MacBook Pro specs for (primarily) scientific computing / data mining, large LaTeX documents

Discussion in 'MacBook Pro' started by UnbiasedDinosaur, Aug 14, 2019.

  1. UnbiasedDinosaur macrumors newbie

    Oct 30, 2015
    I'm trying to decide on specs for a new 13-inch MacBook Pro. I'll primarily use it for scientific computing / data mining (e.g. MATLAB, Python with numpy) and large LaTeX documents (compiling has become frustratingly slow on my current machine, the original 12-inch MacBook). I've become accustomed to a single port, so two vs. four ports makes no difference. For storage, 256GB will suffice. Thus, I'm mostly debating the following:
    • Should I choose 1.7GHz i7 or 2.4GHz i5 processor? (There's no price difference, and Geekbench gives the latter ~5% higher single core score but ~5% lower multi core score, so it's unclear which is "better".)
    • Is the 16GB RAM upgrade worth the $200?
    If instead you think the entry level 1.4GHz is fine for my use cases, or the high-end 2.8GHz is worth the price jump, let me know. Thanks!
  2. Howard2k macrumors 68020


    Mar 10, 2016
    2.4Ghz machine has faster CPU, faster wifi, faster GPU, faster SSD, than the 1.4/1.7. So if the 1.4 doesn't work I would skip the 1.7 and go straight to the 2.4.
  3. UnbiasedDinosaur thread starter macrumors newbie

    Oct 30, 2015
    Thanks for the advice. I figured there was some downside to the 2.4GHz machine, since it's priced the same as the 1.7GHz one but you're also paying for extra ports. But this makes it sounds like the 2.4GHz is just better, and the 1.7GHz is perhaps overpriced?
  4. Howard2k macrumors 68020


    Mar 10, 2016

    For sure. The 1.7Ghz machine is an IQ test. If you choose it you failed. :)

    I'm sure there might be some obscure reason that makes it more sensible than the 2.4Ghz machine but I can't think of one.

    The 1.4Ghz is good value IMO. Just not worth upgrading the CPU at all.
  5. solouki macrumors member


    Jan 5, 2017
    Hi UnbiasedDinosaur,

    Are you using TeXLive? And what do you mean by "large LaTeX documents"?

    I routinely typeset 18000+ 8.5x11 pages of LaTeX filled with math, vector diagrams, data plots, and images, and at least in my experience the faster the Turbo Boost of the CPU the better -- i.e., the typeset time is roughly inversely proportional to Turbo Boost frequency.

  6. bill-p macrumors 68000

    Jul 23, 2011
    Sorry for quoting you out of context, but this is basically your problem.

    Any other MacBook in the current generation will be far faster at any task than this machine. While it's okay for light computing, from my experience, it's just not up to the task of compiling large projects at all.

    So the CPU will be the most important thing. Barring that... 8GB or 16GB of RAM won't matter much unless you run large data sets.

    For mining, you're better off offloading that to a Linux server. Personally, I use AWS because I can get access to some with GPU, and they speed up computation significantly. Unless you're planning on running everything locally on your own laptop.

    Jupyter should be deployed to the cloud.
  7. leman macrumors G3

    Oct 14, 2008
    The big issue with LaTeX is that it cannot utilise multiple CPUs. So it's going to be slowish no matter what. That said, the modern CPUs are fairly fast. MATLAB can be a memory hog depending on your data sizes, so 16GB could be a good idea.

    I would recommend you to get the 2.4Ghz (or 2.8Ghs if the money is no issue) CPU. The 1.7 is slightly faster in burst workloads than the 2.4Ghz, but it will slow down in longer workflows (and in your lien of work you'd be probably doing analyses that last for 30 minutes or more).
  8. UnbiasedDinosaur thread starter macrumors newbie

    Oct 30, 2015
    I'm using TeXstudio, but I assume it's the same LaTeX compiler so the typeset time is similar? Good to know about typeset time vs. Turbo Boost frequency; seems things will speed up over my current machine no matter which Pro I buy.
    --- Post Merged, Aug 15, 2019 ---
    Yes, I was prioritizing portability when I bought the original MacBook, but I've come to regret that as my work has become more intensive.

    Indeed. My typical workflow is to prototype things locally on smaller datasets (<1 hour tests), then to run things on a remote server (code that takes hours or days to complete). My main goal is to speed up those local tests.

Share This Page

8 August 14, 2019