Xeon AI Training Workstation Overview
This Xeon AI Training Workstation is built around the Xeon W7-3565X with 4 RTX Pro 6000 Blackwell 96GB video cards.
This combination enables smooth development in Intel OpenVINO, Intel OneAPI, and Intel openMKL. This combination provides a very powerful solution for machine learning training, fine-tuning, and inference with AI large language models. This machine has a total of 384GB VRAM and runs on 120V power (single 20A or two 15A circuits).
Xeon AI Training Workstation Upgrades/Changes
This Machine Learning AI Workstation can be changed in CPU model, RAM loaded, and any unique storage requirements. This system has dual power supplies to allow 4 for 300W GPUs and a high-wattage CPU. If you need more power, we can upgrade the power supplies up to dual 1600W. This will need specialized local power at your office (separate 15A circuits or a single 20A).
Modern Xeon W can be heavily overclocked, often being able to set the all-core turbo to 4.0Ghz or higher.
768GB of RAM is the highest we can go on Xeon, without resorting to EXTREMELY expensive 128GB or 256GB DIMMs.
Xeon AI Training Workstation Limitations
This deep learning workstation is currently limited to 120 threads, 7 PCI-E devices, and 2TB of memory. Please let us know if you need a more robust platform.
Builds In This Set
Single-Socket AI Inference Desktop Workstations (24B – 192B)
Ryzen 9 24B | Ryzen 9 32B | Ryzen 9 48B | Ryzen 9 96B
Core Ultra 9 24B | Core Ultra 9 32B | Core Ultra 9 48B | Core Ultra 9 96B
Threadripper 96B | Threadripper 192B | Threadripper Pro 96B | Threadripper Pro 192B
Xeon 2500 96B | Xeon 2500 192B | Xeon 3500 96B | Xeon 3500 192B
Single-Socket AI Training Desktop Workstations (384B)
Threadripper 384B | Threadripper Pro 384B | Xeon 3500 384B
Heavy Compute Desktop Workstations
Dual EPYC 9005 Low | Dual EPYC 9005 Medium | Dual EPYC 9005 Extreme | Dual EPYC 9005 Ultimate
