Advice Needed: Best GCP VM Configuration for Heavy Blender Fluid Simulation & GPU Rendering

Hi everyone,
I’m currently running Blender on a Google Cloud Platform (GCP) VM, mainly for heavy liquid (Mantaflow) simulations and Cycles GPU rendering. However, the performance is far slower than expected.

Even though I’m using a high-end underlying rig, the GCP VM feels extremely slow for tasks that should run much faster—baking, simulation, and GPU rendering all take a very long time, and viewport performance is poor.

Before I upgrade or change instances, I need advice from anyone with experience running Blender on GCP.

What is the ideal GCP VM configuration for Blender?

Looking for recommendations on:

1. Best Machine Type

  • n1, n2, n2d, c2, c3, or a2?
  • Number of vCPUs that Blender benefits the most from (for simulation, baking, physics engine).

2. RAM Requirements

  • How much memory is optimal for high-resolution fluid sims?
  • Is 32–64 GB enough, or should I go higher?

3. GPU Choice

  • T4 vs L4 vs A100 vs H100?
  • Is a single powerful GPU better than 2 weaker ones in GCP for Cycles?
  • Any experiences with GPU passthrough or driver issues?

4. Disk Performance

  • Should I switch to Local SSD for caching & baking?
  • Is Persistent Disk too slow for heavy Blender projects?

5. Known Bottlenecks

Has anyone run into these issues before on GCP?

  • Slow fluid baking
  • Poor GPU utilization
  • Slow viewport updates
  • RAM spikes during Mantaflow
  • CPU clock speed bottleneck in VM vs bare metal

My current problems:

  • Simulation times are much longer than they should be.
  • GPU rendering feels underutilized.
  • Overall Blender responsiveness is slow.

So I’m looking for recommendations on which exact GCP setup works best for Blender simulations + rendering.

Any help or shared experience would be greatly appreciated!