I'm contemplating the purchase of one of the new shiny Macbook Pros with that new fancy graphics chip. Now I am wondering if these chips are (or could eventually be) beneficial for making music in any way, shape or form.

I don't understand much about these things (coming to think of it, I don't even really know what a chip is, let alone why there have to be dedicated graphics chips), but it seems to me that a chip that excels at rendering 3-D-graphics should also be able to serve reverbs like Space Designer, etc.

Can graphic chips be used by plugins?

No. They can do graphics, nothing else. The only thing they might help is sluggish screen redraw, or speed up scrolling thousands of regions.

Yes, it is possible in the future. GPUs are becoming more and more powerful every day. In fact, many GPUs are even more powerful than CPUs. What one can do is offload some calculations from the CPU to a GPU (much like you do in distributed computing over the network; Logic Node is one such example).

Technologies, such as Grand Central and OpenCL that will be featured in Apple's next release of Mac OS X—Snow Leopard—will do exactly that.

Here's what Apple has to say about it:

(Italics mine.)

Veerrryyy interesting.

From what I understand, these new MBPs have 2 separate graphic chips. You currently have to decide which one you use before you start applications; the other seems to sit around idly, which I find kind of baffling.

Is the hardware of the current crop of MBPs sufficient to do all the magic that you are describing once the software is there?
Or is some different, future hardware setup necessary?
In other words, am I gonna bite myself in the ass half a year down the road if my newly purchased unibody MBP is not capable of taking advantage of such a new feature?

There was a synth software a few years ago that required a good GPU as it used the DSP technology to make the sounds.

DSP is DSP. It is like when the Velocity Engine came about...no one knew how to use it. And now...most programmers just ignore it because CPU time is cheap. But native CPU time is getting to a standstill and all the real research is in creating dedicated DSP hosts -- most in the form of GPUs. As programmers get use to using these things and going outside the boundaries, it will happen.

Today? And with the current chipset? Give it a couple years before anyone is doing anything interesting with it...

I'm no where close to knowledgeable about this stuff (unlike the posts above) but it's really cool technology that we're eventually surely going to benefit from more than most other "pro" users I'd have thought. After all, unless you work with DV and audio, our basic graphics needs have been met for quite some time and are unlikely to increase . . . will be interesting to see what happens over the next couple of years

Difficult to say at this point. The addition of an Nvidia chipset (not just the GPU) is an interesting move. Snow Leopard (and even 10.5.6) will definitely tap into the new possibilities even on the current hardware (by current, I mean the newly introduced portables). There are reports already that the new laptops use hardware H.264 decoding. This may be the first (albeit a little too obvious) step toward distributed processing within a single workstation.

So, don't hold your breath yet. Not only does the operating system have to provide the technologies necessary for "all the magic" as you put it; developers will need to update their software in order to take advantage of the new technologies. These things take time.

By the way, eMagic was one of the first companies to use AltiVec in its products.

Well, I seem to buy my computers in 6 or 7 year increments. I went from a Quadra 650 to the Ti-Book which I currently own, and now I am looking to extend myself again. So I have a long breath.
I am well aware that new machines are pretty much obsolete after about 2 weeks, but I would hate to purchase a new MBP and then see a new one come out half a year down the road that can run 237 Space Designers due to "distributed processing", while mine can´t.

Any categorical statement is complete bullshit. (Self-referentiality rules!)

Anyway.....

Graphics chips used to be just for graphics, but these days they are programmable, and people make them do amazing things, things that have nothing to do with graphics.

Go to nvidia.com and search for "CUDA courses" or cuda documentation. CUDA is their programming language (framework, something) so that programmers can write code for the GPU. Any code. CUDA is just C with some extensions.

For a proof of what you can do with GPUs, the Los Alamos "Roadrunner" computer was the first one to break the petaflop barrier, and it got there thanks to the fact that next to every regular cpu there was in IBM Cell processor, the same that's in your Playstation 3.

Now for the ifs and buts.

Not everything works on these GPUs. They have amazing power and bandwidth if your data is already on the chip. Getting it on and off the GPU is very slow. That means that only certain algorithms are suited for running on GPUs. (Example: genome sequencing. Data is linear in genome length, work is quadratic.) The Los Alamos team had to invest a gawdawful amount of work to rewrite codes to be efficient on Roadrunner.

What does this mean? Sorry, I don't know enough about DSP algorithms. I think they can benefit from a modest speedup on such chips. So if Apple will standardize on CUDA-programmable chips, then Emagic may actually program Logic to take advantage of them. This is very much like software on Windows using SSE to speed up. It's extra work for the programmer, but it can be done.

Victor.

Well, let's just hope then developers prove me wrong soon!