List of compatible (and really useful) Opencl GPUs ?

photo by FA

30. September 2016 15:27

Hi Maikol,
Late 2015 iMac 5K GPU is not supported and not compatible with CO9.3 and I have turn OpenCL off as while generating previews, it was giving weird results.

Here is the list of supported GPUs

https://www.phaseone.com/en/Search/Arti ... nguageid=1

0

Wesley

30. September 2016 18:16

[quote="fatihayoglu" wrote:
Hi Maikol,
Late 2015 iMac 5K GPU is not supported and not compatible with CO9.3 and I have turn OpenCL off as while generating previews, it was giving weird results.

Here is the list of supported GPUs

https://www.phaseone.com/en/Search/Arti ... nguageid=1

5K iMac is supported. The list is not kept up to date.
All 1GB and higher, OpenCL GPU should work.

0

Christian Gruner

30. September 2016 18:18

[quote="fatihayoglu" wrote:
Hi Maikol,
Late 2015 iMac 5K GPU is not supported and not compatible with CO9.3 and I have turn OpenCL off as while generating previews, it was giving weird results.

Here is the list of supported GPUs

https://www.phaseone.com/en/Search/Arti ... nguageid=1

The same GPU runs fine in our office ? The GPU is not used for generating previews, so I am a little puzzled ?
Please contact our Support team.

0

Christian Gruner

30. September 2016 18:37

[quote="Wesley" wrote:
[quote="fatihayoglu" wrote:
Hi Maikol,
Late 2015 iMac 5K GPU is not supported and not compatible with CO9.3 and I have turn OpenCL off as while generating previews, it was giving weird results.

Here is the list of supported GPUs

https://www.phaseone.com/en/Search/Arti ... nguageid=1

5K iMac is supported. The list is not kept up to date.
All 1GB and higher, OpenCL GPU should work.

Over 1 gb and a benchmark over 0.80 (look in the CaptureICP.log on Mac or the ImgCoreExternal.log on Windows to see the number). 0.80 is pretty slow btw. Mac Pro D700 benchmarks to 0.11 pr adapter (as does the new AMD 480), and a AMD R9 Nano benchmarks to 0.055

0

Michael Seminatore

1. Oktober 2016 07:22

Thank you guys for your answers !

So the list in question is not up to date, that's what I had felt.

Also, Christian, is there somewhere I can find a list of those CaptureICP.log benchmark values you are talking about ?

All I seem to find on the web is this thread on another forum : https://www.tonymacx86.com/threads/750- ... ne.157179/.

I'd say you guys should post an up to date list of these values. 😉

0

photo by FA

1. Oktober 2016 16:19

Hi Christian,

Please see the topic

Also, if you have a chance to look case number 228053, you will see that the reply from Support team:

---------------------
Hi Faith,

This is not a bug, but rather a compatibility issue with the graphics card. More on this here: https://www.phaseone.com/en/Search/Arti ... nguageid=1

Kind regards,

Phase One Technical Support
---------------------

So, according to the Support team, my GPU is not compatible with CO, or is not supported. The same thing.

Also;
2016-10-01 17:14:07> OpenCL benchMark : 0.148000, this is just now from CaptureONeICP.log. So it is not that terrible I believe.

0

Eric Nepean

3. Oktober 2016 03:37

What a useful observation. From my 2015 27" iMac 17,1 (the middle model, with 4GHz i7, M390 GPU, 24GB RAM, using an external SSD)

2016-10-02 23:23:27> OpenCL : found platform Apple, OpenCL Version : OpenCL 1.2 (Jun 30 2016 20:18:53)
2016-10-02 23:23:27> OpenCL Device : AMD Radeon R9 M390 Compute Engine
2016-10-02 23:23:27> OpenCL Driver Version : 1.2 (Aug 29 2016 22:17:00)
2016-10-02 23:23:27> OpenCL Compute Units : 16
2016-10-02 23:23:27> OpenCL : Loading kernels
2016-10-02 23:23:27> OpenCL : Loading kernels finished
2016-10-02 23:23:27> OpenCL : Benchmarking
2016-10-02 23:23:27> OpenCL : Initialization completed
2016-10-02 23:23:27> OpenCL benchMark : 0.162652

The list of supported GPUs lists no GPUs at all for Macs for COP9 - just Intel HD Graphics. Which leaves users with no leg to stand on when the support team says "your GPU is not supported".

0

Michael Seminatore

4. Oktober 2016 15:11

I guess the real useful thing we'd also need to know before getting a new machine is this : in which areas of C1 will using OpenCl really be an improvement ?

- only for heavy batch processing ?

- When culling ? Importing ?

- When editing ? Will it make a real difference when moving sliders around or drawing a mask for example ?

- On exporting ?

By knowing where OpenCl is really used in C1 Pro 9, and thus deciding if it's really useful for their workflow, one could make a better informed decision when choosing the computer they'd use.

Cheers,
M

0

photo by FA

6. Oktober 2016 12:42

To be honest, first the PO team needs to make their minds up and decide whether they support certain GPUs or not or certain GpUs are compatible or not. Then we should know the benefit of using OpenCL. Otherwise that information will be a unuseful info.

0

Christian Gruner

6. Oktober 2016 13:33

[quote="Maikol" wrote:
I guess the real useful thing we'd also need to know before getting a new machine is this : in which areas of C1 will using OpenCl really be an improvement ?

- only for heavy batch processing ?

- When culling ? Importing ?

- When editing ? Will it make a real difference when moving sliders around or drawing a mask for example ?

- On exporting ?

By knowing where OpenCl is really used in C1 Pro 9, and thus deciding if it's really useful for their workflow, one could make a better informed decision when choosing the computer they'd use.

Cheers,
M

If you search the forum , you will find that I have answered that question several times before here.

Here are some of the posts:
viewtopic.php?f=61&t=21628&p=102289&hilit=opencl#p102289
viewtopic.php?f=61&t=24229&p=112785&hilit=opencl#p112785
viewtopic.php?f=61&t=24229&p=112785&hilit=opencl#p112737
viewtopic.php?f=62&t=22459&p=107945&hilit=opencl#p107945

In general, OpenCL will make a huge difference if the GPU is fast and capable enough, compared to the CPU.

0

Christian Gruner

6. Oktober 2016 13:37

[quote="fatihayoglu" wrote:
To be honest, first the PO team needs to make their minds up and decide whether they support certain GPUs or not or certain GpUs are compatible or not. Then we should know the benefit of using OpenCL. Otherwise that information will be a unuseful info.

As a rule of thumb, there is no whitelist, nor a model-specific blacklist.
In general, it is more up to the drivers than the specific hardware if the cards work properly in Capture One.
That also means that even new cards will (/should) work, as they are running the same drivers as previous cards.

0

Michael Seminatore

6. Oktober 2016 14:40

Well, as I'm now very close to buying that new mac (well my budget being limited it's going to be a used one actually), here are some more specific models and combinations that I'm considering, so maybe Christian or others can give us some feedback regarding opencl :

- Mac Pro 5.1 with 6 core @3.33Ghz and 1 AMD Radeon R9 280X

or

- Mac Pro 5.1 with 6 core @3.33Ghz and 2 AMD Radeon R9 280X

or

- MacBook Pro (2013 or 2014) 15" Retina i7 @ 2.8Ghz

Which one of these should be the faster all around for editing in Capture One ?

Would the R9 in the old mac pro (or even double R9) really bring some big improvement with openCl ? Enough to really overpower the macbook pro ?

Cheers,
Maikol

0

Michael Seminatore

6. Oktober 2016 14:45

Oops I think we wrote our posts at the same time Christian ! 😊

0

Christian Gruner

6. Oktober 2016 14:47

[quote="Maikol" wrote:
Which one of these should be the faster all around for editing in Capture One ?

The Mac Pro verson without a shadow of a doubt.

[quote="Maikol" wrote:
Would the R9 in the old mac pro (or even double R9) really bring some big improvement with openCl ? Enough to really overpower the macbook pro ?

Absolutely, in terms of processing and adjustmentspeed, however, there is one big catch. The AMD cards are likely reflashed firmware wise to allow them to work on a Mac, and requires special drivers. If that is the case, you are a little on your own. We recently did a fix for a GTX 980 TI in a mac pro, so we try to support our customers using these unsupported (OSX wise) combinations, but they have a risk of breaking somewhere in process.

0

Michael Seminatore

6. Oktober 2016 15:02

[quote="Christian Gruner" wrote:
[quote="Maikol" wrote:
Which one of these should be the faster all around for editing in Capture One ?

The Mac Pro verson without a shadow of a doubt.
[quote="Maikol" wrote:
Would the R9 in the old mac pro (or even double R9) really bring some big improvement with openCl ? Enough to really overpower the macbook pro ?

Absolutely, in terms of processing and adjustmentspeed, however, there is one big catch. The AMD cards are likely reflashed firmware wise to allow them to work on a Mac, and requires special drivers. If that is the case, you are a little on your own. We recently did a fix for a GTX 980 TI in a mac pro, so we try to support our customers using these unsupported (OSX wise) combinations, but they have a risk of breaking somewhere in process.

Thanks Christian !

Hmm yes the R9 280x are indeed flashed PC cards. I didn't know they would require special drivers ! Maybe it is too risky then ?

Would you recommend one card in particular for maximum openCl power in a mac pro 5.1 ? Like that GTX 980 TI you did a fix for ?

0

Permanently deleted user

6. Oktober 2016 21:59

I am running a flashed (and slightly overclocked) AMD Radeon HD 7950 in a MacPro 5.1 , 6-core 3.33 GHz.
No glitches or problems or additional drivers - under OS X 10.11.6 with C1 9.0.3 Pro - smooth performance.
I had been considering an R9 card before but decided against it because I feared it's power consumption might overstrain the MacPro's power supply. The guy I purchased my card from had measured different cards in terms of current drain - and I did not want to risk frying my MacPro.

Performance is from my point of view @ 30" and 2560 x 1600 resolution absolutely o.k. - and noticeably faster than with my MacBook Pro 15" retina - bench logs below:

MacPro5,1 (Mid 2010), 6-Core Intel Xeon 3,33 GHz - NEC PA302W @ 2560 x 1600 @ 60 Hz:
> Logging is now active.
> CPU: GenuineIntel [Intel(R) Xeon(R) CPU W3680 @ 3.33GHz]
> CPU features: MMX, SSE, SSE2, SSE3, SSSE3, SSE41, SSE42, CX8, RDTSCP, POPCNT
> OpenCL initialization...
> OpenCL : found platform Apple, OpenCL Version : OpenCL 1.2 (Jun 30 2016 20:18:53)
> OpenCL Device : AMD Radeon HD 7950 Compute Engine
> OpenCL Driver Version : 1.2 (Aug 29 2016 22:17:00)
> OpenCL Compute Units : 28
> OpenCL : Loading kernels
> OpenCL : Loading kernels finished
> OpenCL : Benchmarking
> OpenCL : Initialization completed
> OpenCL benchMark : 0.123876
> Exiting
> Shutting down

MBP15â€œ retina 11.3 (Mid 2014), i7, 2,8 GHz:
> Logging is now active.
> CPU: GenuineIntel [Intel(R) Core(TM) i7-4980HQ CPU @ 2.80GHz]
> CPU features: MMX, SSE, SSE2, SSE3, SSSE3, SSE41, SSE42, AVX, AVX2, FMA, F16C, MOVBE, CX8, RDRAND, RDTSCP
> CPU features: POPCNT, BMI1, BMI2, LZCNT
> OpenCL initialization...
> OpenCL : found platform Apple, OpenCL Version : OpenCL 1.2 (Jun 30 2016 20:18:53)
> OpenCL Device : Iris Pro
> OpenCL Driver Version : 1.2(Aug 29 2016 22:20:39)
> OpenCL Compute Units : 40
> OpenCL : Loading kernels
> OpenCL : Loading kernels finished
> OpenCL : Benchmarking
> OpenCL : Initialization completed
> OpenCL benchMark : 0.928320
> OpenCL Device : GeForce GT 750M
> OpenCL Driver Version : 10.10.13 310.42.25f01
> OpenCL Compute Units : 2
> OpenCL : Loading kernels
> OpenCL : Loading kernels finished
> OpenCL : Benchmarking
> OpenCL : Initialization completed
> OpenCL benchMark : 1.591360
> Exiting
> Shutting down

Hope that helps,

Oliver

0

Michael Seminatore

6. Oktober 2016 22:45

Oliver, thank you so much for having taken the time for this!

Yes it does help, as this is precisely the mac pro I'm considering, and the power consumption of the R9 is also something I'm concerned about. Add to that what Christian wrote about, and this R9 route might very well be a bad idea...

In the meantime I found the thread about the 980ti bug that's been solved by the C1 team (here it is : viewtopic.php?f=61&t=22218&hilit=980ti ), and as this is now solved, it also looks like a nice option, as this card seems to really be a beast in opencl.

But then there's also the power consumption question !

Ooooh bugger how complicated this all is ! 😜 😁

0

Wolfgang Glathe

21. Oktober 2016 03:28

Hi,

I don't know if I'm too late for you with my answer, but maybe it is still helpful.
I bought my Mac Pro 4.1 (Early 2009) late 2009 😊 - dual 2.26, Geforce GT120. A few years ago my first SSD, later a Sonnet Tempo SSD PCI Express card. Over the years my very reliable all day working horse - for me my best machine ever! This year I started thinking about more powerful hardware and choosing C1P for my photography. A new Mac or "pimp-my-great-2009-cMP"? After weeks of reeding, calculating and comparing I decided to stay with my cMP and get him a full fresh-up: firmware-upgrade to a Mac Pro 5.1, new CPUs (2x 6-core X5690 3.47 GHz), a new GeForce GTX 980 Ti from MacVidCards, 48GB of now supported faster RAM, a very fast M.2-SSD from Samsung on a PCIe-Adapter and a new Apple-WLAN/Bluetooth-module from OSXWifi. I got it all together for less than a new i7-iMac (took a little time to wait for good prices). At this time Phase One fixed the issues with C1P and the GTX 980 TI (thanks!) and now I am a happy C1P-user on a (for me) very snappy and fast, still very reliable machine!

Now to the requested numbers: 😊
CaptureOneICP.log reports:

> Logging is now active.
> CPU: GenuineIntel [Intel(R) Xeon(R) CPU X5690 @ 3.47GHz]
> CPU features: MMX, SSE, SSE2, SSE3, SSSE3, SSE41, SSE42, CX8, RDTSCP, POPCNT
> OpenCL initialization...
> OpenCL : found platform Apple, OpenCL Version : OpenCL 1.2 (Aug 10 2016 17:16:39)
> OpenCL Device : GeForce GTX 980 Ti
> OpenCL Driver Version : 10.13.65 367.15.10.05f01
> OpenCL Compute Units : 22
> OpenCL : Loading kernels
> OpenCL : Loading kernels finished
> OpenCL : Benchmarking
> OpenCL : Initialization completed
> OpenCL benchMark : 0.071200

Power consumption:
I did a CPU stress test (System Load von Marcel Bresink) with 99% load on all 24 virtual cores and some "heavy" OpenCL-/ CUDA-benchmark- and demo-tools (I just forgot the names) for an hour - no problem! No hiccups, all temperature-sensors reported no problems, the fans just went a little higher, thats it.

Drivers for the GTX 980 TI:
Nvidia is very fast in providing new drivers for new macOS-versions. A few days after every macOS-release (including betas) Nvidias software asks me for downloading a new driver-version.

So after the first months of using my â€žnewâ€œ old Mac Pro (now with macOS Sierra) everything is very stable and feels very fast. I have no real comparison except the OpenCL benchmark number above.
Maybe this is helpful: I processed 104 Olympus OM-D RAWs with different adjustments and layers to full size JPEG. It took:
226 sec - only with CPU (2,17s for each)
88 sec - with GPU (0,85s for each)
So the GTX 980 Ti does this job 2,6 times faster than my both 6-/12-core 3,47GHz XEON CPUs.

I hope, there is something helpful for you.

Greetings, Wolle

0

Michael Seminatore

21. Oktober 2016 06:21

Hi Wolle !

You are not too late, thank you very much indeed for this very useful information !

I'm sure it will also help other users wondering about these questions.

The 2.6 x improvement in rendering time is indeed impressive. 🤓

How would you quantify the improvement in snappiness while using C1 (sliders, browsing, etc.) compared to this ?

Thanks again and all the best !
M

0

Permanently deleted user

21. Oktober 2016 13:53

This sounds nice, thank you!

I have a Sapphire HD 7950 in my Mac Pro 5,1 and the rendering times are similarly impressive.

Where I have problems though is that Capture One's sliders aren't as fluid as I'd wish and rotating images is very jaggy, not fluid at all. And this is want makes me consider an even more powerful graphics card IF this would solve these problems.

So here is my question, Wolle: how is C1's responsiveness while editing images. Do you feel any hesitation, locking, jagginess?

Thank you!

0

Wolfgang Glathe

22. Oktober 2016 04:08

Hi,

it is difficult for me, to give a "real" impression of how snappy my cMP is with or without OpenCL. Most sliders for things like color, HDR, sharpness, clarity, structure, curves etc. react nearly instantaneous on my Olympus OM-D RAWs with OpenCL ON. If I would think of "perfect fluid" is 25 fps, most of the sliders give me a feeling of 18-23fps with sometimes a small delay of maybe 1/4 sec behind my movement. Rotating, keystone correction etc. are a little bit slower - something like 8 fps (guidelines displayed), 13 fps (no guidelines displayed when "scrolling a slider" with the touchpad).

Browsing:
I'm not very long with C1P, so at the moment one catalog has around 6500 pictures. Browsing is good - again not absolutely smooth, but just a bit cosmetic difference from that. (OK, browsing 63000 private pictures in Fotos is very fluid on this cMP, but I think that is not a fair comparison 😉)

That's all very imprecise, no real scientific "hard" numbers. For me rotating images etc. doesn't feel jaggy. Not really smooth (like rotating a small JPEG in Pages 😉) but very good to work, at no time feeling like a handicap for me.

I think about capturing a video of the monitor when doing such rotations etc. and make a video-analysis: counting the different displayed rotation-states and with given fps of the video calculation the frames per second of the C1P-rotation.
OK, not enough time for such thingsâ€¦ 😊

Greetings,
Wolle

PS: My displays: 30" (2560 x 1600) and 23" (1920 x 1200)
PPS: I'm sorry, my English is not so goodâ€¦ 😊

0

Michael Seminatore

22. Oktober 2016 14:52

Thanks again Wolle !

And your english is perfectly fine. 😉

Cheers,
M

0

Wolfgang Glathe

23. Oktober 2016 23:19

PPPS: Barefeats just tested some new and old graphic cards from Nvidia and AMD in a cMP (2010), also the GTX 980 Ti. A lot of OpenGL-tests but also CUDA and OpenCL - mixed, so not perfectly comparable. The results are really impressive for the GTX 980 Ti. Maybe it is a bit interesting:

http://barefeats.com/sierra_rx480.html

0

Michael Seminatore

26. Oktober 2016 22:05

Hi everybody,

As a follow up, just wanted to let you guys what I eventually did, in case it might help someone else :

- I got a second hand 2009 MacPro 4.1 for around 500â‚¬,

- Put some SSD's in there and installed El Capitan,

- Replaced the 8 Go @ 1066Mhz RAM with 32 Go @ 1333Mhz.

After all the Q&A here, I found a Sapphire HD7970 Vapor X (equivalent to the R9 280X) for pretty cheap (120â‚¬!), so I thought I would give it a try, and sell it back if it wouldn't work.

It's an unflashed PC card, I installed it this evening with two 6 pins to 8 pins PCI Express power adapters, and it worked immediately, with no problem so far. Just got no Apple boot screen, but I don't need it for now.

Just checked in the CaptureOne Log, and found that the OpenCl benchmarks is 0.098420 !!!

Sounds not bad ! 😎

Will see if it works well on the long run, but so far my old 2.66Ghz Quad Core seems to be flying ! 😄

Next step will be to replace the processor with the hexacore @ 3.33 Ghz, to get the full potential of the new RAM (it's for now limited to 16Go @ 1066 MHz by the W3520 processor).

All in all, this will be an under 1000â‚¬ pretty serious powerhouse ! 💡

Hope this helps !

Cheers
M

0

Permanently deleted user

27. Oktober 2016 00:11

[quote="WolleG" wrote:
PPPS: Barefeats just tested some new and old graphic cards from Nvidia and AMD in a cMP (2010), also the GTX 980 Ti. A lot of OpenGL-tests but also CUDA and OpenCL - mixed, so not perfectly comparable. The results are really impressive for the GTX 980 Ti. Maybe it is a bit interesting:

http://barefeats.com/sierra_rx480.html

This looks great!

The GTX 980 Ti is doing very well in this report and it looks like there is no need to wait for 1080 card to get flashed.

0

Kommentare