Opinion on Hardware

[colin_c]

Hi All,

I'm currently in the process of buying a workstation for RNA-seq expression analysis (mRNA, miRNA, lnRNA snoRNA).

I was wondering what people thought of the following hardware:

HP Z820 Workstation
2 Xeon E5-2687Wv2 3.40Ghz 25MB 1866 8C
256GB DDR-3-1866 (8 X 32GB) 2CPU LR RAM
4 x 2TB 7200 SATA
RAID 5

Thanks!
Colin

[dpryan]

Looks good. The RAM is a bit overkill for what you're doing at the moment, but of course it's better to have it and not need it than to need it and not have it

[colin_c]

Looks good. The RAM is a bit overkill for what you're doing at the moment, but of course it's better to have it and not need it than to need it and not have it

Thanks for the reply! Is there an arguments for decreasing the RAM to 128GB and increasing the number of cores to 24?

Colin

[dpryan]

128 is still enough for things like STAR, which might benefit from the increased core availability (though I've never benchmarked that). If you're planning to use DEXSeq or anything like that, then the increased number of cores would be beneficial. I can't think of anything normally used in RNAseq that'll need more than 128GB memory, though if you're going to need to assemble the transcriptomes (i.e., you're not using one of the common model organisms) then perhaps that'll need it (it's not something I ever need to do, so I can't provide any advice there).

[colin_c]

128 is still enough for things like STAR, which might benefit from the increased core availability (though I've never benchmarked that). If you're planning to use DEXSeq or anything like that, then the increased number of cores would be beneficial. I can't think of anything normally used in RNAseq that'll need more than 128GB memory, though if you're going to need to assemble the transcriptomes (i.e., you're not using one of the common model organisms) then perhaps that'll need it (it's not something I ever need to do, so I can't provide any advice there).

The majority of my work involves a non model organism so transcriptome assembly is a possibility.

In terms of the number of cores, I guess the trade off is run time?

Thanks!
Colin

[dpryan]

Yeah, more cores will generally decrease run time (to a point, at least).

[colin_c]

Yeah, more cores will generally decrease run time (to a point, at least).

Thanks for you help! Much appreciated!

[GenoMax]

SATA disks probably would become the bottleneck with a large number of cores. You should probably stick with your original configuration but you may want to include a good (intel) SSD for your OS.

[Brian Bushnell]

If you want to do assembly of eukaryotes, I do not recommend increasing core-count at the expense of memory. 256g will already be limiting in some scenarios, depending on the size of the organisms you're working with. Many assemblers do not scale well with large numbers of processors, anyway, though mappers should scale linearly.