Seqanswers Leaderboard Ad

Collapse

Announcement

Collapse
No announcement yet.
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #16
    Something to consider too, is that the 4600v2 is due out in 2014 Q1. So, depending on when you have to spend your money by, or when you need to get the work started, it may be worth it to wait a few months if you're set on the 4600 series. You'll probably see at least 2 additional cores, another .1-.2 GHz and maybe a little more range in the turbo boost at the same price levels going from 4600 to 4600v2. So the 4620 might be more like a 10-core 2.3GHz with turbo to 2.8.

    Comment


    • #17
      I'll just add my $0.02 worth -

      Consider increasing your storage by several fold. That machine may last you several years but your data storage needs will grow over time. You also need to consider data backups and the hardware you may need for that. If you have access to a cluster or some other shared big memory machine then you could always do your biggest jobs there. Its convenient to be able to do all of your work on your own machine but its really inconvenient to have to continually shuffle files around because you main file system is too small. Its also really inconvenient to lose data because of a hardware failure. Make sure you have reliable off-site backup.

      Comment


      • #18
        Thanks for your input, Wallysb01. I will certainly not go with RAID5. Actually, the colleague's machine I have been using has both a RAID1 and RAID0 array as you describe, and it has worked nicely for Trinity. I will certainly up the storage, likely with a few 4TB 3.5" SATA drives.

        You provide a valid reason for consideration of 2x 2600 v2 instead of 4x 4600, especially since the higher end 2600 V2s (e.g. the E5-2697V2) are available with up to 12 cores. This then becomes a 24 faster core vs. 32 slower core situation...

        Once again, thank you everyone.

        Comment


        • #19
          Hello again,

          I have yet another consideration. Looking at various configurations, the lower price of the AMD Opteron CPUs quickly becomes apparent, in comparison to Xeon 2600v2 and 4600 CPUs.

          Is there any reason to shy away from an Opteron 6300 series CPU for the aforementioned applications?

          I can go to a 4 CPU system, be able to get 512 GB RAM at a more reasonably price (as I'd able to use 32 * 16 GB DIMMs, instead of the pricier 16 * 32 GB), and still come in at a far less expensive machine...

          And have no fear, I've heeded the advice and I'm going to several 4 TB drives. One question there - is RAID cache in the form of an LSI Nytro MegaRAID particularly useful for assembly and mapping?

          Thanks again.

          Comment


          • #20
            The Opterons are substantially slower than the Intel equivalents. Also, don't get fooled by the core count the Opterons boast. For floating point calculations (which is much of bioinformatics) those cores are cut in half. That's because the thing to count is the module for floating point, and each module has two cores. Its a bit like Intel hyper threading.

            That said, it is a cost effective way to get a high memory machine. So, if you really want 512GB of RAM for <$10K (or there abouts), AMD is generally what you'll be looking at. But be prepared for up to 50% slower performance (single/lowly threaded stuff will especially suffer). Those 6300s are just plain old too. AMD didn't really move forward much from 6200 to 6300, so its the same basic technology from about 2 years ago. Supposedly the Warsaw series of workstation CPUs from AMD will be coming out early 2014, but who really knows if that's true, and even if they show up, who knows how competitive they will be with E5-2600 v2s.

            As for the other question, I don't have any experience with that kind of set up, so hopefully someone else can jump in.

            Comment


            • #21
              Originally posted by Wallysb01 View Post
              For floating point calculations (which is much of bioinformatics) those cores are cut in half.
              Much of bioinformatics is floating point??? I'll disagree with this. IMHO much of bioinformatics is string manipulation.

              That said, the Opterons are single-CPU slower than Intel chips. Offset by more cores and larger memory that can be used as a ram-disk.

              My main machines (mainly because there are more of them available to me) are 4-year-old AMD-based machines that have the same number of cores (16) but with more memory than the most recent Intel machines that I can also use. For what it is worth, the newer Intel machines are about 3 times faster than my older AMD machines in my normal processing (demultplexing, filtering, de-novo assembly, de-novo transcriptome).

              As usual it comes down to a cost basis. Would I want two AMD-based slower machines or one Intel-based fast machine? Given the number of samples I process (as a core lab) more is better than faster. For your work the balance may be different.

              Comment


              • #22
                Originally posted by westerman View Post
                Much of bioinformatics is floating point??? I'll disagree with this. IMHO much of bioinformatics is string manipulation.
                Well, to be truthful, many are both. Think of sequence alignment for example. You're manipulating strings, but then coming up with alignment scores. So, you'll still end up choking on those floating points if you tell the AMD machine to try to cram 32 threads of floating point calculations into 16 modules.

                That said, the Opterons are single-CPU slower than Intel chips. Offset by more cores and larger memory that can be used as a ram-disk.
                Its not more cores if you count Opteron modules vs Intel cores. Regardless, you could do some benchmarks, as I have with 2xE5-2630s vs 4x6320s and for many things the Intel machine is much faster. The AMD machine is only roughly even in things that scale well past the 12 cores on the 2630s and are not utilizing much floating point calculations. But despite having 32 cores vs 12 cores, the AMD machine never gains much of a lead on the Intel machine. Of course that AMD machine has 512GB of RAM and the Intel one has 128GB, so some jobs just won't fit on that Intel machine.

                My main machines (mainly because there are more of them available to me) are 4-year-old AMD-based machines that have the same number of cores (16) but with more memory than the most recent Intel machines that I can also use. For what it is worth, the newer Intel machines are about 3 times faster than my older AMD machines in my normal processing (demultplexing, filtering, de-novo assembly, de-novo transcriptome).
                I suppose de bruijn graphs are all strings, as is demultplexing and a good chunk of filtering (depending on type), so for you AMD does appear to make more sense. For most projects though, de novo assembly and read processing are just the first of many steps. Things like genome annotation pipelines, RNA-seq DE analysis, multiple sequence alignment and conservation analysis are going to have more floating point calculations than you have.

                As usual it comes down to a cost basis. Would I want two AMD-based slower machines or one Intel-based fast machine? Given the number of samples I process (as a core lab) more is better than faster. For your work the balance may be different.
                That is certainly the balance. AMD does sit in a nice spot once you get past the 2x8/10 core Intel machines price wise (the 12 core 2600s are crazy expensive). Basically they hit the gap between workstations with Intel 2600s and 4600s. They offer more RAM than the 2600s (assuming similar price points) and are lower cost (by a lot) than the 4600s.

                Personally, I'd suggest the 2x2600s with 256GB of RAM, and for the odd job that can't fit in 256GB, get access to clusters with 512GB/1TB RAM machines, such as blacklight or mason, which are part of XSEDE, or DIAG, or any others you may be able to find. Paying all that money for 512GB or 1TB of RAM when you only need it for handful of jobs over the machine's lifetime is just kinda crazy, IMO anyway.

                Comment


                • #23
                  While it is interesting to configure/price/build a machine, administering it on a daily basis introduces another level of complexity. If one is primarily a biologist (with interest in bioinformatic analysis) then it may be best to leave the systems administration part to the pros (if your institution has central computing support use them).

                  A mis-configured/vulnerable server can be a real threat (especially a powerful one with configuration such as that being discussed in this thread) for not only you but many others.

                  Comment


                  • #24
                    Originally posted by GenoMax View Post
                    While it is interesting to configure/price/build a machine, administering it on a daily basis introduces another level of complexity. If one is primarily a biologist (with interest in bioinformatic analysis) then it may be best to leave the systems administration part to the pros (if your institution has central computing support use them).

                    A mis-configured/vulnerable server can be a real threat (especially a powerful one with configuration such as that being discussed in this thread) for not only you but many others.
                    Thank you for your concern. The machine will be administered by (and located in) our IT department. I was presented with options by our IT department, but they are not computational biologists, and as such I wanted to submit some of those options to the community here. I have also conferred with colleagues, looking at their configurations, and that's largely what you're seeing reflected here - Opteron vs. Xeon options, etc.

                    I figured since I will be the primary user of the machine, it was best to become versed in the options now, rather than regret my ignorance 6 months down the line!

                    Comment


                    • #25
                      Originally posted by anth View Post
                      Thank you for your concern. The machine will be administered by (and located in) our IT department. I was presented with options by our IT department, but they are not computational biologists, and as such I wanted to submit some of those options to the community here. I have also conferred with colleagues, looking at their configurations, and that's largely what you're seeing reflected here - Opteron vs. Xeon options, etc.

                      I figured since I will be the primary user of the machine, it was best to become versed in the options now, rather than regret my ignorance 6 months down the line!
                      Thank you for clarifying that. People may tend to overlook that important aspect i.e. need for ongoing management of the workstation/server.
                      Last edited by GenoMax; 10-23-2013, 02:24 PM.

                      Comment

                      Latest Articles

                      Collapse

                      • seqadmin
                        Strategies for Sequencing Challenging Samples
                        by seqadmin


                        Despite advancements in sequencing platforms and related sample preparation technologies, certain sample types continue to present significant challenges that can compromise sequencing results. Pedro Echave, Senior Manager of the Global Business Segment at Revvity, explained that the success of a sequencing experiment ultimately depends on the amount and integrity of the nucleic acid template (RNA or DNA) obtained from a sample. “The better the quality of the nucleic acid isolated...
                        03-22-2024, 06:39 AM
                      • seqadmin
                        Techniques and Challenges in Conservation Genomics
                        by seqadmin



                        The field of conservation genomics centers on applying genomics technologies in support of conservation efforts and the preservation of biodiversity. This article features interviews with two researchers who showcase their innovative work and highlight the current state and future of conservation genomics.

                        Avian Conservation
                        Matthew DeSaix, a recent doctoral graduate from Kristen Ruegg’s lab at The University of Colorado, shared that most of his research...
                        03-08-2024, 10:41 AM

                      ad_right_rmr

                      Collapse

                      News

                      Collapse

                      Topics Statistics Last Post
                      Started by seqadmin, Yesterday, 06:37 PM
                      0 responses
                      11 views
                      0 likes
                      Last Post seqadmin  
                      Started by seqadmin, Yesterday, 06:07 PM
                      0 responses
                      10 views
                      0 likes
                      Last Post seqadmin  
                      Started by seqadmin, 03-22-2024, 10:03 AM
                      0 responses
                      51 views
                      0 likes
                      Last Post seqadmin  
                      Started by seqadmin, 03-21-2024, 07:32 AM
                      0 responses
                      68 views
                      0 likes
                      Last Post seqadmin  
                      Working...
                      X