macbook i5 dual vs. i7 dual vs. i7 quad

I am purchasing a new macbook pro and wonder which processor will give the best MatLab performance, or rather if the difference in performance will be significant.

all are with-in price range, but I would opt to spend less money if there is no sig diff between the i5 and i7, or the i7 dual and i7 quad.

If anyone can help, it would be much appreciated.

Most of my applications right now are integration of systems of ODEs and PDEs. Also, numerical solvers in multiple dimensions.

I will be starting to do more quant finance in the coming months and year. But am not sure what methods that will involve...except for the processing of large amounts of data.

I do not use the Parallel Processing Toolkit, but this is not something I purposefully deciding...maybe I should be(?)

Along those lines, there is use for GPUs when it comes to plotting vortex ring solutions etc... Some of my fellow grad students are using a GPU workstation.

But I can use the workstation, maybe remotely from a laptop (?)

The reason I need a laptop is, of course, portability. This is a must at this point...I can worry about upgrading my tower later.

Will

[Jan]

Please insert additions to your own question by editing, not as comment.

Answer 1

[Walter Roberson]

For these sorts of things we need to have an idea of what kind of problems you will be computing, and whether you will be using the Parallel Processing Toolkit.

[Walter Roberson]

None of the MacBook Pro line are able to be used for Matlab's GPU computing.

Some of the models that existed until recently did use NVidia graphics boards, but those graphics boards did not support a high enough level of CUDA for Matlab to use.

The new models of MacBook Pro recently announced use either ATI Radeon graphics or graphics built in to the CPU itself, and thus are not even close to supporting GPU computation.

[William]

As I mentioned, I could use a GPU workstation. The main purpose is portability. If there is a worthwhile laptop with an nividia card...i would be interested and possibly run unix.

Answer 2

[Lasse]

Here's the thing, Matlab seems not yet to be optimised for these current multicore processors (since core 2 duo). Effectively, any other tasks than matrix manipulations utilise only a single thread at a time (in the dual core i7 in the MDP there are 4 threads). This means that until this is fixed, it does not matter much what the processor is, if you're not making your code explicitly parallel. ODEs are particularly a pain the youknowwhat.

[Knut]

Your arguments does not support that "it does not matter much what the processor is". Rather, it supports that "having more than one core will often not give substantial speedup in MATLAB".

Modern multi-core processors have substantial differences in single-threaded/single-core performance. I would assume that this translates into different MATLAB performance.

[John D'Errico]

It is NOT true that matlab does not use more than one core, although very much depends on what you are doing. I have an 8 core machine, and I can often find between 2 and all 8 cores hard at work. On big problems, you can even hear the fan kick on when things really start to crank. The thing is, matlab does not fire up the extra cores unless it sees something that it can benefit from. So on manipulations that involve large arrays, yes, things will happen. On a loop, maybe not, even if you think that matlab is working hard, unless of course you are using the explicit parallel processing toolbox.

[Jan]

Some mutlithreaded function: SUM, MIN, MAX, FILTER, ... For small arrays a single thread is faster, because starting a thread needs a lot of time.