I have a home brew (read, Citizen Science) 32 core intel machine in my garage! I am currently using it for low throughput coarse-grain molecular dynamics simulations and some basic rna seq calculations. As my work has turned more towards cheminformatics with a bit of bioinformatics thrown in, I have become increasingly interested in buying a GPU to be used by all three areas. My problem is that I know nothing about GPUs with bio/chem informatics tools like I do with computational chemistry tools. If I was to add a poor GPU (NVidia £50-£100), it will bottle neck the CPU performance whilst performing computational chemistry calculations. Essentially, if you don't have much cash to spend, aim for a GPU that performs the non-bonded interactions (GPU based) at least as well as the bonded interactions (CPU based), otherwise there is no benefit.
However, I don't have a lot of disposable cash and I don't know how the bioinformatics community has embraced GPUs. Compared with a pure 2.33ghz 32 core machine, what GPU spec will I begin to see any benefit from having? I have around £500 - £600 to spend. I might add, I don't have any particular science/topic in mind, as generic as saying GWA, ran seq and SNP related software.
Thanks Anthony
I only know a handful tool that utilizes GPU. Example: https://nvlabs.github.io/nvbio/nvbowtie_page.html