About 10 months ago I finished my dual 5090 build, but I found that the VRAM was insufficient for my AI research workloads. To eliminate the VRAM bottleneck, I recently upgraded my setup to dual NVIDIA Pro 6000 Workstation cards.
While my main work is training image generation models, the massive 192GB VRAM buffer has the added bonus of allowing me to run large LLMs like Minimax M2.1 and GLM 4.7 fully locally. This has been very helpful when processing private data.
I also switched to Linux as my daily driver a few months ago. It has been the best decision I have made regarding system stability and resource management. I currently dual boot Windows on a separate drive specifically for gaming.
The total cost for the PC came out to just over €20k.
Full Spec List:
- CPU: AMD Ryzen 9950X3D
- GPUs: Dual NVIDIA Pro 6000 Workstation Edition
- Motherboard: Asrock X870E Taichi
- RAM: 96GB G.Skill Ripjaws M5 Neo RGB DDR5-6000, CL30
- Storage (Linux): WD BLACK SN850X 8TB
- Storage (Windows): Samsung 990 Pro 4TB
- Shared Storage for both OS: Samsung 990 Pro 4TB
- PSU: Seasonic Prime TX-1600 Noctua Edition
- AIO: Lian Li Hydroshift 360TL
- Case: Lian Li o11 EVO XL (Front mesh panel + upright GPU bracket)
- Fans: 13x Lian Li TL series fans
- GPU 1 Mounting: Lian Li VG4-4-V2W Vertical Kit (w/ LINKUP PCIE 5.0 22cm riser)
- GPU 2 Mounting: Upright w/ LINKUP PCIE 5.0 90cm riser cable
Happy to hear your thoughts and answer any questions about the build or the specific hardware choices.