Jump to content

OpenCL Oceanwave & Bandwidth Bench - 07. March 2013

OpenCL AMD NVIDIA

  • Please log in to reply
343 replies to this topic

#121
mitch_de

mitch_de

    InsanelyMacaholic

  • Retired
  • 2,902 posts
  • Gender:Male
  • Location:Stuttgart / Germany
First: OpenCL always detects at least one OpenCL device - the cpu.
The gpu maybe NOT detected if the gpu isnt supported in general (like ATI 3xxx gpu or Geforce 7xxx) OR the gpu is unknown by the Apple drivers and must be patched for OpenCL. OpenGL may work normal, even on unpatched drivers.

"Cool, not sure how hard that will be for you to code. Might you have any thoughts as to why some parts of OSX see the 8800GT and others see the 5770?"
OpenCL is an special thing in this case compared to OpenGL.
The OpenCL code must be edited for using not the standard (first in the OpenCL info list (log of the app).

#122
diskeeper

diskeeper

    InsanelyMac Protégé

  • Members
  • PipPip
  • 58 posts
  • Gender:Male

Now can you do a comparison to the MP3,1 value?


MacPro3,1 : 615,34

MacPro5,1 : 31,45

#123
oswaldini

oswaldini

    InsanelyMac Geek

  • Members
  • PipPipPip
  • 144 posts
  • Gender:Male
  • Location:Cracow, Poland

MacPro3,1 : 615,34

MacPro5,1 : 31,45


Ohh come on guys ! Just simply edit AppleGraphicsPowerManagament.kext will fix it ! You need add your DevID to plist.

#124
Diwad

Diwad

    InsanelyMac Geek

  • Members
  • PipPipPip
  • 120 posts
Yeah, editing APGM will give you much more GPU power plus better gpu power management I think. It should eat less power when you add your ID to APGM.

Simanko oswaldini;)

#125
mitch_de

mitch_de

    InsanelyMacaholic

  • Retired
  • 2,902 posts
  • Gender:Male
  • Location:Stuttgart / Germany
NEW Version V1.2
- error handling for non OpenCL gpus (cant do OpenCL at all) OR OpenCL errors by OpenCL driver failures (Fermi with unpatched driver). Users get an messagebox insted of unknown fps results
- updated results list
- no changes in the main OpenCL bench
Sorry, but no option to select gpu for OpenCL - very complex Apple code to add that feature.

#126
Peter_sm

Peter_sm

    InsanelyMac Sage

  • Members
  • PipPipPipPipPip
  • 281 posts
When running this first I got about 14fps..... then I modified the AppleGraphicsPowerManagement.kext, now I got 65fps :unsure: But I don't know is good enough, see below.

I can play MKV (high bitrate) better in Windows7 than I can in OSX. But that perhaps is that I installed codecs in W7, do I need some codecs in OSX also ????


HACKINTOSH  MacBookPro6,1 Mac OS X 10.7.2 Intel(R) Core(TM) i7 CPU		Q 720  @ 1.60GHz 1596 MHzGPU			  GeForce GT 230M	 1100 MHz   _FPS= 65.97


#127
mitch_de

mitch_de

    InsanelyMacaholic

  • Retired
  • 2,902 posts
  • Gender:Male
  • Location:Stuttgart / Germany
Yep, decoding video depends on the codec softwarespeed. MKV is an container - codec can be everything, but most MP4 or h.264 :D
On windows some gpu drivers accererate videodecoding, on OS X only some do that and only if the software uses Apple video decoding hw accell framework. Normally doesnt matter (beside higher cpu loads) but may be a problem for slower cpus.

#128
mitch_de

mitch_de

    InsanelyMacaholic

  • Retired
  • 2,902 posts
  • Gender:Male
  • Location:Stuttgart / Germany
Added LUXMARK (OpemCL Benchmark) info + dl Link + result link on first posting.
LUXMARK can bench ALL OpenCL gpus - if you have more than one (and both not lowend) , maybe interesting.
It has much higher CPU usage (beside OpenCL) / load so cpu speed does matter compared to OceanWave with very less cpu loads.
To get close to the luxmark HEROs you need 8 * GTX 580 for results around 70000! :)

Attached Files



#129
HFW

HFW

    InsanelyMac Protégé

  • Members
  • PipPip
  • 84 posts
  • Gender:Male
  • Location:England
HACKINTOSH MacPro5,1 Mac OS X 10.7.2 Intel® Core™ i7 CPU 920 @ 2.67GHz 3500 MHz GPU ATI Radeon Barts XT Prototype 930 MHz FPS= 386.16
(AMD Radeon HD 6870)

LuxMark:

Attached File  Screen_Shot_2011_11_21_at_15.47.29.png   418.96KB   17 downloads

8 * GTX 580? who needs that much power? o.o

#130
mitch_de

mitch_de

    InsanelyMacaholic

  • Retired
  • 2,902 posts
  • Gender:Male
  • Location:Stuttgart / Germany
Added many submitted FPS results in the list on first posting.
THANKS.
If you submit very different FPS (like 25 and 80 fps) please comment what you have done (Mac modell in .plist, changes in AGPM of the gpu,...)

With that many results we can see that OceanWave produced valid results, like same GPU type much differnt CPU - same result. :(
HACKINTOSH MacPro3,1 Mac OS X 10.7.2 CPU Q9400 @ 3400 MHz GPU GeForce 9800 GT 1500 MHz FPS= 98.68
HACKINTOSH MacPro3,1 Mac OS X 10.7.2 CPU Q6600 @ 2997 MHz GPU GeForce 9800 GT 1500 MHz FPS= 98.68
If your result is much slower to similar gpu check if your gpu may run in slow mode (1/3 - 1/2 gpu of max. clock) even the gpu has lot of work. In such cases the OpemGL performance will be also much lower than expected. Has to do with AGPM settings for that gpu / mac modell type.

#131
jm_1990

jm_1990

    InsanelyMac Protégé

  • Members
  • Pip
  • 29 posts
  • Gender:Male
  • Location:Argentina
Motherboard: ASUS P8Z68-V PRO
BIOS: 1101 (All default settings. No Over-clocking, No XMP)
CPU: Intel Core i7 2600K 3.4 GHz
GPU: nVidia GeForce GTX 560 Ti (Gigabyte GV-N560OC-1GI 900 MHz 1024 MB)
RAM: 16 GB (4 x 4 GB) (G.Skill Sniper 1600 MHz)
OS: Mac OS X Lion 10.7.2
Boot: Chameleon 2.1 R1713
SMBios: Mac Mini Server (Mid 2011)
Kexts Installed: FakeSMC, ALC8xxHDA, HDAEnabler892, AppleHDA (10.6.2)

I'm not using any DSDT or similar. Vainilla system. AppleIntelCPUPowerManagement patched with SpeedStepper 1.2. Added my device ID in NVDAGF100Hal, patched GeForceGLDriver and libclh.dylib for OpenCL support. Edited AppleGraphicsPowerManagement with this:

   <key>Macmini5,3</key>
   <dict>
    <key>Vendor10deDevice1200</key>
    <dict>
	 <key>BoostPState</key>
	 <array>
	  <integer>0</integer>
	  <integer>1</integer>
	  <integer>2</integer>
	  <integer>3</integer>
	 </array>
	 <key>BoostTime</key>
	 <array>
	  <integer>3</integer>
	  <integer>3</integer>
	  <integer>3</integer>
	  <integer>3</integer>
	 </array>
	 <key>Heuristic</key>
	 <dict>
	  <key>ID</key>
	  <integer>0</integer>
	  <key>IdleInterval</key>
	  <integer>250</integer>
	  <key>SensorOption</key>
	  <integer>1</integer>
	  <key>SensorSampleRate</key>
	  <integer>4</integer>
	  <key>TargetCount</key>
	  <integer>5</integer>
	  <key>Threshold_High</key>
	  <array>
	   <integer>60</integer>
	   <integer>60</integer>
	   <integer>60</integer>
	   <integer>100</integer>
	  </array>
	  <key>Threshold_Low</key>
	  <array>
	   <integer>0</integer>
	   <integer>75</integer>
	   <integer>90</integer>
	   <integer>98</integer>
	  </array>
	 </dict>
	 <key>LogControl</key>
	 <integer>1</integer>
	 <key>control-id</key>
	 <integer>17</integer>
    </dict>
    <key>LogControl</key>
    <integer>1</integer>
    <key>default-control-id</key>
    <integer>17</integer>
   </dict>

AGPM works fine. Stays in 2 most of the time when I'm using the computer. Goes 0 when I run this test. And I get this:

Posted Image


But according the first page I should get about 420 fps with this GPU and CPU. Am I doing something wrong? Thanks



#132
mitch_de

mitch_de

    InsanelyMacaholic

  • Retired
  • 2,902 posts
  • Gender:Male
  • Location:Stuttgart / Germany
You can try to run the bench a few times. The faster the gpu the more little diff in the results.
Your result is OK for that card. Other, faster results may come from little more systembus/pcie speed and/or little faster cpu andor little faster VRAM speed/ GPU {censored}.
AGPM State 0 (when run this test) is perfect - runs as fastest gpu clock.

#133
jm_1990

jm_1990

    InsanelyMac Protégé

  • Members
  • Pip
  • 29 posts
  • Gender:Male
  • Location:Argentina
Thanks dude. Maybe the guy who get the 420 fps must be using the new 560 Ti version with more CUDA cores. If it matters I run LuxMark too. This are the results for GPU only and CPU+GPU:

Posted Image


Posted Image


The guy who got 8 GeForce GTX 580 is insane! I thought you can only get two card by SLI and four GPU if the card is Dual GPU. Who manufactures a motherboard with 8 x16 PCI Express slots? And you'll need PCI Express 4.0 to get to work with full speed. I don't know I think is fake. What do you think?



#134
mitch_de

mitch_de

    InsanelyMacaholic

  • Retired
  • 2,902 posts
  • Gender:Male
  • Location:Stuttgart / Germany
Yep, OpenCL can, if so coded use more than one GPU. But depends on the code it uses much more cpu load for that.
8 GPUs may not be get full load by today CPU+systembus speed+system pcie speed - bottleneck in this case.
But shows how fast gpus can be in the future or used with very special universitary usage (mostly complex simulations).

#135
Regi Yassin

Regi Yassin

    Who am I ?

  • Members
  • PipPipPipPipPip
  • 278 posts
  • Gender:Not Telling
ASUS P8P67 Deluxe
i7 2600k (no OC)
Gigabyte 580 SOC
MacPro4,1

OpenCL_OceanWave submited, this is the screenshots

Attached File  Screen Shot 2011-12-13 at 1.15.43 AM.png   460.34KB   21 downloads

Attached File  Screen Shot 2011-12-13 at 1.05.43 AM.png   384.93KB   18 downloads

#136
mitch_de

mitch_de

    InsanelyMacaholic

  • Retired
  • 2,902 posts
  • Gender:Male
  • Location:Stuttgart / Germany
Great Luxmark value. (Info: Luxmark is also cpu dependent - faster cpu gives higher (some) luxmark value on same gpu - even on the gpu only bench).

#137
Hangten

Hangten

    InsanelyMac Sage

  • Members
  • PipPipPipPipPipPip
  • 404 posts
  • Gender:Male
  • Location:Laguna Beach, CA.
Posted Image

Here's my score, delayed by a month! LOL

#138
ujen

ujen

    InsanelyMac Protégé

  • Members
  • PipPip
  • 81 posts
Heres my run on my new X79 build with a 3930k and a GTX 570 OC:

Posted Image

And heres my Luxmark result:

Posted Image

#139
Badeendje

Badeendje

    InsanelyMac Geek

  • Members
  • PipPipPipPip
  • 213 posts
Shameless Crosspost;

Posted Image

HIS 6870

22 fps difference over my old score with an old version of the benchmark, and even more with the latest version with the stock .2 kexts!
Ran the benchmark a couple of times since this much difference has not been seen with any update since I started testing.

#140
mitch_de

mitch_de

    InsanelyMacaholic

  • Retired
  • 2,902 posts
  • Gender:Male
  • Location:Stuttgart / Germany
NEWS: AMD tested new AMD 7970 vs 6970 with Luxmark (win) - 2,3 times faster. AMD needs that, because GTX 580 was much faster in OpenCL. But beside faster HW the complete new Architecture (now more like Fermi) will takes time for OpenCL drivers on non Windows systems. Belongs to Apple if and when it supports (and devleops) for that new Highend GPUs. The new mobile 7xxxx GPUs are rebranded 6xxxM gpus so there is no need for much develops in drivers.

Attached File  Bildschirmfoto 2011-12-20 um 10.57.44.jpg   227.48KB   19 downloads





0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users

© 2014 InsanelyMac  |   News  |   Forum  |   Downloads  |   OSx86 Wiki  |   Mac Netbook  |   PHP hosting by CatN  |   Designed by Ed Gain  |   Logo by irfan  |   Privacy Policy