Build 284 – Leibniz: Machine Learning Workstation

Leibniz is a machine learning workstation, designed around twin 96-core EPYC 9654 processors and an RTX A6000 48GB video card.

Interested In Buying This PC Build?

Buy Now

Navigate here

2024 Build Archive

Support

  • Please note that we are unable to provide extended user-specific software support, you will need to be able to troubleshoot.
  • If you require IT support for your PC and we are unable to help, please contact NerdsToGo Cary at (919)-999-2181 or salescarync@nerdstogo.com
  • We do not work on PCs we didn't build. For computer repair & assembly, please contact Black Sun PC at (980) 866-5497 or jeremy@blacksunpc.com
Machine Learning Workstation

Description

Leibniz is a machine learning workstation, designed around twin 96-core EPYC 9654 processors and an RTX A6000 48GB video card.

General System Design

This build is designed around twin 96-core EPYC 9654 processors and an RTX A6000 48GB video card.

Software Workload/Customer Need

This customer needed a powerful new workstation to handle their growing professional workload that would last a long time.

This build is designed for machine learning, LLM development, and highly multithreaded CPU rendering.

This build was influenced by our Dual EPYC 9004 Extreme Workstation.

System Spec

Processor (CPU): 2x AMD EPY​C 9654

CPU Cooler: 2x Silver​Ston​e Tec​hnolo​gy X​E360-S​P5

Motherboard: Gigabyte M​Z73-LM0 (r​ev. 2.0)

Memory: Kingston 2​4x64GB DDR​5-5600 ECC

Main Storage (NVME): Kingston Fury Renegade 4TB

Backup Storage (HDD): 4x Toshiba MG09ACA18TE 18TB RAID5

Video Card (GPU): Nvidia RTX A6000 48GB

Power Supply (PSU): be quiet! Dark Power Pro 13 1600W Titanium

Case: Fractal Design Meshify 2 XL

11x Phanteks PH-F120T30, 4x Noctua NF-A4x20, 1x Silverstone FHS80X

OS: Windows 11 Pro for Workstations

Facebook Build Gallery | YouTube Video Archive

Design Rationale

  • The dual EPYC 9654s gave the customer a great amount of single-core performance for general use, and multicore performance for heavily-threaded tasks. We reviewed the 9754, but determined that the base/boost frequencies were lower than desired, and there was much lower cache/core than preferred for AI development.
  • To ensure the CPU runs at full speed, liquid cooling with dual 360mm coolers is recommended.
  • The customer wanted a stable motherboard.
  • The customer wanted at least 1.5TB of RAM, adding more RAM was relatively easy to do.
  • The customer requested a specific storage setup.
  • The RTX A6000 48GB gave the customer enough VRAM for their models, and could be swapped for an RTX 6000 ADA in the future.
  • The overall design of the build meant that an 1600W power supply was recommended.
  • We used a large full-tower case to fit the build.
  • We added extra fans and custom acrylic bracing to cool the RAM and VRM more effectively.
  • We used our highest-end fans to ensure proper airflow and static pressure.

Customer Review

Build Gallery

Related Builds

Build 307 – ThreadLightly: Machine Learning Workstation

ThreadLightly is a machine learning workstation, designed around a 24-core Threadripp​er ​7960X processor and an RTX 5090 32GB video card.

Build 284 – Leibniz: Machine Learning Workstation

Leibniz is a machine learning workstation, designed around twin 96-core EPYC 9654 processors and an RTX A6000 48GB video card.

Build 273 – ZebraSuper: AI Inference Workstation

ZebraSuper is an AI inference workstation, designed around a 16-core i7-13700K processor and an RTX 4090 24GB video card.

Build 214 – The Enterprise: AI Development Workstation

The Enterprise is an AI development workstation, designed around a 24-core i9-13900K processor and an RTX 4080 16GB video card.

Build 206 – Hammerhead: Machine Learning Workstation

Hammerhead is a machine learning workstation, designed around a 24-core i9-13900K processor and an RTX 4090 24GB video card.

Build 37 – Deep Shadow: Machine Learning Workstation

Deep Shadow is a machine learning workstation, designed around an 8-core i9-9900K processor and twin RTX 2080Ti 11GB video cards.

Interested In Buying This PC Build? Contact Us!

See our brief update on the status of components, including pricing and lead time

Search

We are unable to provide free or aged systems at this time. No soliciting.