Custom RTX 5090 AI Workstation (Value Build) vs System76 Thelio Major (1x RTX PRO 6000 Blackwell) for Local AI: Local AI Performance

Side-by-side comparison of 2 AI workstation configurations. All specs from manufacturer datasheets.

Specification Custom RTX 5090 AI Workstation (Value Build) System76 Thelio Major (1x RTX PRO 6000 Blackwell)
Form Factor desktop tower desktop tower
GPU Model NVIDIA GeForce RTX 5090 NVIDIA RTX PRO 6000 Blackwell Workstation Edition
GPU Count 1 1
Gpu vram each gb 32 96
Total GPU VRAM (GB) 32 96
Memory Architecture dedicated VRAM (GDDR7) dedicated VRAM (GDDR7)
Gpu memory bandwidth each gbps 1792 1792
System RAM (GB) 64 128
System Memory Type DDR5-5600 DDR5 ECC
ECC Memory Yes
CPU Model AMD Ryzen 9 9950X AMD Ryzen Threadripper 9980X (64-core)
CPU Vendor AMD AMD
Cpu cores 16 64
Storage primary gb 2048 4096
Storage Type NVMe SSD (PCIe 5.0) NVMe SSD (PCIe 5.0)
NVLink Support No (RTX 5090 does not support NVLink) No (Blackwell workstation GPUs do not support NVLink)
Power supply watts 1200 1500
Estimated peak power watts 850 900
Cooling air (custom AIO optional) professional air cooling
Dimensions Mid-tower ATX Mid-tower workstation
Operating system Linux (Ubuntu 24.04) / Windows 11 Pop!_OS / Ubuntu (pre-installed)
CUDA Support Yes Yes
Metal Support No No
Linux Support Yes Yes (first-class)
Windows Support Yes No (not sold with Windows)
Wsl support Yes (WSL2 with CUDA) No
Warranty (Years) Varies by component (typically 3yr GPU) 1 (extendable)
Onsite Support No No (mail-in)
Release Year 2025 2026

Category Winners

Most Accelerator Memory

System76 Thelio Major (1x RTX PRO 6000 Blackwell)

Lowest Price

Custom RTX 5090 AI Workstation (Value Build)

Lowest Power

Custom RTX 5090 AI Workstation (Value Build)

Best Value (VRAM/$)

Custom RTX 5090 AI Workstation (Value Build)

Model-Fit Comparison (Q4 Quantization)

Calculated estimates using the site's central ModelFitService.

Model Custom RTX 5090 AI Workstation (Value Build)System76 Thelio Major (1x RTX PRO 6000 Blackwell)
Llama 3.1 8B
Llama 3.1 70B
Llama 3.2 1B
Llama 3.2 3B
Qwen 2.5 7B
Qwen 2.5 14B
Qwen 2.5 32B
Qwen 2.5 72B
Mistral 7B
Mixtral 8x7B
DeepSeek R1 7B
DeepSeek R1 32B
DeepSeek R1 70B
Phi-3 Medium 14B
Gemma 2 9B
Gemma 2 27B
Stable Diffusion XL
Flux.1 Dev
Flux.1 Schnell

Which Should You Choose?

Custom RTX 5090 AI Workstation (Value Build)

Custom Build
$4,000

Best value CUDA workstation. RTX 5090 with 32GB VRAM handles 7B-14B models at full speed, 33B with quantization.

Best for: Budget CUDA development and inference

Limitation: 32GB VRAM limits model size; DIY means no unified warranty

System76 Thelio Major (1x RTX PRO 6000 Blackwell)

System76
$15,000

Linux-first 96GB VRAM workstation. Best for CUDA developers who want maximum single-GPU memory without enterprise overhead.

Best for: Linux-based CUDA development with large VRAM needs

Limitation: No Windows option; mail-in warranty only; premium price

Some links are affiliate links. See our disclosure. Specifications from manufacturer datasheets — no hands-on testing was performed.