Linux multi-GPU training and inference
System76 Thelio Major (2x RTX 6000 Ada)
System76 · Verified 2026-07-23
From $18,000
Dual RTX 6000 Ada with Linux-first support. 96GB combined for 70B-class models over PCIe P2P.
Main limitation: No Windows; mail-in warranty; higher cost than DIY
Some links are affiliate links. We may earn a commission at no extra cost to you. Full disclosure.
Key Specifications
| Specification | Value |
|---|---|
| Form Factor | desktop tower |
| GPU | NVIDIA RTX 6000 Ada Generation |
| GPU Count | 2 |
| GPU Memory per GPU | 48 |
| Total Accelerator Memory | 96 |
| Memory Architecture | dedicated VRAM (GDDR6) |
| GPU Memory Type | GDDR6 |
| Memory Bandwidth (per GPU) | 960 |
| System RAM | 256 |
| RAM Type | DDR5 ECC |
| ECC Memory | Yes |
| Max System RAM | 768 |
| CPU | AMD Ryzen Threadripper 7980X (64-core) |
| CPU Cores | 64 |
| CPU TDP | 350 |
| Primary Storage | 4096 |
| Storage Type | NVMe SSD (PCIe 5.0) |
| PCIe Generation | PCIe 5.0 |
| PCIe x16 Slots | 3+ |
| NVLink Support | No (RTX 6000 Ada has no NVLink — NVIDIA datasheet 2504660) |
| Power Supply | 1500 |
| Est. Peak Power | 1300 |
| Cooling | professional air cooling |
| Dimensions | Mid-tower workstation |
| OS | Pop!_OS / Ubuntu (pre-installed) |
| Linux Support | Yes (first-class) |
| Windows Support | No (not sold with Windows) |
| WSL Support | No |
| CUDA Support | Yes |
| Metal Support | No |
| Warranty | 1 (extendable) |
| Onsite Support | No (mail-in) |
| Release Year | 2025 |
What Models Can It Run?
Calculated estimates using weight-only memory analysis at Q4 (4-bit) quantization. Actual requirements vary by framework, context length, and runtime overhead. See methodology.
| Model | Params | Required Memory (Q4) | Fits in 96GB? |
|---|---|---|---|
| Llama 3.1 8B | 8B | 5.52 GB | ✓ Yes |
| Llama 3.1 70B | 70B | 37.14 GB | ✓ Yes |
| Llama 3.2 1B | 1B | 2 GB | ✓ Yes |
| Llama 3.2 3B | 3B | 3.01 GB | ✓ Yes |
| Qwen 2.5 7B | 7B | 5.01 GB | ✓ Yes |
| Qwen 2.5 14B | 14B | 8.53 GB | ✓ Yes |
| Qwen 2.5 32B | 32B | 18.06 GB | ✓ Yes |
| Qwen 2.5 72B | 72B | 38.14 GB | ✓ Yes |
| Mistral 7B | 7B | 5.01 GB | ✓ Yes |
| Mixtral 8x7B | 46.7B | 25.44 GB | ✓ Yes |
| DeepSeek R1 7B | 7B | 5.01 GB | ✓ Yes |
| DeepSeek R1 32B | 32B | 18.06 GB | ✓ Yes |
| DeepSeek R1 70B | 70B | 37.14 GB | ✓ Yes |
| Phi-3 Medium 14B | 14B | 8.53 GB | ✓ Yes |
| Gemma 2 9B | 9B | 6.02 GB | ✓ Yes |
| Gemma 2 27B | 27B | 15.05 GB | ✓ Yes |
| Stable Diffusion XL | 6.6B | 4.81 GB | ✓ Yes |
| Flux.1 Dev | 12B | 7.52 GB | ✓ Yes |
| Flux.1 Schnell | 12B | 7.52 GB | ✓ Yes |
Max model size: 70B class models (~185B params at Q4). These are calculated estimates, not measured results.
Pros & Cons
Pros
- ✓ Very large accelerator memory (96GB)
- ✓ Linux support
- ✓ ECC memory for reliability
Cons
- ✗ No NVLink support (multi-GPU limited to PCIe P2P)
- ✗ No Windows; mail-in warranty; higher cost than DIY
Compare with Other Workstations
Apple Mac Studio M2 Ultra (192GB)
Apple Mac Studio M2 Ultra (64GB)
ArsenalPC MES2X Dual RTX 5090 AI Workstation
BIZON G3000 G2 (1x RTX 5090)
BIZON G3000 G2 (4x RTX 5090)
BOXX APEXX 8R (1x RTX PRO 6000 Blackwell)
Custom Dual RTX 4090 Training Workstation
Custom RTX 5090 AI Workstation (Value Build)
Sources & Verification
Primary source: System76 Thelio Major Product Page
https://system76.com/desktops/thelio-major
Verification date: 2026-07-23
Data quality: Vendor specification — from manufacturer datasheet. No hands-on testing was performed.