OpenCL and Macbook Pro 15 Zoll, Mitte 2012
After testing C1 now for 3 Shootings i found out, that my GT650M and OpenCl not working together.
Why? This Computer is not too Old. C1 had the Version 7.2.2. 7.2.3 does drive crazy so i downgrade.
What is needed to work with OpenCl on this Maschine?
UPDATE:
Started an OpenCL Benchmark and how i see, my GT650M do a better Job than my other Devices.
Why? This Computer is not too Old. C1 had the Version 7.2.2. 7.2.3 does drive crazy so i downgrade.
2014-07-26 11:10:08> Logging is now active.
2014-07-26 11:10:08> OpenCL initialization...
2014-07-26 11:10:08> OpenCL : found platform Apple, OpenCL Version : OpenCL 1.2 (Jun 3 2014 12:43:41)
2014-07-26 11:10:08> OpenCL Device : GeForce GT 650M
2014-07-26 11:10:08> OpenCL Driver Version : 8.26.26 310.40.45f01
2014-07-26 11:10:08> OpenCL Compute Units : 2
2014-07-26 11:10:08> OpenCL device not used due to insufficient capabilities
2014-07-26 11:10:08> Shutting down
2014-07-26 11:10:08> OpenCL allocations : 0 bytes
What is needed to work with OpenCl on this Maschine?
UPDATE:
Started an OpenCL Benchmark and how i see, my GT650M do a better Job than my other Devices.
Mac & i OpenCL Benchmark 1.0 (1)
2014-07-26 09:36:58 +0000
MacBookPro9,1
(
{
deviceAdressBits = 64;
deviceID = "<00440201 00000000>";
deviceName = "HD Graphics 4000";
deviceType = 4;
deviceVersion = "OpenCL 1.2 ";
globalMemSize = 1073741824;
localMemSize = 65536;
maxClockFrequency = 1200;
maxComputeUnits = 16;
maxWorkGroupSize = 512;
preferedVectorWidthFloat = 1;
},
{
deviceAdressBits = 32;
deviceID = "<00270201 00000000>";
deviceName = "GeForce GT 650M";
deviceType = 4;
deviceVersion = "OpenCL 1.2 ";
globalMemSize = 536870912;
localMemSize = 49152;
maxClockFrequency = 774;
maxComputeUnits = 2;
maxWorkGroupSize = 1024;
preferedVectorWidthFloat = 1;
},
{
deviceAdressBits = 64;
deviceID = "<ffffffff 00000000>";
deviceName = "Intel(R) Core(TM) i7-3615QM CPU @ 2.30GHz";
deviceType = 2;
deviceVersion = "OpenCL 1.2 ";
globalMemSize = 17179869184;
localMemSize = 32768;
maxClockFrequency = 2300;
maxComputeUnits = 8;
maxWorkGroupSize = 1024;
preferedVectorWidthFloat = 4;
}
)
HD Graphics 4000, VW 1
4096 x 4096 pixel in 1004.11 ms
16.71 megapixel per second
70.18 GFLOPS
HD Graphics 4000, VW 4
4096 x 4096 pixel in 1299.08 ms
12.91 megapixel per second
54.24 GFLOPS
HD Graphics 4000, VW 8
4096 x 4096 pixel in 1386.18 ms
12.10 megapixel per second
50.83 GFLOPS
GeForce GT 650M, VW 1
4096 x 4096 pixel in 462.39 ms
36.28 megapixel per second
152.39 GFLOPS
GeForce GT 650M, VW 4
4096 x 4096 pixel in 612.76 ms
27.38 megapixel per second
115.00 GFLOPS
GeForce GT 650M, VW 8
4096 x 4096 pixel in 738.67 ms
22.71 megapixel per second
95.39 GFLOPS
Intel(R) Core(TM) i7-3615QM CPU @ 2.30GHz, VW 1
4096 x 4096 pixel in 4587.50 ms
3.66 megapixel per second
15.36 GFLOPS
Intel(R) Core(TM) i7-3615QM CPU @ 2.30GHz, VW 4
4096 x 4096 pixel in 1144.57 ms
14.66 megapixel per second
61.56 GFLOPS
Intel(R) Core(TM) i7-3615QM CPU @ 2.30GHz, VW 8
4096 x 4096 pixel in 861.06 ms
19.48 megapixel per second
81.83 GFLOPS
All OpenCL devices together, VW (
4,
4,
8
)
4096 x 4096 pixel in 296.22 ms
56.64 megapixel per second
237.88 GFLOPS
done.
0
-
The 650M is basically too slow for CO. Compared to the normal CPU in a mid 2012 mbp, the GPU would actually be slower.
We primarily disable the slow GPU due to stability and performance. Slow cards needs compromises in the code, that in-turn can cause stability-issues along with performance issues on faster cards.0 -
[quote="Christian Gruner" wrote:
The 650M is basically too slow for CO. Compared to the normal CPU in a mid 2012 mbp, the GPU would actually be slower.
We primarily disable the slow GPU due to stability and performance. Slow cards needs compromises in the code, that in-turn can cause stability-issues along with performance issues on faster cards.
Well, when i see it right my Nvidia is faster than my I7 CPUGeForce GT 650M, VW 1
4096 x 4096 pixel in 462.39 ms
36.28 megapixel per second
152.39 GFLOPSIntel(R) Core(TM) i7-3615QM CPU @ 2.30GHz, VW 1
4096 x 4096 pixel in 4587.50 ms
3.66 megapixel per second
15.36 GFLOPS
or i am wrong?0 -
Well, with version 7.2.3 i got the following console message 2014-07-28 18:08:08> Logging is now active.
2014-07-28 18:08:08> OpenCL initialization...
2014-07-28 18:08:08> OpenCL : found platform Apple, OpenCL Version : OpenCL 1.2 (Jun 3 2014 12:43:41)
2014-07-28 18:08:08> OpenCL Device : HD Graphics 4000
2014-07-28 18:08:08> OpenCL Driver Version : 1.2(Jun 9 2014 13:24:09)
2014-07-28 18:08:08> OpenCL Compute Units : 16
2014-07-28 18:08:08> OpenCL : Building kernels
2014-07-28 18:09:21> OpenCL : Benchmarking
2014-07-28 18:09:23> OpenCL : Initialization completed
2014-07-28 18:09:23> OpenCL benchMark : 2.777920
2014-07-28 18:09:23> OpenCL Device : GeForce GT 650M
2014-07-28 18:09:23> OpenCL Driver Version : 8.26.26 310.40.45f01
2014-07-28 18:09:23> OpenCL Compute Units : 2
so, it look slike that my GFX card is not too slow to working with OpenCL 😉
I test it out.0 -
TAO, what specific MBP do you use? There were models with both 512 MB vram and 1 GB vram back in 2012. Capture One only runs properly with 1 GB vram or more, so the “automatic†setting in 7.2.2 disabled OpenCL when this is not met. With C1 7.2.3 though there’s a bug where this startup checking of requirements isn’t made, resulting in potential crashes while in use as well as processing errors in final JPEG/TIFF files.
Cheers,
-Sascha0 -
[quote="nggalai" wrote:
TAO, what specific MBP do you use? There were models with both 512 MB vram and 1 GB vram back in 2012. Capture One only runs properly with 1 GB vram or more, so the “automatic†setting in 7.2.2 disabled OpenCL when this is not met. With C1 7.2.3 though there’s a bug where this startup checking of requirements isn’t made, resulting in potential crashes while in use as well as processing errors in final JPEG/TIFF files.
Cheers,
-Sascha
I have The 512mb Version. At The Moment i have no Problems with crashes.0 -
[quote="TAOGde" wrote:
I have The 512mb Version. At The Moment i have no Problems with crashes.
It depends a lot on circumstances: file size, resolution, number of images, steps made while developing them, layers / local adjustments, how often you restart C1 during work on a session, do you use the loupe, and so on. As it’s not very predictable OpenCL has been disabled on “automatic†for a long time with little video memory (what you observed with 7.2.2), with 7.2.3 this startup check was removed (by accident? bug?).
But as long as it works for you, enjoy! 😊
Cheers,
-Sascha0
Post is closed for comments.
Comments
6 comments