Matlab, hardware

[George Ormonde]

Any advice, My main application is matlab,
With a variety of parallel and sequential work loads,

the 3 build I'm looking at are
~$900 for Ryzen 1700 ---The performance of the Ryzen in compute is tempting, but the platform seems like a risk at the moment.
~$900 for i7-7700K ---- Boring safe choice, might be the best
~$1200 for I7-6800K ---- more than I want to spend, but may be worth it for the compute and option to expand memory and gpu in future

any advice ?

 

[lehtv]

When does it need to be built? If there's no hurry, you could just wait and see how the AM4 platform is received. I'm not sure what risks there are other than "it's new".

 

[dave_the_nerd]

...you really think you need that much oomph for Matlab?

Is that what you do all day every day for a living, or are you a college student who thinks they need their own server farm to major in CS, when they really just need a laptop and some Adderall?

 

[Knavish]

How familiar are you with writing parallel Matlab code? Lots of Matlab is single threaded, but you can force it to run "stupidly parallel" operations with functions like parfor and blockproc. Unless you're processing 1000's of images or huge datasets, I don't think it's worth spending on a super high-end CPU.

I've definitely run out of memory and been limited by disk I/O, but I don't feel like I'm typically waiting for the CPU to finish processing. Of course, your datasets may be completely different than mine. Half of the fun of Matlab is trying to figure out how to rewrite functions as linear algebra operations and using things like sparse matrices to make your code more efficient vs. just brute forcing it with a CPU.

p.s. I typically run Matlab on a Mac laptop with 16GB ram.

 

[freeskier93]

...you really think you need that much oomph for Matlab?

Is that what you do all day every day for a living, or are you a college student who thinks they need their own server farm to major in CS, when they really just need a laptop and some Adderall?

When I was in school one of the many uses we had for Matlab/Simulink was simulating aircraft dynamics. We could actually run the simulations in real time and pipe the data over to FlightGear to do the visualization. Even got to the point where you you give in control surface inputs. Essentially a flight simulator running a crude physics engine with Matlab. If you wanted to run long term simulations faster then real time (to see dynamic stability) you need something with quite a bit of oomph.

My laptop could barely run faster then real time so an hour long simulation could take 30-40 minutes to run. On my desktop the same simulation might take 5-10 minutes to run. When you have to run these a bunch of times it all adds up.

Orbital mechanics was also a real killer. Even using an efficient solver that used variable step sizes, something like plotting a earth to moon and return trajectory could take a long time to run.

Plus, Matlab isn't exactly an efficient language. It's designed to be quick and easy to use such that development time is short and outweighs possible code innefficiency in the short term. Although Matlab is definitely getting a lot better when compared to other languages like C.

Definitely glad I had a powerful desktop for Matlab though.