have any meaning for C1?
  
  THanks
  BeO

  
  
  There are some technical and resource questions regarding the list, so no promises yet. Generally,, buy the fastest card you can afford. The more CUDA cores (nvidia) or Stream Processors (amd) the faster. Also, the faster ram-bus the faster processing, but this is marginal compared to the number of cores/processors on the GPU.
  
  The Benchmark is to a certain extent linear.
  
  The threshold for enabling of OpenCL currently has no external factors involved. If the ram amount is under 1024 mb, or the benchmark is over 2.5, CO will not initialize the OpenCL Pipeline.
  
  OpenCL 2.0 gives a small performance gain of 3-5% (in raw benchmark, so the gain will be even lower when all factors of processing are added to the equation).

  0
• 

  BeO

  • 7. Januar 2016 19:59

  

  Thank you much for your answer Christian.
  
  

  [quote="Christian Gruner" wrote:
  The threshold for enabling of OpenCL currently has no external factors involved. If the ram amount is under 1024 mb, or the benchmark is over 2.5, CO will not initialize the OpenCL Pipeline.

  
  
  Makes sense. Then the user can decide to switch opencl off in the preferences and see if CPU or GPU is faster.
  
  Cheers
  BeO

  0

Post ist für Kommentare geschlossen.