LOGO

Framework Desktop PC: Gaming & Local AI Inference

February 26, 2025
Framework Desktop PC: Gaming & Local AI Inference

Framework Launches its First Desktop Computer

Framework, renowned for its commitment to creating modular and easily repairable laptops, has now introduced its inaugural desktop computer. This compact PC delivers performance that surpasses its diminutive size.

Innovative Internal Components

A key feature of this new device lies in its internal hardware. Framework is among the first manufacturers to integrate AMD’s recently unveiled Strix Halo architecture, also recognized as the Ryzen AI Max processors. This all-in-one processing unit is engineered to provide substantial computational power.

Essentially, Framework has developed a PC specifically tailored for two distinct user groups: individuals seeking a remarkably small form factor gaming PC, and those desiring the capability to execute large language models directly on their personal machines.

Design and Customization

The Framework Desktop’s exterior presents a more playful aesthetic than that of a conventional computer. It boasts a compact 4.5L volume, constructed around a mini-ITX mainboard, making it smaller than both the PlayStation 5 and the Xbox Series X.

The front panel is highly customizable, featuring 21 interchangeable plastic square tiles. Customers purchasing directly from the Framework website can select from a variety of tile colors and patterns to personalize their desktop’s appearance.

Beyond the standard ports typically found on a mini-ITX mainboard, the device incorporates Framework’s signature expansion card system – two slots are located on the front, and two on the rear. A diverse selection of modules is available, including USB-C and USB-A ports, a headphone jack, an SD card reader, and storage expansion options.

Internal Specifications

The internal architecture is straightforward: the mainboard houses AMD’s accelerated processing unit, alongside a cooling fan, heat sink, power supply, and two M.2 2280 NVMe SSD slots for storage.

AMD’s Strix Halo APU is directly soldered to the mainboard. Framework offers two distinct configurations: the AMD Ryzen AI Max 385 and the AMD Ryzen AI Max+ 395. The higher-end model features 16 CPU cores, 40 graphics cores, and 80MB of cache, while the base configuration provides 8 CPU cores, 32 graphics cores, and 40MB of cache.

Memory Considerations

A notable design decision concerns the RAM. Framework provides 32GB to 128GB of RAM, which is soldered directly onto the mainboard. This means that users will not have the option to upgrade or add more RAM in the future.

According to Framework CEO Nirav Patel, this choice was made to maximize memory bandwidth. “To enable the massive 256GB/s memory bandwidth that Ryzen AI Max delivers, the LPDDR5x is soldered,” he explained in a company blog post.

He further stated that extensive efforts were made to explore alternative solutions, but ultimately, a modular memory configuration capable of achieving the required throughput with the 256-bit memory bus proved technically unfeasible.

AI and Gaming Capabilities

Despite this limitation, the availability of up to 128GB of unified memory opens up significant possibilities for running large language models. Llama 3.3 70B can operate smoothly using tools like Ollama, llama.cpp, and other open-source platforms designed for local AI processing.

Other open-weight models from developers such as Mistral, Nous, Hermes, and DeepSeek are also expected to function effectively. Framework also offers the mainboard as a standalone component. The company has even constructed a mini-rack utilizing four Framework Desktop mainboards in parallel for AI testing purposes.

Pricing and Availability

The entry-level Framework Desktop is priced at $1,099, while the top-of-the-line version costs $1,999. Consistent with other Framework products, the company guarantees compatibility with both Windows and popular Linux distributions, including Ubuntu, Fedora, and the gaming-oriented Bazzite.

Preorders are currently being accepted, with shipments scheduled to commence in early Q3 2025.

#Framework#desktop PC#gaming PC#AI inference#local AI#customizable PC