Computer Recommendations

Computer Recommendations

Computer Recommendations for Image Analysis #

There are many considerations when building/buying a computer for image analysis. The two main considerations are:

  1. The type of data you will be analyzing
  2. The operating system that supports the software you are going to use.

The type of data to be collected will be the largest factor in the need for computational hardware. If you are going to analyze a single 2-D widefield image, even a older laptop will be capable of using ImageJ to process this file. However, if you collect 3-D to 5-D datasets, you can quickly find your self needing more hardware. There is no perfect computer. There is, however, a range of configurations that will work for most data.

The range of computers for image analysis go from a powerful laptop to a set of connected servers or “cluster.” The following specifications focus on the middle ground of a useable workstation that can be used in a lab setting. Each section is broken up into three recommendation levels. The Good section is lest expensive and should work with small to medium 5-D confocal datasets. The Better section is a good middle ground. The Best section describes hardware needed for large datasets from microscopes like the SiMView and the MOSAIC.

The right computer from you might be a some combination from each of these categories.

CPU #

  • Good - Fast CPUs (4GHz or faster)
    • Small processing in Fiji/ImageJ
  • Better - Single CPU with a lot of cores (12 or more)
    • Larger datasets using parallel processing
  • Best - Multiple CPUs with a lot of cores (24 or more each)

GPU #

  • Good - Older NVida graphics card list of GTX/RTX
    • Even the smallest older NVida graphics cards are able to help with rendering speeds
    • NVida is the only company that supports CUDA (a requirement for Hydra Image Processor)
  • Better - One or more RTX graphics card (current RTX 2080)
    • Some programs like Hydra can make use of multiple graphics cards for processing.
    • 3D+ visualization will benefit from multiple graphics cards only when tied together with SLI
  • Best - Multiple Quadro class graphics cards (curret Quadro RTX 8000)
    • The largest memory graphics cards are typically the Quadro class with over 32 Gigabytes of memory

Memory #

Programs that can load all of the data in memory will run faster than having to go to the hard drive for data. Getting as much memory as possible will always help image analysis.

  • Good - 64 Gigabytes
  • Better - 128 Gigabytes
  • Best - As much as you can afford (~One Terabyte)

Data Storage #

With all other tings, there are tradeoffs in storage solutions. The two opposing forces here are speed and size. Fast storage is typically small and much more expensive per unit storage. We recommend having a hierarchical storage solution. Use a fast hard drive for working on a dataset and a large storage solution for longer term saving.

Primary (working) #

  • Good - 8 Terabyte 7200 RPM Traditional Hard Drive
  • Better - 2+ Terabyte Solid State Drive (SATA)
  • Best - 1+ Terabyte NVMe Flash Drive

Secondary (storage) #

  • Good - 8+ Terabyte 7200 RPM Traditional Hard Drive
  • Better - Raid 5 or 6 Array of Hard Drives
  • Best - Network Attached Storage (NAS) or Local Server Storage


Last modified Apr 20, 2020