Workstation configuration

Hello there!

I’m not sure this is the right category to ask this question under, but I would guess that the mods are able to move it around.

We’re building a workstation in the lab and I have a few questions in regard to hardware, and I rely on the brains here that know the architecture better than me to help out :slight_smile:
Our primary workloads are basic Fiji tasks (segmentation, particle tracking…) and that on very large images and some image analysis on some proprietary softwares that mostly rely on GPU.

  1. When choosing a CPU, should we give priority to the number of cores or to the frequency? I am well aware that different workloads and different platforms ask for different things.
  2. Since there will be a moderate amount of work using CLIJ but also on the aforementioned proprietary softwares, I am wondering wether it makes any sense to buy one of the Quadro cards or should we save the buck and go with the usual GeForce lineup?
  3. Is there a major influence of the RAM frequency on the performance? In other words - if we choose to go with AMD platform should we invest into 3600MHz RAM that is optimal for the Infinity Fabric and what not or will that ceiling never be met with 3000MHz RAM?

Cheers and thanks for your input!

See my answer to a previous post with a similar question.
To answer your questions more specifically:
1- More cores will be beneficial for multithreading-enabled software which should be the case for Fiji. In multithreading applications, clock speed has less influence and higher CPU frequency doesn’t always mean faster. Go for highest CPU frequency with few cores only if you’re going to run a lot of single threaded applications.
2- Check the requirements/recommendations for the proprietary software. The vendors usually provide this information.
3- Higher RAM frequency doesn’t always translate in faster processing because speed also depends on the RAM latency and how it plays with the CPU and other components. RAM is often not the bottleneck in data processing (unless you don’t have enough). For images, the amount of GPU RAM may have a bigger impact.

1 Like