Zum Hauptinhalt gehen

⚠️ Please note that this topic or post has been archived. The information contained here may no longer be accurate or up-to-date. ⚠️

GTX 970 OpenCL not working?

WPNL
Below the ticket I created, since there's no response of any kind I figured to drop it here also 😊

Hi guys,
I upgraded mu GPU from a GTX285 to a GTX970 but there's no improvement, it looks like opencl is not used. I found this in my log:
2015-03-28 09:52:56.288> Logging is now active.
2015-03-28 09:52:56.288> OpenCL initialization...
2015-03-28 09:52:56.641> OpenCL : found platform NVIDIA CUDA, OpenCL Version : OpenCL 1.1 CUDA 7.0.29
2015-03-28 09:52:56.666> OpenCL Device : GeForce GTX 970
2015-03-28 09:52:56.666> OpenCL Driver Version : 347.88
2015-03-28 09:52:56.666> OpenCL Compute Units : 13
2015-03-28 09:52:56.906> Shutting down

Can you help me out, thanks a lot!
0

Kommentare

25 Kommentare

Datum Stimmen
  • Paul Topol
    Hi, I have a similar problem. Got the same card and display is very slow.
    Move a slider then wait, move a slider then wait.........

    Hope somebody comes up with a solution.

    Paul
    0
  • Filthy Lucre
    Have you enabled opencl in the c1 prefs?
    0
  • Wolfram HĂślzel
    Could this be also a problem of the 3.5 GB-RAM Problem?
    0
  • Paul Topol
    [quote="PhaseoneUser40673" wrote:
    Have you enabled opencl in the c1 prefs?


    Of course
    0
  • dee jjjaaaa
    [quote="NN635531119132438914UL" wrote:
    Below the ticket I created, since there's no response of any kind I figured to drop it here also 😊

    Hi guys,
    I upgraded mu GPU from a GTX285 to a GTX970 but there's no improvement, it looks like opencl is not used. I found this in my log:
    2015-03-28 09:52:56.288> Logging is now active.
    2015-03-28 09:52:56.288> OpenCL initialization...
    2015-03-28 09:52:56.641> OpenCL : found platform NVIDIA CUDA, OpenCL Version : OpenCL 1.1 CUDA 7.0.29
    2015-03-28 09:52:56.666> OpenCL Device : GeForce GTX 970
    2015-03-28 09:52:56.666> OpenCL Driver Version : 347.88
    2015-03-28 09:52:56.666> OpenCL Compute Units : 13
    2015-03-28 09:52:56.906> Shutting down

    Can you help me out, thanks a lot!


    which log are you talking about ? ImgCoreOCL.log ?

    what is in ImgCore.log or ImgCoreExternal.log ?
    0
  • Christian Gruner
    What kinda files are you working on ?

    What you see in the imagecoreOCL log is perfectly normal.
    0
  • Filthy Lucre
    [quote="Christian Gruner" wrote:

    What you see in the imagecoreOCL log is perfectly normal.

    ?? Seems incomplete to me. Mine looks like this:

    2015-02-08 14:48:40.112> Logging is now active.
    2015-02-08 14:48:40.775> OpenCL initialization...
    2015-02-08 14:48:40.978> OpenCL : found platform NVIDIA CUDA, OpenCL Version : OpenCL 1.1 CUDA 6.5.30
    2015-02-08 14:48:40.978> OpenCL Device : GeForce GTX 650
    2015-02-08 14:48:40.978> OpenCL Driver Version : 344.75
    2015-02-08 14:48:40.978> OpenCL Compute Units : 2
    2015-02-08 14:48:41.072> OpenCL : Loading kernels
    2015-02-08 14:48:48.824> OpenCL : Loading kernels finished
    2015-02-08 14:48:48.824> OpenCL : Benchmarking
    2015-02-08 14:48:49.246> OpenCL : Initialization completed
    2015-02-08 14:48:49.246> OpenCL benchMark : 1.104896
    2015-02-08 14:53:35.138> OpenCL allocations : 17944080 bytes
    2015-02-08 14:53:35.850> Shutting down
    0
  • Christian Gruner
    As i said, this completely as it should be.

    This log builds the kernels for the GPU's. When the kernels are built, or already exists (already built once), the compiler shuts down. This log has nothing to do with the in-app use of OpenCL.
    0
  • Paul Topol
    This comes from the ImgCoreExternal.log file.

    Note the benchmark number.
    maybe that's an indication showing how slow my system is.
    Any help would be appreciated.


    2015-03-31 08:41:45.007> Logging is now active.
    2015-03-31 08:41:49.402> OpenCL initialization...
    2015-03-31 08:41:49.503> OpenCL : found platform AMD Accelerated Parallel Processing, OpenCL Version : OpenCL 2.0 AMD-APP (1642.5)
    2015-03-31 08:41:49.503> OpenCL : found platform NVIDIA CUDA, OpenCL Version : OpenCL 1.1 CUDA 7.0.29
    2015-03-31 08:41:49.521> OpenCL Device : GeForce GTX 980
    2015-03-31 08:41:49.521> OpenCL Driver Version : 347.88
    2015-03-31 08:41:49.521> OpenCL Compute Units : 16
    2015-03-31 08:41:49.685> OpenCL : Loading kernels
    2015-03-31 08:41:50.327> OpenCL : Loading kernels finished
    2015-03-31 08:41:50.327> OpenCL : Benchmarking
    2015-03-31 08:41:50.456> OpenCL : Initialization completed
    2015-03-31 08:41:50.456> OpenCL benchMark : 0.168960
    2015-03-31 08:50:34.696> Exiting
    2015-03-31 08:50:34.747> Shutting down
    2015-03-31 08:50:35.248> (ERROR) Timed out waiting for executers [IC_External].
    2015-03-31 08:50:35.751> (ERROR) Timed out waiting for executers [PYRAMID_IC_External].
    2015-03-31 08:50:36.254> (ERROR) Timed out waiting for executers [Processor for creating thumbnails].
    2015-03-31 08:50:36.755> (ERROR) Timed out waiting for executers [PYRAMID_Processor for creating thumbnails].


    Paul
    0
  • Christian Gruner
    [quote="Paul811" wrote:
    This comes from the ImgCoreExternal.log file.

    Note the benchmark number.
    maybe that's an indication showing how slow my system is.
    Any help would be appreciated.


    2015-03-31 08:41:45.007> Logging is now active.
    2015-03-31 08:41:49.402> OpenCL initialization...
    2015-03-31 08:41:49.503> OpenCL : found platform AMD Accelerated Parallel Processing, OpenCL Version : OpenCL 2.0 AMD-APP (1642.5)
    2015-03-31 08:41:49.503> OpenCL : found platform NVIDIA CUDA, OpenCL Version : OpenCL 1.1 CUDA 7.0.29
    2015-03-31 08:41:49.521> OpenCL Device : GeForce GTX 980
    2015-03-31 08:41:49.521> OpenCL Driver Version : 347.88
    2015-03-31 08:41:49.521> OpenCL Compute Units : 16
    2015-03-31 08:41:49.685> OpenCL : Loading kernels
    2015-03-31 08:41:50.327> OpenCL : Loading kernels finished
    2015-03-31 08:41:50.327> OpenCL : Benchmarking
    2015-03-31 08:41:50.456> OpenCL : Initialization completed
    2015-03-31 08:41:50.456> OpenCL benchMark : 0.168960
    2015-03-31 08:50:34.696> Exiting
    2015-03-31 08:50:34.747> Shutting down
    2015-03-31 08:50:35.248> (ERROR) Timed out waiting for executers [IC_External].
    2015-03-31 08:50:35.751> (ERROR) Timed out waiting for executers [PYRAMID_IC_External].
    2015-03-31 08:50:36.254> (ERROR) Timed out waiting for executers [Processor for creating thumbnails].
    2015-03-31 08:50:36.755> (ERROR) Timed out waiting for executers [PYRAMID_Processor for creating thumbnails].


    Paul

    The benchmark for that card seems right. It's not a slow card at all.
    0
  • dee jjjaaaa
    [quote="Paul811" wrote:

    2015-03-31 08:41:49.521> OpenCL Device : GeForce GTX 980
    2015-03-31 08:41:50.456> OpenCL benchMark : 0.168960


    it was noted by P1 staff here that the lower OpenCL benchMark number the better (non intuitive of course).

    for your reference my notebook with GTX 870m (notebook card + one generation older + one tier lesser) is "0.328736" and iHD4600 in i7-4810MQ is "0.887280"
    0
  • Paul Topol
    Thanks Christian,

    So why do I have the slow response to almost every adjustment?

    Could it have something to do with full disks?

    Running windows 7 64bit with 16gb memory.

    Would be better to use a ssd drive?

    Thanks

    Paul
    0
  • Christian Gruner
    [quote="Paul811" wrote:
    Thanks Christian,

    So why do I have the slow response to almost every adjustment?

    Could it have something to do with full disks?

    Running windows 7 64bit with 16gb memory.

    Would be better to use a ssd drive?

    Thanks

    Paul


    I have too little information to speculate. Please create a Support case, and they will get you going.
    0
  • WPNL
    I'm sorry for not replying earlier, I'm subscibed to this topic but didn't get any notification, thought it was dead...

    ImgCoreOCL.log shows:
    OLD CARD (look at the date)
    2015-03-24 19:21:42.469> OpenCL initialization...
    2015-03-24 19:21:42.485> OpenCL : found platform NVIDIA CUDA, OpenCL Version : OpenCL 1.1 CUDA 6.5.45
    2015-03-24 19:21:42.506> OpenCL Device : GeForce GTX 285
    2015-03-24 19:21:42.506> OpenCL Driver Version : 341.44
    2015-03-24 19:21:42.506> OpenCL Compute Units : 30
    2015-03-24 19:21:42.607> OpenCL : Loading kernels
    2015-03-24 19:21:43.170> OpenCL : Loading kernels finished
    2015-03-24 19:21:43.170> OpenCL : Benchmarking
    2015-03-24 19:21:43.514> OpenCL : Initialization completed
    2015-03-24 19:21:43.514> OpenCL benchMark : 0.558944

    NEW CARD
    2015-03-31 22:59:17.701> OpenCL initialization...
    2015-03-31 22:59:17.904> OpenCL : found platform NVIDIA CUDA, OpenCL Version : OpenCL 1.1 CUDA 7.0.29
    2015-03-31 22:59:17.919> OpenCL Device : GeForce GTX 970
    2015-03-31 22:59:17.919> OpenCL Driver Version : 347.88
    2015-03-31 22:59:17.919> OpenCL Compute Units : 13
    2015-03-31 22:59:18.091> OpenCL : Loading kernels
    2015-03-31 22:59:18.435> OpenCL : Loading kernels finished
    2015-03-31 22:59:18.435> OpenCL : Benchmarking
    2015-03-31 22:59:18.576> OpenCL : Initialization completed
    2015-03-31 22:59:18.576> OpenCL benchMark : 0.209920

    I created a ticket an T.S is helping also, I've just sent them this:
    Thanks a lot for getting back at me. I've tested a few times as you suggested.
    My method was:
    1 Get a directory with 100 RAW images
    2 Reset all parameters
    3 Auto level all of them and measure the time it took

    I've set openCL to auto or never and closed/opened P1 with the applied setting

    The results are:
    Run 01, "auto" - 23 seconds
    set openCL to "never", close P1, open P1, reset all, auto level...
    Run 02, "never" - 20 seconds
    set openCL to "auto", close P1, open P1, reset all, auto level...
    Run 03, "auto" - 20 seconds
    reset all, auto level
    Run 04, "auto" - 21 seconds
    set openCL to "never", close P1, open P1, reset all, auto level...
    Run 05, "never" - 20 seconds
    reset all, auto level...
    Run 06, "never" - 22 seconds

    I'm a bit (understatement) disappointed in the difference, I hope you sympathise with that.
    I've installed GPU-Z and it's remarkable but it shows the CPU load is not reaching over 5%!
    That seems like a pointer to me?
    0
  • Christian Gruner
    What camera are the raw-files from?
    0
  • WPNL
    SONY A77-II
    0
  • Christian Gruner
    As it is a Bayer based camera you should have full acceleration. Please use our support, and they'll get you going.
    0
  • WPNL
    Thanks! I'm working on it with the technicians, as soon as it becomes clear what's happening I'll post the results.
    By the way, I've ordered a Radeon 290X so I can test this (and the GTX285 and GTX970) coming weekend weekend, head to head...
    0
  • Filthy Lucre
    [quote="NN635531119132438914UL" wrote:
    Thanks! I'm working on it with the technicians, as soon as it becomes clear what's happening I'll post the results.
    By the way, I've ordered a Radeon 290X so I can test this (and the GTX285 and GTX970) coming weekend weekend, head to head...

    We'll all be looking forward to your report!
    0
  • WPNL
    Hi guys,
    T.S. pointed out that OpenCL is only used in specific processes:
    "Applying auto levels does not use the GPU, hence no change. OpenCL processing is used in two areas in Capture One:
    Processing files from Raw into jpeg/tiff, etc. (OpenCL for Processing)
    Refresh rate of images on screen while performing adjustments (OpenCL for Display)
    The test I was describing earlier was to convert 100 files to jpeg or tiff from the process recipe tool."
    I thought it was used throughout the whole application, my misunderstanding... So I tested by processing 100 ARW into full-size JPG. It immediately became clear that there was a significant gain, see attached screenshots.

    1 - OpenCL set to "auto / auto", estimated time on start of the export was 1.55 minutes, and ended at 2.10 minutes.
    Look at the shape of the graphs, I was a bit sceptic, certainly because the threads of my CPU were not maxing as expected.
    http://phodograf.com/xfer/opencl_0010.png

    2 – OpenCL set to “never / neverâ€, estimated time on start of the export was 9.55 minutes, and ended at 4.00 minutes (exact, funny) in a BSOD… let’s not get into that… I’m pretty sure that it would have taken 10 minutes to complete
    Look at the shape of these graphs :
    http://phodograf.com/xfer/opencl_0030.png


    Finally, I still have the R9 290X on it’s way and just to be sure I’m going to repeat this on that card.
    I’m really curious 😊
    0
  • WPNL
    Test results updated (export 100ARW to Full size JPG):
    GTX970, 1.30m / 1.33m
    R9 290X, 1.30m / 1.32m

    This was not a professional perfectly controlled benchmark, just ran the batch and used a stopwatch (digital of course).
    I think there are many factors that can influence the result, some specs of my rig:
    CPU: Intel i7 875K overclocked to 3,99Ghz
    RAM: 16Gb 1600Mhz
    HD: Samsung 850 EVO 250Gb / WD 2Tb disk (location of the photos)
    0
  • Wolfram HĂślzel
    Thank you for the testresulats. Are there any difference during the work?
    0
  • WPNL
    I didn't work or test for a long time but one thing was remarkable. I had the 290X installed :
    When altering an image by using a slider (exposure for instance) the preview on the screen was changing real-time but after a second it would "change" into the final preview. When using sharpness it looked like the image was softened a bit after a second.
    The GTX doesn't have this, the preview stays as it is when "sliding"
    0
  • SFA
    [quote="NN635531119132438914UL" wrote:
    Test results updated (export 100ARW to Full size JPG):
    GTX970, 1.30m / 1.33m
    R9 290X, 1.30m / 1.32m

    This was not a professional perfectly controlled benchmark, just ran the batch and used a stopwatch (digital of course).
    I think there are many factors that can influence the result, some specs of my rig:
    CPU: Intel i7 875K overclocked to 3,99Ghz
    RAM: 16Gb 1600Mhz
    HD: Samsung 850 EVO 250Gb / WD 2Tb disk (location of the photos)


    The i7 875k is rated at about 5.5 at Passmark.

    My i7 3820QM is rated at about 8.5. It's already a 3 year old standard.

    I would guess that motherboard bus architecture around the 875 might not be the fastest there is by current standards. If so the Samsung 850 may not be able to perform as well as it could.

    I have a 500Gb 840 which performs well and a 1Tb mSata 840 that only gives half max rated speed because it's fitted in an older spec comms port. But the preformance the mSata drive offers is still adequate for my purposes.

    You may find that the system rather than the components is restricting performance - it's certainly worth checking if only to eliminate the possibility.

    HTh.


    Grant
    0
  • WPNL
    Grant,
    I'm aware of that, a mainboard /cpu upgrade is planned also but first I have to have budget 😊
    What I'm pretty sure of is that this investment was the quickest and most cost efficient way to faster processing and 4K compatibility.

    Anyway.. Of all guys with more up to date hardware than mine I've not yet found any concrete comparison or information about the initial subject...

    So... How long does it take on your rig?... So I know what I will be working for haha
    0

Post ist fĂźr Kommentare geschlossen.