Hello everyone!
I am planning to buy and set up at home work station for simulations.
Could you recommend which CPU (choosing between ryzen 7 3700x and
intel core i7) and graphics card to buy ?
I am using molecular dynamics for polymer system modeling.
Thanks)
Hi Alexandra, what matters the most is figuring out ahead of time what force fields you will be using, and whether they support GPU-accelerated versions using the KOKKOS or GPU packages. Then familiarize yourself with the strengths and requirements of each of these packages: https://lammps.sandia.gov/doc/Speed_packages.html
One important factor for GPU-accelerated vs. CPU is whether you can afford running in single or mixed precision. Run some quick tests on a CPU to see if it makes a difference e.g. in mechanical properties before deciding whether you need double precision (which consumer GPUs are not as strong with).
Giacomo
It is not that simple. These days you cannot just say “buy this not that” and always be correct. You have to go back 20 years or so, where things were this simple (and you had fewer choices to begin with).
What is most important is to configure a balanced system, i.e. to not spend extreme amounts of money to get the fastest component (e.g. CPU), and then save elsewhere (e.g. amount and speed of RAM, or storage). you also want to consider reliability and quality. a fast computer is useless if it crashes regularly under load.
furthermore, you need to factor in that you are likely not only running simulations but also will be doing analysis. that usually requires much more RAM (e.g. when loading long trajectories) and is also affected by the GPU power. but then again, a very high-end GPU is only justified if you plan to use it for GPU computing and then you have to research very carefully what is applicable and whether you may even be better off with 2 GPUs. at that point you may also need to worry about cooling and power consumption (but then again, you will have a nice warm home office in winter…).
if this sounds like there are more chances to make mistakes than there are to make the right choice that this is not far from reality. however, you can also see through the looking glass the other way and realize that it is impossible to get the perfect hardware for your budget unless you become an expert in several aspects of computer hardware and then just accept that your choice will not be perfect and go for something reasonable that is not too extreme in any direction and well within your budget. perhaps leave some space you can later double up on RAM modules and storage devices. I have found that a machine with lots of RAM and a moderate CPU is much longer useful than a machine with a very fast CPU but little RAM. or to quote a former colleague of mine, “RAM in a computer is like displacement in a car engine: you cannot have enough of it”
Axel.