Jump to content

fermi/gtx 480/gtx 470


cmf
 Share

1,729 posts in this topic

Recommended Posts

strange....

for me: OceanWave Display: 13 fps

why?

Maybe take a look in the Console log (within the app) - maybe som errors listed in the opencl info and , after running OceanWave

other errors.

Mine looks like:

OpenCL OceanWave:-------------------------------------------------------------------------------
-------------------------------------------------------------------------------
Connecting to NVIDIA GeForce 9600 GT... (my CPU : C2D 2.9 Ghz)
-------------------------------------------------------------------------------
Building compute program...
-------------------------------------------------------------------------------
Building compute kernels...
Starting event loop...
ESC key abort / [GPU] Compute: 44.02 ms  Display: 22.72 fps

 

OpenCL may not work in all cases you want / Also Apple itself is in the beginnings of OpenCL dev - OceanWave is orig. Apple Source Code, unmodded by me.

 

But for sure nice to see your results - if they stay very different - OceanWave isnt usable as bench in the Final of my smallluxGPU, like Luxball Ultra oder LuxMark.

Link to comment
Share on other sites

OpenCL OceanWave:-------------------------------------------------------------------------------
-------------------------------------------------------------------------------
Connecting to NVIDIA GeForce GTX 470...
-------------------------------------------------------------------------------
Building compute program...
-------------------------------------------------------------------------------
Building compute kernels...
Starting event loop...
ESC key abort / [GPU] Compute: 75.52 ms  Display: 13.24 fps

Link to comment
Share on other sites

Maybe compiler (it does compile on the fyl at runtime) isnt good for that gpu type with that OpenCL source code.

GTX 460 / 480 seems to not have that "problem".

Doesnt matter until real OpenCL apps rise :(

 

What you can do (and othersa for compare)

Run an very lowlevel OpenCL Bandwith bench - shows device device memtransfer speed (VRAM speed) and speed over the PCIe (host> gpu, gpu>host).

 

I get with NV 9600 GT, C2D 2.8 GHz, 280 MHz FSB :

/Developer/GPU Computing/OpenCL/bin/darwin/release/[b][size=4]oclBandwidthTest[/size] [/b]Starting...
WARNING: NVIDIA OpenCL platform not found - defaulting to first platform!
Running on...
GeForce 9600 GT
Quick Mode
Host to Device Bandwidth, 1 Device(s), Paged memory, direct access
  Transfer Size (Bytes)	Bandwidth(MB/s)
  33554432			2464.3
Device to Host Bandwidth, 1 Device(s), Paged memory, direct access
  Transfer Size (Bytes)	Bandwidth(MB/s)
  33554432			1879.2
Device to Device Bandwidth, 1 Device(s)
  Transfer Size (Bytes)	Bandwidth(MB/s)
  33554432			30777.4

 

Yours should be much higher - with fast CPU/Bus System also the 2 first ones (+20 - +50%). But most diff should be dev-dev speed to mine (+50 - +150%).

 

Compare that to others (with GTX 4xx) - should not be very much different with similar CPU type.

Otherwise some prob (idle mode / low MHz = less speed ....).

oclBandwidthTest.zip

Link to comment
Share on other sites

Now we have native support for nvidia Fermi on Mac os x 10.6 Link

post-193013-1304582011_thumb.png

DOWNLOAD Only Nvidia Drivers

 

sorry for my english

 

I tried this driver and now I've a 100% working GTX 460!!!

With this drivers you don't need the video loop or iTunes methods.

What you have to do is just replace the kexts into the system/library/extensions with the kext in the rar file.

I just had a problem: I wasn't able to overwrite them.

The solution was: backup the original kexts into a new folder, delete them from extensions folder and write the new kexts into the extensions folder.

The last thing that you have to do is repair permission and rebuild extensions.mkext (I used Kext Utility or Kext Helper, I don't remember now... I'm at work in this time!!!).

Link to comment
Share on other sites

Now we have native support for nvidia Fermi on Mac os x 10.6 Link

 

I tried this driver - I have a 465 GTX - I get about 1fps slower on Cinebench (18.63 as opposed to 19.54 on the driver from the nVidia site). I'll let you know what kind of reliability I get...

 

Patrick

Link to comment
Share on other sites

that's what I see

WARNING: NVIDIA OpenCL platform not found - defaulting to first platform!

Running on...

GeForce GTX 470

Quick Mode

Host to Device Bandwidth, 1 Device(s), Paged memory, direct access
  Transfer Size (Bytes)	Bandwidth(MB/s)
  33554432			1801.7

Device to Host Bandwidth, 1 Device(s), Paged memory, direct access
  Transfer Size (Bytes)	Bandwidth(MB/s)
  33554432			5708.3

Device to Device Bandwidth, 1 Device(s)
  Transfer Size (Bytes)	Bandwidth(MB/s)
  33554432			72315.6

logout

Link to comment
Share on other sites

I tried this driver - I have a 465 GTX - I get about 1fps slower on Cinebench (18.63 as opposed to 19.54 on the driver from the nVidia site). I'll let you know what kind of reliability I get...

 

Patrick

For me, 9600 GT OpenGL near same, BUT OpenCL (some OpenCL benches, not all) massiv slower (1/2) like Luxball Ultra.

Also other user reported same with 9800 GT (ist 8800 GT renamed)..

New drivers : (iMac 2011)
MacPro3,1 MacPro3,1 OpenCL GPU   0:			GeForce 9800 GT 	1500 MHz
CPU: 			Intel® Core(tm)2 Quad  CPU   Q9300  @ 2.50GHz 	3442 MHz
Result=Luxball Ultra  160,2 sec sec Mac OS X 10.6.7

Normal drivers:
MacPro3,1 MacPro3,1 OpenCL GPU   0:			GeForce 9800 GT 	1500 MHz
CPU: 			Intel® Core(tm)2 Quad  CPU   Q9300  @ 2.50GHz 	3442 MHz
Result=Luxball Ultra  86,2 sec sec Mac OS X 10.6.7

Priority is OpenGL - clear, but have an focus of that later if OpenCL does matter in apps.

Link to comment
Share on other sites

that's what I see

WARNING: NVIDIA OpenCL platform not found - defaulting to first platform!

Running on...

GeForce GTX 470

Quick Mode

Host to Device Bandwidth, 1 Device(s), Paged memory, direct access
  Transfer Size (Bytes)	Bandwidth(MB/s)
  33554432			1801.7

Device to Host Bandwidth, 1 Device(s), Paged memory, direct access
  Transfer Size (Bytes)	Bandwidth(MB/s)
  33554432			5708.3

Device to Device Bandwidth, 1 Device(s)
  Transfer Size (Bytes)	Bandwidth(MB/s)
  33554432			72315.6

logout

All OK with that result / OpenCL . Some OpenCL code will get slow with iMac 2011 update / some not. will be fixed with 10.6.8 update i think.

Link to comment
Share on other sites

Great !

Here some other (emailed to me):

Vladimir: MacPro3,1 OpenCL GPU 0: GeForce GTX 480 1401 MHz

CPU: Intel® Core™2 Quad CPU Q9450 @ 2.66GHz 3072 MHz

Result=Luxball Ultra 20,5 sec sec Mac OS X 10.6.7

OceanWave Display: 59.41 fps

 

that's me :(

 

 

opencl is tricky, things like loops, branching can slow down things without much reason, even worse, same opencl code can run differently in different OSes, due to different driver optiomizations

Link to comment
Share on other sites

After installing the drivers I just get a black screen with some garble when the desktop is supposed to come up. Running a 580GTX.

 

10.6.7

No injectors or anything. (at least I dont think so)

 

runs fine in safe mode (full res, dual screen) just no acceleration

 

Any ideas?

 

I added an EFI hex into my boot plist, still no go. The details show up in safe mode now, but normal boot = black garble.

 

Here is a pic of the screen I get at boot

photo.JPG

Link to comment
Share on other sites

What options do we have for graphics injection on 10.7?

 

I am using Netkas Chameleon 10.7 for boot with non fermi card.

 

What options are available for injection; boot loader with injection or kext for fermi in Lion?

 

Sorry if it has been asked and I missed it.

 

Ty.

Link to comment
Share on other sites

Why OpenGL is so slow in GTX480?

 

I haven't been able to play Portal 2 or L4D on OSX without stuttering since I had my GTX285. :(

 

@Gidi

 

Same here, I manage to get my new EVGA GTX 460 working, but it is unbelievably slower than my GTX 285, I use CAD to work on schematic/drawing...noticed that it is very slow:(

Link to comment
Share on other sites

My Twin Frozer GTX465 (while somewhat quieter and possibly cooler) is no faster than my GTX275.

 

That said, no GFx crashes/KPs since installing the native drivers above... I wonder if I should put back the 275 and see how it performs with these drivers.

 

Patrick

Link to comment
Share on other sites

hello, I do not speak English, I'm using a translator, I apologize in advance.

I urgently need help, some h264 video compression I can not see them, importing the clips to Final Cut Pro is no sound but the video viewer does not go on, it makes me see the shots of still images or pre-made ​​video desktop background, even when put in time on canvas does the same thing, the same clip from the same problem with MPEG Streamclip but not with quick time.

same problem in acquisition of minidv always final cut pro, in the preview window you can see flashes of video images captured previously with other backgrounds and windows, but the sound is heard and the acquisition is done correctly and audio that video, I just do not have the preview.

which may be the problem?

thanks

 

ah, on the laptop that's obviously less accurate nn I fixed these problems, handle them with great h264 me I have no problem acquiring and nos, so my question is probably missing some video codec or the proper functioning of hardware video nn .

 

I found someone who has the same problem with a lot of videos to make the idea of the problem.

from what I can understand the problem is the graphics card series 400

 

http://www.insanelymac.com/forum/index.php...t = # entry1682680

 

this is his video:

 

thanks

Link to comment
Share on other sites

The guide is outdated. If you mean the one in the first post.

 

Install the Quadro 4000 drivers for Snow Leopard from Nvidia. Make sure you are running at least 10.6.5 first.

Your model identifier must be MacPro3,1, 4,1 or 5,1. If it isn't, drag the installer off of the disk image, right click, show contents, find distribution.dist inside, open it with textedit and change the model identifier requirement so that it matches yours (see System Profiler). It will be obvious what to do once you're looking at it.

 

Install Chameleon 2.0 RC5 and set GraphicsEnabler=y in com.apple.Boot.plist.

 

Use the DVI output that's closest to the motherboard when looking straight at the back of your PC case.

Link to comment
Share on other sites

 Share

×
×
  • Create New...