byteLAKE’s CFD Suite (AI-accelerated CFD) — recommended hardware for AI training at the Edge (part 3/3)
In previous parts of this miniseries (part 1, part 2) I described how we picked a reference edge server that we recommend for the AI Training of the CFD Suite. From the perspective of this article the main conclusions are:
- if you are not yet ready or do not have needs to migrate to more complex HPC infrastructures, Lenovo ThinkEdge SE450 Edge Server (Product Guide, Press Release) powered by 2 NVIDIA A100 80GB Tensor Core GPUs (Learn More) is byteLAKE’s recommended hardware configuration to perform CFD Suite’s AI Training at the Edge;
- CFD Suite’s AI Training’s performance improves if we add more NVIDIA GPUs per node. The speedup between 1xA100 and 2xA100 was stable for all benchmarked meshes and varied from 1.76 to 1.87.
Here, let me quickly summarize how to get access to the complete CFD Suite package, including software license (CFD Suite) and reference hardware.
CFD Suite — licensing model
I have already described our model in details in one of my previous posts (link). It basically consists of 2 parts:
- one time cost related to CFD Suite’s adaptation, integration and AI Training so that it can make accurate predictions for selected traditional CFD solvers. This fee includes related consultancy and is always estimated case by case.
- yearly license fee that gives access to the CFD Suite software, updates and customer support.
There is also an option to get trained on AI Training. We made it possible so that some of our clients and partners have the possibility to (re-)train the CFD Suite AI models themselves. By default, byteLAKE pre-trains the models for the initially identified CFD solvers as part of the one time cost.
Hardware
CFD Suite has been optimized for various hardware configurations to ensure optimal results for both: AI Training and inference (predictions). Lenovo’s SE450 powered with 2 NVIDIA A100 80GB Tensor Core GPUs is a recommended edge platform that makes it possible to leverage the full potential of byteLAKE’s CFD Suite. It enables both: AI Training and inference (predictions).
To learn more about the product, make sure that you follow byteLAKE on social media and subscribe to a dedicated blog post series available at: www.byteLAKE.com/en/AI4CFD-toc.
Appendix
- download the full report: https://www.bytelake.com/en/download/4018/
- read the report on SlideShare:
