ROBERT BARUCH  A rapidshare download? Really?

Could this be released on, say, SourceForge? And under, say, an open source license?

I would love to be able to see the source code so that I could learn from it. I haven't yet been able to download the zip because "there aren't available free user slots" on rapidshare :(  
(Version V 2.0)

N00B32  Hi,

I added a table for better comparison of the OpenCL benchmarks:

http://wiki.osx86project.org/wiki/index.php/OpenCL

What would be the best benchmark for evaluation of OpenCL performance? Galaxies?

@mitch_de:

could you provide in this standard benchmark a build number visible while benchmarking (better comparison)?

Thanks

Jason   
(Version V 2.0)

MITCH_DE  from dev:

Ibuild an new Galaxis V2 32K. Apple updated OpenCL API usage , perhaps helps ATI to run ? But i think ATI will need really 10.6.2 update.

Others (NVIDIA) will have a bit higher GPU GigaFlops with that version!

Be patient in CPU modes - CPU has much work to compute the 32000 stars. GPU runs much faster here but also has less that 20 FPS (GTX285) 9 FPS 8800GT.

Link (will be official if some report : work)

http://rapidshare.com/files/286234291/Galaxies_32K_V2.zip

Users with Mobile Nvidia should better use the also new V2 8K version.

GigaFlops for that mobile GPUs same as with 32K version, but more FPS especially in CPU mode

on lowend C2D mobile

http://rapidshare.com/files/286235157/Galaxies_8K_V2.zip  
(Version V 2.0)

STARKNIGHTS  Still not working on my MacPro with HD4870.

Console outputs:

Build Log for Device[0]:

then after 15 seconds or so:

setupComputeDevices() failed: -7

and then I'll have to force quit the app.

Hope the X.6.1 update will fix the ati drivers.

PS.

Schöne Grüße aus Bonn ;)  
(Version V 1.0)

BEN PAPWORTH  This is fantastic.

I'm lucky enough to own a 17" MacBook Pro - and it's very interesting that I can use both GPUs at once (9400M as renderer and 9600M as compute engine), as I didn't think this would be possible.

9600M = 46 GFLOPS

9400M = 21 GFLOPS

Hybrid = 25 GFLOPS

Given the speed of calculations, it looks like hybrid CPU+GPU must be using 9400M for GPU compute (or hybrid is not nearly as efficient). It would be a great addition if the benchmark gave two hybrid options (one for each GPU), or identified which GPU it was using.

On another note, do you think it's possible to make Galaxies use multiple GPUs for compute? I realise that this is a big ask - as you've already indicated that the 9400M GPU can only process 4K elements...

Keep up the good work!  
(Version V 1.0)

MITCH_DE  by uploader:

Old OpenCL "Benchmark" was removed and now its replaced by Galaxies OpenCL Demo-Benchmark V1.0

- Galaxies is an much more real world OpenCL Demo and Benchmark

- no more an terminal application - OpenGL fullscreen support

- computes an universum star szene on CPU, GPU and Hybrid (CPu+GPUs) - switch with key s between that modes

- does measure computed GigaFlops and sure you also will see the speed in the moving speed of the stars - on an MacPro 1,1 an NVDIA GTX 285 is about 5 times faster - soon up to 8 times than CPUs

- do not use or report OpenCL V0.25 results anymore, the new GigaFlop results are much, much more real for OpenCL speed

- results, questions & answers you can find on the home forum:

http://www.insanelymac.com/forum/index.php?showtopic=182874

- happy benching !  
(Version V 1.0)

ê¹€ 승균  ...........................................................

.................. OpenCL Bench V 0.25 by mitch ...........

...... C2D 3GHz = 12 sec vs Nvidia 9600GT = 0,93 sec ......

... time results are not comparable to older version! .....

...........................................................

Number of OpenCL devices found: 5

OpenCL Device # 0 = GeForce GT 120

Device 0 is an: GPU with max. 1400 MHz and 32 units/cores

Now computing - please be patient....

time used: 1.629 seconds

OpenCL Device # 1 = GeForce GT 120

Device 1 is an: GPU with max. 1400 MHz and 32 units/cores

Now computing - please be patient....

time used: 3.984 seconds

OpenCL Device # 2 = GeForce GT 120

Device 2 is an: GPU with max. 1400 MHz and 32 units/cores

Now computing - please be patient....

time used: 1.743 seconds

OpenCL Device # 3 = GeForce GT 120

Device 3 is an: GPU with max. 1400 MHz and 32 units/cores

Now computing - please be patient....

time used: 1.636 seconds

Error: clGetDeviceInfo CL_DEVICE_NAME

ERROR NUMBER = -33

BUFFER =GeForce GT 120

When testing CPU, i got this error message.

Program is tested on MAC PRO (Quad cores)x2 with GeForece GT 120x4  
(Version V025)

SOMMOP87  Very good speed.  
(Version V025)

MINORGROOVE  Nice speedup on my early 2008 MacBook Pro 2.6 Hz machine with GeForce 8800M GT.

...........................................................

.................. OpenCL Bench V 0.25 by mitch ...........

...... C2D 3GHz = 12 sec vs Nvidia 9600GT = 0,93 sec ......

... time results are not comparable to older version! .....

...........................................................

Number of OpenCL devices found: 2

OpenCL Device # 0 = GeForce 8600M GT

Device 0 is an: GPU with max. 940 MHz and 32 units/cores

Now computing - please be patient....

time used: 2.958 seconds

OpenCL Device # 1 = Intel(R) Core(TM)2 Duo CPU T9500 @ 2.60GHz

Device 1 is an: CPU with max. 2600 MHz and 2 units/cores

Now computing - please be patient....

time used: 14.791 seconds

Now checking if results are valid - please be patient....

:) Validate test passed - GPU results=CPU results :)  
(Version V025)

MITCH_DE  Comment from the develeoper:

ATI is working on the problem that also near all other OpenCL Apple Demos (like GALXIES, qJulia,...)fail on ATI 4850 + ATI 4870.

They found already some bugs and will fix them with 10.6.1 /10.6.2 or an ATI Driver/Framework update seperately.  
(Version V025)