Lenovo RevLive 2021: Workstation Pipelines for Manufacturing
How we configured Lenovo P620 workstations to process multi-threaded simulation meshes, CAD models, and real-time GPU path-tracing simultaneously.
Modern manufacturing software demands a new breed of hardware. Standard office computers or light developer laptops crash immediately when trying to solve millions of finite element equations or open massive assembly files containing complex carbon layups.
To run simulation-first engineering workflows without waiting hours for computation, you must structure your hardware layout to match your software architecture. This post details our Lenovo RevLive 2021 presentation, focusing on how we configured our Lenovo ThinkStation P620 machines to handle high-fidelity simulation and rendering in parallel.
Table of contents
- The Performance Bottleneck in Advanced Manufacturing
- Structuring the Workstation for Multi-Application Loads
- Tuning CPU Cores, Memory, and GPU VRAM
- Summary: Optimizing the Engineering Engine
The Performance Bottleneck in Advanced Manufacturing
In our workshop, a typical design iteration involves running three intense computational tasks simultaneously:
- Mechanical Modeling: Running CAD software to design complex composite surfaces.
- Finite Element Analysis: Meshing and solving stress/flow models in Ansys.
- Real-Time Rendering: Raytracing the finished models in NVIDIA Omniverse.
If the workstation lacks CPU core capacity or memory bandwidth, these applications choke each other out. The engineer is forced to wait, interrupting their creative flow and extending project timelines.
KEY TAKEAWAY: A slow workstation is not just an inconvenience; it is a financial drain that actively limits the complexity and quality of the products you can develop.
Structuring the Workstation for Multi-Application Loads
At RevLive 2021, we explained our transition to AMD Threadripper Pro architecture on the Lenovo P620 workstation. Unlike standard CPUs, this platform offers a high core count alongside massive PCI Express bandwidth.
Figure 1: Multi-screen engineering layout powered by Lenovo ThinkStation hardware.
This architecture allows us to dedicate specific CPU cores to background Ansys simulations while keeping other cores free for smooth CAD modeling. We can run a structural load test in the background and continue modifying components in Fusion 360 without experiencing interface lag.
Tuning CPU Cores, Memory, and GPU VRAM
To optimize workstation throughput, we balanced three hardware pillars:
| Pillar | Configuration | Engineering Impact |
|---|---|---|
| CPU Compute | AMD Threadripper Pro (32 Cores) | Allows simultaneous background FEA solving and foreground CAD design. |
| Memory Bandwidth | 128GB DDR4 ECC RAM (8-Channel) | Prevents data bottlenecks when loading massive mesh structures containing millions of nodes. |
| GPU Compute | NVIDIA RTX A6000 (48GB VRAM) | Enables real-time raytraced rendering of complex assemblies without running out of video memory. |
Summary: Optimizing the Engineering Engine
Balancing CPU, RAM, and GPU capacity is essential to unlock the full performance of modern simulation and rendering software.
Key takeaways:
- Core allocation prevents lag: Dedicate separate CPU cores to solving background simulation models.
- Memory channels matter: High-fidelity meshes require large memory bandwidth to transfer data to the processor.
- Large VRAM is critical: Rendering complex composite material files in real time requires large GPU memory buffers.
Q&A
Q: Why choose AMD Threadripper Pro over standard consumer processors? A: Consumer CPUs are limited to 2 or 4 memory channels and fewer PCIe lanes. The Threadripper Pro’s 8-channel memory and 128 PCIe lanes are crucial for preventing data bottlenecks during large mesh calculations.
Q: Do you use local workstations or cloud computing? A: We use a hybrid approach. We run real-time design, visual rendering, and initial simulation validations locally on our Lenovo P620s. For massive optimization iterations, we scale out to cloud compute nodes.