Forum:Bioinformatics Workstation Suggestions

Question

[akh22]

Hi

We will invest some money in building a lab workstation dedicated to run bioinformatics analysis suites for RNAseq and scRNAseq, and I am curios to see what sort of configurations most use. We are looking at Dell XPS with i9 (whopping 8 cores), 64GB DDR4, NVIDIA® GeForce® GTX 1080 8GB GDDR5X and 2TB M.2 PCIe NVMe SSD (Boot) + 2TB 7200RPM 3.5" SATA HDD (Storage). And I will add Dell's TB3 PCIe. Is this overkill or not sufficient?

Thanks in advance

[colindaven]

2TB of storage will fill up very quickly. I'd get 6-8TB if possible, or a couple of ext HDs of that size.

My desktop has 32 GB of RAM these days, so a workstation would benefit from 128GB I feel. The 8 cores might become 16 pretty quick if you turn on hyperthreading if available.

The only proven benefit I've seen so far of a GPU is in image analysis or Nanopore basecalling. I'd like one, but not sure I'd save on the RAM to get one.

Answer 1

[jared.andrews07]

The 1080 is likely wasted money unless you're doing protein structure modeling or something along those lines.

64 GB memory should be sufficient for most scRNA datasets <100k cells. And there are ways to do larger analyses by using the loom format, etc.

What you have sounds fine otherwise. RAM and harddrives are easy to upgrade/replace later as well.

Answer 2

[curious]

We use a workstation with 6 cores (12 threads) and 64GB RAM for our human RNA-seq samples. It has been great for us (most intense thing we use is STAR). I can't speak for scRNA-seq.

Answer 3

[Dave Carlson]

Obviously much will depend on your budget and your specific research questions, but as @jared.andrews07 noted, the 1080 is probably not worth the money unless you have specific tasks in mind that utilize graphical processing. I would take the money you plan to put into the GPU and get something with more cores and/or RAM. If you plan to do any transcriptome assembly in particular, you might want to consider increasing the RAM as 64 GB may not cut it.

[akh22]

My plan initially was to build a Linux box with Ubuntu or KDE but nobody in my lab wanted to deal with command line stuff too much so I decided to build a Window machine instead. I was hoping I could use CUDA to take an advantage of GPGPU but that won't happen. Anyway, thanks for all the suggestions.