AI builder maxes out rig with 80GB VRAM after NVIDIA discount rejection
After a 3-month rejection from NVIDIA Inception, one builder lands a rare RTX Pro 5000.
A dedicated AI enthusiast on Reddit shared the completion of their ultimate local AI workstation, detailing a months-long struggle with supply shortages and pricing. The builder originally aimed for an NVIDIA RTX Pro 6000, applying to the NVIDIA Inception program for a discount that would have brought the ~$8.5k card down by $1k. After a three-month wait, the application was rejected—and by then, the card's price had jumped to $13.5k, far beyond budget.
Instead, the user sourced the last available RTX Pro 5000 in their country, combining it with existing hardware to achieve a total of 80GB VRAM (across multiple GPUs, including a 5090), 192GB system RAM, 17TB of disk space, and an AMD 9950X3D CPU. The system runs on a 1300W ATX 3.1 PSU, which the builder calculates will hit ~95% load under simultaneous full GPU/CPU stress—a risky margin, mitigated by power-limiting the 5090. The rig is intended for heavy AI workloads like running Q_8 quantized models and multi-GPU ComfyUI pipelines. The post celebrates the persistence required and encourages others building AI rigs to stay the course.
- Build includes 80GB VRAM (RTX Pro 5000 + another GPU) and a 9950X3D CPU with 192GB RAM and 17TB storage.
- Original plan for RTX Pro 6000 fell through after NVIDIA Inception rejection and price surge from $8.5k to $13.5k.
- Power supply at 1300W runs near 95% capacity under full load, with plans to power-limit the 5090 for safety.
Why It Matters
Shows the extreme lengths AI professionals go to for local compute amid GPU shortages and pricing volatility.