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.
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.
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.
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.