We have a big Max7 application 95% built in OpenGl for video

The application is built to use the 2 GPUs of our MacPro (Yosemite 10.10.3)

We have some problems and need to know if there Is a maximum

use of VRAM with Jitter or maximum opengl objects to use.

If so would a standalone of some treatments help us bypass the problems.

We have a big Max7 application 95% built in OpenGl for video
and multiple displays (6)
The application is built to use the 2 GPUs of our MacPro (Yosemite 10.10.3)

We have some problems and need to know if there Is a maximum
use of VRAM with Jitter or maximum opengl objects to use.

urgent-maximum-vram-use-with-jitter

I made a permanent install w. nMP D500. full OpenGL.

I had to pass to max6-64Bit to load 3 Go of textures on the graphic card. I did not test in Max7 64bit.

max6/7 32 bit is limited to 2 Go. it comes from 32 bit system in general... and not max.

Given that others components of my install works only in 32 bit, I had to make standalone of them and send textures to the main 64 bit application with syphon or netsend or udp.

It4s also the way to go if you want to split your processor ressources.

As said in another thread, Max and other realtime software don't make any use of the 2nd GPU in nMP. Only Professionnel OpenCL software. So it's money for nothing...

I don't think neither i managed to use the 3Go of Vram of the second card.

So with a D700 and max 64 bit, you could load 4Go of Vram minus the max app.

A solution is to dynamically upload images on the same set of jit.gl.texture. So it won't create new mem ressources.

The future of realtime with osx looks very inconsistent. OpenGl is badly optimized and graphic cards lag 2 years behind. and horribly expensive.

Personnaly I swaped to Win Laptop. Schenker W705 _ 2x512 SSD _ i7-4790S _ GTX980M

Compare to nMP D700: Half the price, 20% more power, compatible w kinect2... portable _ 2xDP+Hdmi

so you can get 7x 720p w. matrox or 5x 1080p

I made a permanent install w. nMP D500. full OpenGL.
I had to pass to max6-64Bit to load 3 Go of textures on the graphic card. I did not test in Max7 64bit.
When you reach the limit, it just crash.
max6/7 32 bit is limited to 2 Go. it comes from 32 bit system in general... and not max.

Given that others components of my install works only in 32 bit, I had to make standalone of them and send textures to the main 64 bit application with syphon or netsend or udp.
It4s also the way to go if you want to split your processor ressources.

As said in another thread, Max and other realtime software don't make any use of the 2nd GPU in nMP. Only Professionnel OpenCL software. So it's money for nothing...
I don't think neither i managed to use the 3Go of Vram of the second card.
So with a D700 and max 64 bit, you could load 4Go of Vram minus the max app.

The future of realtime with osx looks very inconsistent. OpenGl is badly optimized and graphic cards lag 2 years behind. and horribly expensive.
Personnaly I swaped to Win Laptop. Schenker W705 _ 2x512 SSD _ i7-4790S _ GTX980M
Compare to nMP D700: Half the price, 20% more power, compatible w kinect2... portable _ 2xDP+Hdmi
so you can get 7x 720p w. matrox or 5x 1080p
or get a Datapath X4.

I tried with Max7 in 64bit mode and I get error like jit.movie @colormode 

then my fps that would usualy be around 56 fps, goes right down to 18fps.

As you mentionned , a 20% increase of performance is sometimes exactly what you need.

Also, SPA as you mentioned in the last statement

>> So with a D700 and max 64 bit, you could load 4Go of Vram minus the max app.

the minus the max app part, would you include 

OR did you just meant -minus Max the app size none the less.

I tried with Max7 in 64bit mode and I get error like jit.movie @colormode 
then my fps that would usualy be around 56 fps,  goes right down to 18fps. 

As you mentionned , a 20% increase of performance is sometimes exactly what you need. 

Also, SPA as you mentioned in the last statement
>> So with a D700 and max 64 bit, you could load 4Go of Vram minus the max app.
the minus the max app part, would you include 
OR did you just meant -minus Max the app size none the less. 

Thanks for your answer SPA about the total amount of VRAM.

Knowing that in the first place I would have never bought a MacPro with 2 video cards. USELESS.

Even a MacPro with Yosemite is really borderline (yeah voodoo psycho lol ...).

I am wondering if there is a way to calculate the amount of VRAM use per video sequence and gl. objects?

I had an extra 1,5 go of VRAM before some additions of about 20 gl objects (16 gl.pix and 4 jit.movie) for a final resolution output of 960 x 960 (most processing being done at 320 x 320 dim).

Thanks for your answer SPA about the total amount of VRAM.
Knowing that in the first place I would have never bought a MacPro with 2 video cards. USELESS.
Even a MacPro with Yosemite is really borderline (yeah voodoo psycho lol ...).

it's a MacPro D500(3Go) with osx.9 w. max6-64bit

a lot of textures and planes in poly. basically png FHD

So i simply use 'activity monitor' and launch my appli without charging the textures/planes.

and then it starts to load 1 by 1 some 200 png and you can observe the amount of Ram for max increasing.

My empirical vram limit test is ... empirical, and goes through ram in max.

with 64 bit, it crashes at 3,5 Go Ram that is for me: 500Mo for max and objects and then loading 3 Go for textures...

My need was to have ALL textures loaded on the graphic card to be super reactive. it's certainly not the best optimisation.

But with swaping the image/movie uploaded on a texture, i don't think you will increase the vram amount.

MH, difficult to evaluate the 'weight' of an openGL object... but i think that texture (FHD) take much more memory on the graphic card.

You can always do a quick poly abstraction and see how max react with changing the number of instances.

Or simply use 'OpenGL Driver Monitor' from 'Graphic Tools' from Xcode for better estimate.

it's a MacPro D500(3Go) with osx.9 w. max6-64bit
my appli is quite specific.
a lot of textures and planes in poly. basically png FHD
So i simply use 'activity monitor' and launch my appli without charging the textures/planes.
and then it starts to load 1 by 1 some 200 png and you can observe the amount of Ram for max increasing.
My empirical vram limit test is ... empirical, and goes through ram in max.
with 32 bit, it crashes at 2Go Ram
with 64 bit, it crashes at 3,5 Go Ram that is for me: 500Mo for max and objects and then loading 3 Go for textures...

My need was to have ALL textures loaded on the graphic card to be super reactive. it's certainly not the best optimisation.
But with swaping the image/movie uploaded on a texture, i don't think you will increase the vram amount.

MH, difficult to evaluate the 'weight' of an openGL object... but i think that texture (FHD) take much more memory on the graphic card.
You can always do a quick poly abstraction and see how max react with changing the number of instances.

Or simply use 'OpenGL Driver Monitor' from 'Graphic Tools' from Xcode for better estimate.


I just saw in another posting that opening Max in "low resolution mode"

is helping for Retina machines. I wonder if it could help us for our problem

Would be nice to have advice from Cycling on all this and clear procedure.

I am spending a lot of money and time on a lot of voodoo.

I just saw in another posting that opening Max in "low resolution mode"
is helping for Retina machines. I wonder if it could help us for our problem
with Yosemite. 

Would be nice to have advice from Cycling on all this and clear procedure.
I am spending a lot of money and time on a lot of voodoo.
It's getting highly annoying !


Back with a variant of this previous problem.

We finally made our application with Max7.0.3 running on Yosemite 10.0.3 and the project is now on display in an exhibition room

It has 6 displays. 1 control monitor and 5 other 1280 x 720 displays. It's all running on Max7 64 bit and loading videos

with poly~ object (thanks to Phiol and SPA).

We decided to restart the MacPro automatically every night so the VRAM could be released after quitting Max

because we came to the conclusion that the VRAM kept increasing until crash.

We didn't see any other way at the time for that.

But it seems (what I can see on remote control and snapshots every 5 min) that the MacPro

doesn't always keep it's displays alignment after restart or reboot. And there is no way to save

Is there a way to keep the MacPro on 24h/24h without restart and release the VRAM ?

This way my display setup wouldn't change

Or if I leave it on without running Max for few hours, will it empty the VRAM slowly ?

We finally made our application with Max7.0.3 running on Yosemite 10.0.3 and the project is now on display in an exhibition room
It has 6 displays. 1 control monitor and 5 other 1280 x 720 displays. It's all running on Max7 64 bit and loading videos
with poly~ object (thanks to Phiol and SPA).

We decided to restart the MacPro automatically every night so the VRAM could be released after quitting Max
because we came to the conclusion that the VRAM kept increasing until crash.
We didn't see any other way at the time for that.
But it seems (what I can see on remote control and snapshots every 5 min) that the MacPro
doesn't always keep it's displays alignment after restart or reboot. And there is no way to save
 the display preferences.

Is there a way to keep the MacPro on 24h/24h without restart and release the VRAM ?
This way my display setup wouldn't change
A script to the terminal ?
Something ?
Or if I leave it on without running Max for few hours, will it empty the VRAM slowly ?

Urgent - Maximum VRAM use with Jitter ?