Good laptop for parallel computing using matlab??

Hello, Looking for a good laptop (has to be a laptop, not desktop) to perform parallel computing on large sets of data. I regularly need to analyse large sets of data, in which even my desktop takes forever to complete. I am therefore learning how to write more of my code using parallel computing functions (e.g. parfor). I would like to buy a laptop (pre-made or custom-built (from a company)) that is as fast as possible. Can anyone please give me some advice on what laptop specifications are essential for this (and matlab in general) when getting a computer. I assume a CPU with as many cores as possible, a good GPU as well (for use of functions such as gpuArray), an SSD drive and as much RAM as possible. Thanks in advance, Tim

[Walter Roberson]

You should probably research "gaming" laptops. And you will have to work hard to keep it cool.

https://www.pcmag.com/roundup/360470/the-fastest-laptops

https://www.laptopmag.com/articles/best-laptop-coolers and https://www.lifewire.com/ways-to-keep-your-computer-cool-2624713

[OCDER]

Don't forget the cooling pad (~$30). And if that's not working to keep things cool, place laptop in the fridge - keep away from perishable food & leave explanatory post-it.

https://www.mathworks.com/help/matlab/matlab_prog/techniques-for-improving-performance.html

[Timothy Olsen]

Hi, Thanks for the answers, sorry for my delay. Your answers were very helpful. Indeed the problem will be keeping the laptop cool. My favourite choose currently is the Dell XPS 15 but as you mention, this has a lot of issues with thermal throttling. I'm hoping a high CPU usage with relatively small GPU usage while running code will avoid some of the problems associated with excess heat. Thanks again. Tim

Answer 1

[FastCar]

Are you sure that your task could be done in parallel? Did you check the memory requirements for the task? Mind that a laptop compared to a workstation or a HPC has limitations in RAM and in number of cores. Also sometimes the process of assigning the task to the cores is so heavy that in Matlab you don't get the speed up you expected.