p
p
p
Thelio Mira: An AI Beast from System76
System76 Mira Desktop Built for AI!
Powerful and Beauty Fitting Into Your Lifesstyle
Thelio Desktop Line From System76 : AMD Ryzen + ThreadRipper, Intel
Core Features
- System76 Thelio Mira Premium: Ryzen 9 9950X (16C/32T @ 5.7 GHz), Linux, Pop!_OS 24.04 COSMIC
- System76 Thelio Mira Elite: 5.7 GHz Ryzen 9 9950X3D (16C/32T @ 5.7 GHz), Linux, Pop!_OS 24.04 COSMIC
- GPU Options: AMD Radeon RX 7900 XTX (24 GB GDDR6) or Radeon AI Pro R9700 (32 GB GDDR6) – both ROCm 7.2 optimized
- Memory: Up to 48/128 GB DDR5
- Storage: Encrypted 2 TB PCIe 5.0 SSD + 16 TB Additional SATA SSDs
- Pre-Installed: PyTorch latest, ROCm 7.2, Ollama / LM Studio, ComfyUI w/ ControlNet, TTS, LTX-2 model loaded
- Custom: TTS + Image-to-Video (I2V) workflows included
- Scale: Single workstation or 3-node cluster for massive LLMs like DeepSeek V4 (February 2026)
ROCm Software
ComfyUI Stable Diffusion
PyTorch Libraries
Deepseek IV FP16 / FP8 Models (Feb 2026)
Lightricks, LTX-2
Ollama Inference
LM Studio Inference
F5 Text-to-Speech Model
Thelio Mira can be configured as a standalone system. Or,can be purchased as a powerful three-node cluster. Your choice.
Using AMD ROCM software 7.2 and PyTorch libaries to make this a multi-node and multi-GPU platform ready to crunch your local AI workloads
AMD GPUs Compared: RX 7900 XTX vs. Radeon AI Pro R9700 (XFX Models)
XFX Radeon RX 7900 XTX
Flagship RDNA 3 consumer/gaming GPU, excellent for high-VRAM local AI inference
XFX Radeon AI Pro R9700
RDNA 4 pro/AI workstation card, optimized for large models, better efficiency & AI accelerators
| Feature | XFX Radeon RX 7900 XTX | XFX Radeon AI Pro R9700 |
|---|---|---|
| Architecture | RDNA 3 | RDNA 4 (newer, AI-optimized) |
| VRAM | 24 GB GDDR6 | 32 GB GDDR6 |
| Memory Bus / Bandwidth | 384-bit / up to ~960 GB/s | 256-bit / up to ~640 GB/s |
| Boost Clock | Up to ~2.5–2.6 GHz | Up to ~2.92 GHz |
| Power Draw (TDP) | 355 W | ~300 W |
| AI / Compute Focus | Strong ROCm support, great for inference | Enhanced (128 AI accelerators, FP8/INT8, up to 2× AI perf vs prior gen) |
| Best For | Heavy compute, large models on budget, 3-node cluster | Larger models, longer contexts, higher precision, future-proof AI workloads |
| Mira System Configuration | Premium Package | Elite Package |
Thelio Mira Solution Pricing Overview
| Model | CPU | Key Specs | Current Price (USD) | Notes |
|---|---|---|---|---|
| Thelio Mira Premium | Ryzen 9 9950X (16C/32T @ up to 5.7 GHz) | 48 GB DDR5 (5600 MHz) 2 TB PCIe 5.0 SSD Pop!_OS 24.04 COSMIC | $5888.00 | Balanced high-performance config; excellent for ROCm AI workloads, clusters, and value. Includes the venerable XFX Radeon RX 7900 XTX GPU |
| Thelio Mira Premium – 3 Node Cluster | 3 × Ryzen 9 9950X (16C/32T @ up to 5.7 GHz) | 3 × (48 GB DDR5 + 2 TB PCIe 5.0 SSD) Pop!_OS 24.04 COSMIC cluster configuration | $17595.00 | Three-node cluster for distributed AI training/inference; ideal for scaling ROCm-based workloads. Includes the venerable XFX Radeon RX 7900 XTX GPU |
| Thelio Mira Elite | Ryzen 9 9950X3D (16C/32T @ up to 5.7 GHz) | 128 GB DDR5 (3600 MHz or higher) 4 TB PCIe SSD Pop!_OS 24.04 COSMIC | $7880.00 | Premium tier with 3D V-Cache for superior gaming/AI inference; extra headroom for massive models. Includes the new XFX Radeon AI Pro R9700 GPU |
| Thelio Mira Elite – 3 Node Cluster | 3 × Ryzen 9 9950X3D (16C/32T @ up to 5.7 GHz) | 3 × (128 GB DDR5 + 4 TB PCIe SSD) Pop!_OS 24.04 COSMIC cluster configuration | $23350.00 | High-end three-node cluster with 3D V-Cache; optimized for demanding inference and large-model workloads. Includes the new XFX Radeon AI Pro R9700 GPU |
Prices are approximate and based on System76’s current preconfigured or the configured options desired at the time of sale ((Please reach out to us for exact builds, discounts, and availability).
Veteran-owned provider of open-source LINUX solutions, AI platforms, and high-performance hardware. Visit madison-digital.net for more.
Both configurations excel in ROCm-based local AI (PyTorch, Ollama, ComfyUI). The RX 7900 XTX offers great value for 24 GB VRAM clusters; the AI Pro R9700 provides extra headroom for massive models like DeepSeek V4.For the three-node cluster we ensure that the cluster is working and provide direction to configure in your environment. Software is all preloaded.