Thelio Mira AI Linux Workstation: 192 GB GPU Memory
System76 has launched the Thelio Mira AI workstation, a powerhouse designed for local AI development, boasting up to 192GB of GPU memory and running Linux. Its staggering price tag, especially for high-end configurations, sparked intense debate on Hacker News regarding its value compared to Apple Silicon and the overall economics of cutting-edge AI hardware.
The Lowdown
System76 has unveiled the Thelio Mira AI, a new workstation specifically engineered for accelerating local AI development. This machine aims to empower users to train, fine-tune, and iterate on demanding AI workloads entirely on their own hardware, bypassing cloud dependence.
The Thelio Mira AI is highly configurable, with key specifications including:
- Up to a 16-core AMD Ryzen 9000 Series CPU.
- Support for up to 192 GB DDR5 RAM.
- Configurations featuring dual NVIDIA RTX Pro 6000 GPUs, offering a combined 192 GB of GPU memory.
- It targets use cases such as multi-GPU AI training, model fine-tuning, computer vision, image generation, simulations, and inference.
- The workstation runs System76's own Pop!_OS, or Ubuntu, and features open-source hardware designs. It is also notably manufactured in the USA.
This robust offering positions the Thelio Mira AI as a premium solution for professionals and researchers who require immense local computational power and memory for advanced AI applications.
The Gossip
Pricing Ponderings
Many commenters were initially shocked by the price, with some mistaking the starting price for the high-end configuration, only to find the fully specced machine costing over $40,000. This led to humorous comparisons, like choosing between a new car and an AI rig, and reflections on how quickly specialized hardware costs escalate.
Apple Allusions
A significant portion of the discussion revolved around comparing the Thelio Mira AI's performance and value to Apple Silicon, particularly the Mac Studio. While some argued that Apple's integrated memory architecture and overall price point make it a strong competitor for AI, others countered that the Thelio's raw FLOPS, higher VRAM capacity, and NVIDIA ecosystem offer superior performance for specific AI workloads, despite the higher cost.
Hardware Huddle
Commenters delved into the technical specifications and potential design choices. Concerns were raised about the use of consumer-grade CPUs (AMD Ryzen 9950X) in such a high-end system, particularly regarding potential PCIe lane limitations for multiple GPUs and NVMe drives. There was also speculation about the cooling solution for dual high-power GPUs and the specific variants of the RTX Pro 6000 cards used, with some noting visible ASUS components for the motherboard.