SEQanswers

Go Back   SEQanswers > Bioinformatics > Bioinformatics



Similar Threads
Thread Thread Starter Forum Replies Last Post
changes to SAM Specs (Sept 7, 2011) BAMseek Bioinformatics 2 10-07-2011 03:46 PM
PacBio RS latest specs anyone? markwatson Pacific Biosciences 10 09-23-2011 08:23 AM
PubMed: Targeted next-generation sequencing for routine clinical screening of mutatio Newsbot! Literature Watch 0 09-14-2011 02:00 AM
which snp calling routine for tophat? rcorbett Bioinformatics 0 08-11-2011 09:48 AM
Bioinformatics Computer: List your specs peromhc Bioinformatics 14 10-09-2009 10:21 AM

Reply
 
Thread Tools
Old 06-16-2008, 09:01 PM   #1
ECO
--Site Admin--
 
Location: SF Bay Area, CA, USA

Join Date: Oct 2007
Posts: 1,358
Post Desktop computer specs for routine analyses? Post your rig

I hope to have some solid data to analyze soon, so I thought it would be a good question to pose to the community.

What specs would you recommend for a machine that could competently deal with common next gen data analysis? Resequencing and de novo assemblies, tag counting, chip seq, whole transcriptome...is there a desktop solution fast enough yet? Dual-quad cores?

I hope to have access to a proper bioinformatics team, but I enjoy learning and tinkering with my own data. I have no concept of the computing power necessary to do even a resequencing alignment to reference.

Disclaimer: I'm obviously not experienced on the dry electrically charged side of the bench...so be gentle. I have some linux/perl knowledge and can code a bit.
ECO is offline   Reply With Quote
Old 06-16-2008, 09:22 PM   #2
ScottC
Senior Member
 
Location: Monash University, Melbourne, Australia.

Join Date: Jan 2008
Posts: 246
Default

Here's my 2c. I don't claim to be any expert, but this is what I considered when I bought my PC. I'm talking standard desktop equipment here.

Firstly, get the best you can afford/manage to justify! None of it will go to waste.

But to be more precise...

It really depends on the software you'll run. A lot of new commercial software that you'll pay thousands of dollars for is still unable to use more than one processor (or core, for that matter). In that case, you're better off getting a processor with a two cores that run at faster speeds, rather than a quad core with cores at lower speeds. But then, when you want to run some software that can use more than one core, maybe you should get the quad. Templated assemblies can be done on a desktop PC fairly quickly depending on your hardware, software and genome size (i.e. bacterial assemblies, CLC Workbench, Intel 8500 in 10 minutes or so) but de novo assemblies will take a long time no matter what you're using (many hours on that same system, same genome, same software).

I'd recommend 8GB of RAM as a minimum. It's cheap now, and lots of things will use it. You have to run a 64 bit OS for that, of course (or trick 32bit windows into using a bit more). Lots of applications will use if not require it. If you match the speed of your RAM to the FSB of your processor, then 666MHz is fast enough, but it doesn't really matter if you get faster than that. It's pretty cheap.

The video card doesnt matter, as long as it's fast enough to move the windows around at high resolution (and as long as you're not planning on doing any of those GPU-assisted assemblies !)

The screen... well more is always better. As far as the amount of info you can display on your screen, it's resolution that's important, not screen size (i.e. you can get a screen that's 2" bigger but has the same resolution - you're not getting extra space, you're just able to sit further away from it because the pixels are bigger).

Disk space is pretty cheap these days too. Two 1TB drives is reasonable, and if you want, you can mirror them for extra security.

Lastly, I reckon it's important to get a nice keyboard and mouse. They're underrated. If you're sitting in front of the damned thing all day, you might as well get comfortable mouse and keyboard!

Last edited by ScottC; 06-16-2008 at 09:28 PM.
ScottC is offline   Reply With Quote
Old 06-16-2008, 11:33 PM   #3
sparks
Senior Member
 
Location: Kuala Lumpur, Malaysia

Join Date: Mar 2008
Posts: 126
Default

Totally agree with Scott. I have Dual core AMD Athlon with 8G RAM and dual 500Gbyte drives, the best quality bits are the display (1680*1050) and the keyboard mouse (Microsoft wireless). I run SUSE 10.3 64-bit.
I'd like to upgrade to the new AMD Phenoms so I can try using 1Gbyte pages but I'm waiting for the 45nm versions.
sparks is offline   Reply With Quote
Old 06-29-2008, 06:30 PM   #4
Torst
Senior Member
 
Location: The University of Melbourne, AUSTRALIA

Join Date: Apr 2008
Posts: 275
Default

Quote:
Originally Posted by ScottC View Post
If you match the speed of your RAM to the FSB of your processor, then 666MHz is fast enough, but it doesn't really matter if you get faster than that.
I have to clarify this point. The RAM on Intel systems is double-pumped which means that 667 Mhz RAM will need a CPU with a 1333 Mhz front side bus to be fully utilised. Desktop CPUs with 1600 Mhz FSB will be released this quarter, so make sure you purchase 800 Mhz RAM if you get one of those (it's the same price as 667 anyway).

--Torsten
Torst is offline   Reply With Quote
Old 06-29-2008, 06:55 PM   #5
ScottC
Senior Member
 
Location: Monash University, Melbourne, Australia.

Join Date: Jan 2008
Posts: 246
Default

Yeah, sorry, I should have been more clear on that... I meant 'match' as in match the appropriate RAM speed for the FSB speed of your processor, not literally 'match the speed'... it is DDR afterall!

Scott.
ScottC is offline   Reply With Quote
Old 10-14-2008, 05:44 AM   #6
rs705
Junior Member
 
Location: USA

Join Date: Sep 2008
Posts: 6
Default

What are the applications you need to be able to use? Prokaryotics, eukaryotics, resequencing?? Depending on the project size and what you plan on doing there are a wide range of requirements we are being told.
rs705 is offline   Reply With Quote
Old 10-16-2008, 01:33 PM   #7
ShaunMahony
Member
 
Location: University Park, PA

Join Date: Apr 2008
Posts: 27
Default

Try to get even more RAM if you want to use the aligners that hash the genome (as opposed to ELAND that hashes the reads). I'm thinking about something like Novoalign, which handles indels & quality scores, but has a memory overhead of ~12GB for vertebrate genomes.
ShaunMahony is offline   Reply With Quote
Old 10-16-2008, 07:24 PM   #8
zee
NGS specialist
 
Location: Malaysia

Join Date: Apr 2008
Posts: 249
Default

A server with 8 cores and at least 16GB of RAM will see you in good shape for the majority of software packages out there.
For reads, alignments, assemblies, etc, having storage + backup of ~ 10 Tb should be ample for most work on sequence data.
zee is offline   Reply With Quote
Old 10-16-2008, 07:41 PM   #9
ScottC
Senior Member
 
Location: Monash University, Melbourne, Australia.

Join Date: Jan 2008
Posts: 246
Default

The specs are getting bigger and bigger as the posts go on through this thread, and it's only 5 months old

I think the original poster was asking about desktop solutions, though.
ScottC is offline   Reply With Quote
Old 10-19-2008, 11:12 PM   #10
zee
NGS specialist
 
Location: Malaysia

Join Date: Apr 2008
Posts: 249
Default

I think that's reasonable. It should be quite cost effective to get a desktop with 8 cores and 16Gb RAM by the end of this year.
Perhaps we should chart a desktop specification by cost where 16 cores and 32Gb RAM would be the most expensive. Anything less than that would be cheaper.
zee is offline   Reply With Quote
Old 10-19-2008, 11:29 PM   #11
ECO
--Site Admin--
 
Location: SF Bay Area, CA, USA

Join Date: Oct 2007
Posts: 1,358
Default

So to update this with my experience so far...I've been doing most of my work (bacteria and arabidopsis) using MAQ on a dual core (E6700) with 8GB of RAM, and I'm reasonably happy.

I can align ~30M reads to A.thaliana (120Mb) on a single core in << 1 hour. To coli, significantly shorter.

SOAP seems nearly as quick but it's not solid friendly so I haven't continued to run as many things through it.

SHRiMP is very easy to run with SOLiD data, but obviously is muuuuuch slower, so much so that I don't use it at all, seems like a compute-farm only solution (2M reads against arabidopsis took something like 16 hours...so 4 samples with 30M reads each is impractical).
ECO is offline   Reply With Quote
Old 10-20-2008, 08:31 AM   #12
dcfargo
Member
 
Location: Chapel Hill

Join Date: Aug 2008
Posts: 22
Default

Hal9000

We have an 8 core 32 gig machine for 'personal' use.
dcfargo is offline   Reply With Quote
Reply

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off




All times are GMT -8. The time now is 05:05 AM.


Powered by vBulletin® Version 3.8.9
Copyright ©2000 - 2020, vBulletin Solutions, Inc.
Single Sign On provided by vBSSO