For the bioinformatic and the computational biology research what are the best and ideal PC or workstations and what should be the specifications and how much it will cost?
My personal thoughts:
OS: Linux. Some people use Macs but more bioinfo* software runs better on Linux. Use the flavor of Linux that your collaborators use or the one that you are used to.
You will probably want to run Windows in a VM for Office / Word / Endnote, etc.
Hardware / CPU: with the same $$$, you can get more slower CPU cores of fewer cores that are faster. With limited funds, I will usually go with higher number of slower cores, that make the machine more efficient when running VMs or otherwise multitasking.
You are likely going to use VMs to run guest OS configurations optimized for specific tasks. Some solutions run seamlessly under Ubuntu, some Fedora, some Centos, etc. Developers will often provide VM images that are preconfigured for their software.
Also: Usually top of the line, fastest CPUs are very expensive: two 3.6Ghz cores may cost more than twelve 2.8GHz cores with same cache and instruction set, so you get more power for the buck with the slower ones, especially with applications that scale well when parallelized.
Hardware / Memory: Generally, the more the better. Size more important than speed. You can usually live with your application running 15% longer, but you can't do anything if it won't run because it can't fit in the memory. Some problems require large amounts of directly-addressable / shared memory. If you know that you will be analyzing huge datasets that can't be efficiently partitioned (like large Hi-C maps) then get enough RAM to fit your calculation. But if you'll be running large-memory tasks only occasionally (say 1% of the time) and you have access to a big machine, e.g. in a collaborator's lab, then you don't have to get 512 GBs of RAM in your desktop. Get 32 or 64 that will be sufficient for 99% of your work.
Hardware / Disk:
A combination of fast SSD's (system, swap, quick access data) and large HDDs (long-term storage). Mirror your SSDs, organize the HDDs in RAID6. I'd go for software RAID rather than hardware : you have more control, and you can recover data more easily in certain situations, e.g. when the controller fails, not just the disks.
Hardware/Power Supply:
Get the bigger one. 750 watts or more.
Display: Get a good one if you can - You may have a bunch of windows open. 32 inch 4k if you can budget it? I've seen 40-inch 4k TVs used as a less expensive alternative.
Graphics: something that will support your monitor's resolution Doesn't have to be top of the line. If you plan to use GPU for computing, get a second one depending on what applications you will be using it for (balance between more threads, more memory or faster clock). I think CUDA/NVidia is supported by more applications than their competition.
This is a vague question and there is no best for such questions. And for some analysis, even supercomputer are slow. Of course you can not have supercomputer. But any system configuration would depend on what you want to do with that.
Software wise most people use Linux or unix like interface. Most of the software and OS are free on this platform.
For the hardware, first you need to decide for your budget yourself in your country. It would be stupid to tell you about price. And go for the best configuration according to your price range. The configuration can be suggested which is a good processor (more GHz), RAM (16 GB or more), storage depends on your data and your usage. Optional is a GPU unit (graphics card)
My personal thoughts:
OS: Linux. Some people use Macs but more bioinfo* software runs better on Linux. Use the flavor of Linux that your collaborators use or the one that you are used to.
You will probably want to run Windows in a VM for Office / Word / Endnote, etc.
Hardware / CPU: with the same $$$, you can get more slower CPU cores of fewer cores that are faster. With limited funds, I will usually go with higher number of slower cores, that make the machine more efficient when running VMs or otherwise multitasking.
You are likely going to use VMs to run guest OS configurations optimized for specific tasks. Some solutions run seamlessly under Ubuntu, some Fedora, some Centos, etc. Developers will often provide VM images that are preconfigured for their software.
Also: Usually top of the line, fastest CPUs are very expensive: two 3.6Ghz cores may cost more than twelve 2.8GHz cores with same cache and instruction set, so you get more power for the buck with the slower ones, especially with applications that scale well when parallelized.
Hardware / Memory: Generally, the more the better. Size more important than speed. You can usually live with your application running 15% longer, but you can't do anything if it won't run because it can't fit in the memory. Some problems require large amounts of directly-addressable / shared memory. If you know that you will be analyzing huge datasets that can't be efficiently partitioned (like large Hi-C maps) then get enough RAM to fit your calculation. But if you'll be running large-memory tasks only occasionally (say 1% of the time) and you have access to a big machine, e.g. in a collaborator's lab, then you don't have to get 512 GBs of RAM in your desktop. Get 32 or 64 that will be sufficient for 99% of your work.
Hardware / Disk:
A combination of fast SSD's (system, swap, quick access data) and large HDDs (long-term storage). Mirror your SSDs, organize the HDDs in RAID6. I'd go for software RAID rather than hardware : you have more control, and you can recover data more easily in certain situations, e.g. when the controller fails, not just the disks.
Hardware/Power Supply:
Get the bigger one. 750 watts or more.
Display: Get a good one if you can - You may have a bunch of windows open. 32 inch 4k if you can budget it? I've seen 40-inch 4k TVs used as a less expensive alternative.
Graphics: something that will support your monitor's resolution Doesn't have to be top of the line. If you plan to use GPU for computing, get a second one depending on what applications you will be using it for (balance between more threads, more memory or faster clock). I think CUDA/NVidia is supported by more applications than their competition.
If you have moderate budget and it will be used by small team then i will suggest you the HP Workstation having Linux operating system with the following specification, (Keeping in mind that almost all modern algorithm rely on GPU from performance gain). It will be roughly cost around $2400.
HP Z4 G4 Workstation
Intel® Xeon® W processor
32 GB memory
1 TB SSD storage
NVIDIA Quadro P2000
Note:
Quadro is a good graphics card
in your case go for SSD as it will give much performance boost in comparison with HDD (SATA)
- Badges
- Science topic
- Similar topics
- Bioinformatics
- Bioinformatics and Computational Biology