3 ms·

“complex simulations” strikes me as particularly wannabe-ish. I really do run pretty complex simulations that take hours, sometimes days, to run to completion (

by qubex 8y ago

“complex simulations” strikes me as particularly wannabe-ish. I really do run pretty complex simulations that take hours, sometimes days, to run to completion (on a maxed-out 2012 Mac Mini with dual SSDs and the maximum amount of memory it will support). It would never occur to me to run such stuff on a laptop that might run out of power, throttle back the processor because of heating, or something else going wrong. “Managing data and complex simulations” seems to allude to ‘em Excel jockeys who should be using a humble database and/or are running some kind of what-if calculations.

et2o 8y ago
Agree with you. I have a ~4k touchbar Macbook Pro and do HPC type of stuff all day... and I rarely consider doing anything more than a toy simulation or computational experiment on my laptop. When I previously used a Macbook Air, my toy PoC stuff ran pretty must just as well. Anything else either quickly gets offloaded to one of my group's servers with 64-80 threads and 100s of GB of RAM or, if it's even bigger, to the local HPC cluster with 1000s of CPUs. Even without access to existing resources: I'm having a hard time imagining a situation where, if I have a fixed budget of ~3-5k to do computational work, it'd be better for me to spend that on a top-end MBP than a two-piece solution of laptop+linux server.
- Fomite 8y ago
  A couple months ago, building projections for a rabies elimination program while sitting in a hotel bar in Zambia, I was pretty pleased I wasn't relying on a server. Edge case, sure, but shrug
  - et2o 8y ago
    Haha fair enough. How's the rabies situation there? Are people still doing multicompartment SIR models? That kind of stuff is hard to parralelize.
    - Fomite 8y ago
      It's still a big deal, but getting better - my university's running a decently successful program in Kenya. And yeah, people still use compartmental models pretty heavily. A lot of my stuff is stochastic (it's about the beginning or end of epidemics, or in small populations) so that's nicely parallel.
Fomite 8y ago
On the other hand, I do this all the time on my laptop, and have found having some more RAM or a couple more cores to be something that would make my life actively easier.

by qubex 8y ago

et2o 8y ago

Agree with you. I have a ~4k touchbar Macbook Pro and do HPC type of stuff all day... and I rarely consider doing anything more than a toy simulation or computational experiment on my laptop. When I previously used a Macbook Air, my toy PoC stuff ran pretty must just as well. Anything else either quickly gets offloaded to one of my group's servers with 64-80 threads and 100s of GB of RAM or, if it's even bigger, to the local HPC cluster with 1000s of CPUs. Even without access to existing resources: I'm having a hard time imagining a situation where, if I have a fixed budget of ~3-5k to do computational work, it'd be better for me to spend that on a top-end MBP than a two-piece solution of laptop+linux server.

Fomite 8y ago
A couple months ago, building projections for a rabies elimination program while sitting in a hotel bar in Zambia, I was pretty pleased I wasn't relying on a server. Edge case, sure, but shrug
- et2o 8y ago
  Haha fair enough. How's the rabies situation there? Are people still doing multicompartment SIR models? That kind of stuff is hard to parralelize.
  - Fomite 8y ago
    It's still a big deal, but getting better - my university's running a decently successful program in Kenya. And yeah, people still use compartmental models pretty heavily. A lot of my stuff is stochastic (it's about the beginning or end of epidemics, or in small populations) so that's nicely parallel.

Fomite 8y ago

On the other hand, I do this all the time on my laptop, and have found having some more RAM or a couple more cores to be something that would make my life actively easier.