Hardware recommendations for a dedicated cryosparc master server

We’re thinking of investing in a dedicated cryosparc master server because the load on our current server is perpetually high. Any recommendations on how much RAM, how many cores to get? I’m assuming the master doesn’t need any GPU’s. The server probably also doesn’t need an SSD cache disk?

Also, is dedicating an entire machine to just the cryosparc master a waste of resources. Does it make sense or is it even possible to have master and worker nodes on the same machine with other workers available as well?

Hi @pgoetz,

Our recommended system requirements and other considerations are outlined at our guide. Here are some additional notes that may be helpful:


  • It’s best to avoid GPUs with less than 11GB VRAM because you may run into memory errors when processing data with larger box sizes and/or K3 super-resolution data in some cases (which is becoming more common)
  • It’s likely that you will be adding more than 2 GPUs to your workstation anyway, but if you are considering using cryoSPARC Live (real-time preprocessing and 2D/3D processing), it’s best to ensure your workstation has 3 or more GPUs. CryoSPARC Live information: https://cryosparc.com/live and https://cryosparc.com/docs/live
  • CUDA11 support (for Ampere-based GPUs like 3080 etc) is to be deployed soon in cryoSPARC. These new GPUs are likely to provide better performance per dollar than previous GPUs but it’s currently very hard to find them and systems for these are not yet well tested for thermal or power load


  • For a workstation, AMD Threadripper CPUs are currently the top choice in terms of performance and cost
  • Make absolutely sure you will have enough power from PSU and enough cooling (case and case fans) for the GPUs. Also be aware that a workstation with 4xGPU will almost definitely be very loud
  • Try to get 256GB CPU RAM if possible, for future proofing

For a fluid user-interface, having the master set up on it’s own machine is ideal. Unless you install the master and worker packages on a single workstation that is powerful (lots of CPU cores, very fast disk drives), the UI may feel sluggish. But also note that if you end up having separate master and worker machines, your system architecture becomes much more complex and becomes heavily dependent on a robust network connection.

It technically doesn’t, but storing your database on a very fast SSD will enable a more fluid UI experience.


Thanks that was extremely helpful. We can consider this issue resolved.

1 Like