Thread Subject: Does Matlab use the GPU at all?

I'm buying a new laptop and will be using it with matlab R2007a (and R2007b
when it arrives). I'm stuck between the MacBook and the MacBook Pro.

Does the MBP's dedicated graphics card affect Matlab performance? ie: does
matlab utilize both the CPU and the GPU.

The only other difference is a 667MHz FSB vs a 800MHz FSB and many CPU-
benchmarks run on the two laptops show only minor performance increases...
so my choice comes down to whether the better MBP graphics card will be used
by matlab or not.

Any ideas? Thanks!

On Aug 18, 12:44 pm, "Jeff " <jkear_REMOVE...@ucsd.edu> wrote:
> I'm buying a new laptop and will be using it with matlab R2007a (and R2007b
> when it arrives). I'm stuck between the MacBook and the MacBook Pro.
>
> Does the MBP's dedicated graphics card affect Matlab performance? ie: does
> matlab utilize both the CPU and the GPU.
>
> The only other difference is a 667MHz FSB vs a 800MHz FSB and many CPU-
> benchmarks run on the two laptops show only minor performance increases...
> so my choice comes down to whether the better MBP graphics card will be used
> by matlab or not.
>
> Any ideas? Thanks!

The last I heard Mathworks was still using X 11 to run Matlab on Macs
and a universal binary version of Matlab for Macs was still not
available. So a better graphics card will not produce any direct
improvement in performance. However, you may see a slight performance
increase with graphics intensive applications. Most likely it won't
affect you but I was disappointed when I ran Simulink on an iMac.

Howard

On Aug 18, 9:38 pm, hrh1818 <hrh1...@comcast.net> wrote:
> On Aug 18, 12:44 pm, "Jeff " <jkear_REMOVE...@ucsd.edu> wrote:
>
> > I'm buying a new laptop and will be using it with matlab R2007a (and R2007b
> > when it arrives). I'm stuck between the MacBook and the MacBook Pro.
>
> > Does the MBP's dedicated graphics card affect Matlab performance? ie: does
> > matlab utilize both the CPU and the GPU.
>
> > The only other difference is a 667MHz FSB vs a 800MHz FSB and many CPU-
> > benchmarks run on the two laptops show only minor performance increases...
> > so my choice comes down to whether the better MBP graphics card will be used
> > by matlab or not.
>
> > Any ideas? Thanks!
>
> The last I heard Mathworks was still using X 11 to run Matlab on Macs
> and a universal binary version of Matlab for Macs was still not
> available. So a better graphics card will not produce any direct
> improvement in performance. However, you may see a slight performance
> increase with graphics intensive applications. Most likely it won't
> affect you but I was disappointed when I ran Simulink on an iMac.
>
> Howard

R2007{a,b} is a Universal binary because Java is the one thing that
Rosetta cannot dynamically translate. But you're right about the X11
thing. Still, if your applications in Matlab use OpenGL for something,
it will be accelerated by your GPU (as long as hardware acceleration
is on). That said, Matlab cannot (without special third-party
software) utilize the GPU on any platform. On Linux and Windows
platforms, Nvidia has a plugin called CUDA which performs accelerated
FFT computations on the GPU using mex files (i.e. code external to
Matlab). Others have written plugins in a similar way. One issue (I
think) that keeps the Mathworks away from using the GPU to accelerate
computations is that GPUs usually only support single-precision
computation (that is, images made for viewing are usually 32-bits at
most, implying that the mathematical operations GPUs have available
are also 32-bit, not 64-bit) instead of Matlab's default double-
precision computation. That means that GPU-acclerated computation has
at least one tradeoff: accuracy. If that's not concern #1, then it
seems that GPU-acceleration is great. In some of my research, 64-bit
precision has proven insufficient, so I could never dream of using GPU-
acceleration.

In short, you can buy a Macbook or a Macbook Pro or an HP or a Dell or
a whatever. No matter which OS you use, you don't get GPU-accelerated
calculations with Matlab. Not without some special programs.

I hope this helps.

Fort

Hi,

We at AccelerEyes have just released Jacket: a full GPU
(CUDA-based) engine for MATLAB. It allows your MATLAB code
to take advantage of both the compute and visualization
benefits of the GPU. You can download a free beta version at:

http://www.accelereyes.com

Hope this helps!

Best,

John Melonakos

"Jeff " <jkear_REMOVE_ns@ucsd.edu> wrote in message <fa7b69$o57$1@fred.mathworks.com>...
> I'm buying a new laptop and will be using it with matlab R2007a (and R2007b
> when it arrives). I'm stuck between the MacBook and the MacBook Pro.
>
> Does the MBP's dedicated graphics card affect Matlab performance? ie: does
> matlab utilize both the CPU and the GPU.
>
> The only other difference is a 667MHz FSB vs a 800MHz FSB and many CPU-
> benchmarks run on the two laptops show only minor performance increases...
> so my choice comes down to whether the better MBP graphics card will be used
> by matlab or not.
>
> Any ideas? Thanks!

Hi,

you can try our Freeware library GPUmat from http://www.gp-you.org. GPUmat enables Matlab code to run on GPUs. Check the documentation to find out how many features we have implemented.

Thanks,
The GP-you Group