In 2021, the portable panorama was bleak in a key section: the  right to repair . Manufacturers made it increasingly difficult to solve problems or update machines on one’s own. It was during this challenging landscape that a small and relatively unknown company presented an alternative. That company was  Framework .

Framework’s  modular laptops  have since emerged as an  absolute reference in repairability and expansion capacity . Over the years, they have maintained this tradition. Recently, the company announced the launch of two new portable models, alongside a surprising addition: a desktop mini-PC that, curiously, seemed contrary to the company’s principles.

 <img alt="Framework Modular Laptop" width="375" height="142" src="https://i.blogs.es/d5ce75/framework-13-review-xataka-portada-alternativa/375_142.jpeg"/>

This innovative machine is none other than the Framework Desktop. It distinguishes itself both externally and internally. The device adopts a mini-PC format with a mini-ITX motherboard and a compact chassis that occupies just  4.5 liters .

Moreover, the Framework Desktop boasts a  differentiating design  featuring a front panel where users can customize up to  21 “tiles” , which can even be printed in  3D . (The designs for customization are available online: GitHub Repository).

Framework Desktop Design

  </div>

In addition, the Desktop continues the tradition of  modular ports  found in Framework’s laptops, allowing users to customize their input and output options. Users can purchase these modules separately and integrate new ones as needed. On the inside, the technical specifications reveal a powerhouse:

  </tbody>
  </th>
             <th class="align-left">
   <p>Framework Desktop</p>
  </th>
</tr>

</thead>
  <tbody>
      <tr>
             <th class="align-left">
   <p>Processor</p>
  </th>
             <td class="align-left">
   <p>AMD Ryzen AI Max 385 (8 cores, 16 threads, up to 5.0 GHz)</p>

AMD Ryzen AI Max+ 395 (16 cores, 32 threads, up to 5.1 GHz)

GPU

Radeon 8050s Graphics (32 nuclei, 2.8 GHz)

Radeon 8060s Graphics (40 cores, 2.9 GHz)

NPU

Up to 50 tops

Memory

32/64/128 GB LPDDR5X-8000 (Welded)

Storage

Flexatx 400W

Front ports

2 x customizable I/O modules

Rear ports

1 x HDMI 2.1

2 x DisplayPort 2.1

2 x USB-C (USB4)

1 x RJ45 (5GBE)

2 x USB-A (USB 3.2 GEN1)

1 x minijack 3.5 mm

Feeding

Flexatx 400W

Dimensions and weight

96.8 x 205.5 x 226.1mm

3.1 kg

4.5 L mini-ITX box

Price

From 1,309 euros

However, in these specifications, it’s surprising that the  AMD SoC  (with the CPU, integrated GPU, and NPU) is soldered to the motherboard, as is the memory.  This seems like a betrayal  of Framework’s core values of interchangeability and repairability, and has already drawn criticism.

An AI Workstation for All Audiences

Some independent reviews have criticized this aspect of the framework. However, the decision stems from a technical necessity. As the company explained in the official announcement:

“To enable the huge  256 GB/s memory bandwidth  offered by Ryzen AI Max, the LPDDR5X was welded. We spent months working with AMD to explore solutions, but ultimately concluded that modular memory with high performance wasn’t technically viable with the 256-bit memory bus.”

F2
Framework Desktop Inside

  </div>

This is a reasonable argument, especially given that while the device has a gaming flair, its true target audience is different. The machines leverage a  unified memory architecture (UMA) , meaning there is no dedicated video memory; instead, the GPU utilizes RAM as video memory. Thus, 32, 64, or  128 GB of RAM  is shared between CPU and GPU. This concept mirrors Apple’s approach with its M1 and M2 chips, allowing flexible assignment of memory resources between CPU and GPU.

The true value of these machines lies in being  affordable workstations  capable of handling local  AI models . Recently, we discussed how OpenAI’s GPT-Oss-20b model demonstrated the importance of having significant graphic memory for such scenarios.

F4
AI Workstation

  </div>

The Framework Desktop presents a strong alternative to Apple’s computers with their unified memory architecture. For instance, if you opt for the  128 GB  RAM model (€2,359), you could allocate up to  112 GB  for video memory. This offers ample capability to work with heavy models like OpenAI’s “Large” variant (GPT-Oss-120b), or as noted by Framework, with Azure AI models.

In fact, the value proposition of this machine lies in its compactness and efficiency; it maintains a  low power consumption  of around 140W and operates quietly. This is not a rival to gaming machines with dedicated graphics. If your focus is serious gaming, this might not be your best bet.

However, if you’re more of an occasional gamer and primarily seek a  powerful workstation  that enables local AI experimentation, this may be among the  best current alternatives  for a compact, efficient, and silent setup.

Other options are evident: either invest in a high-end Mac with the highest possible amount of unified memory, or opt for a PC setup that combines multiple dedicated graphics cards (leading to higher costs and energy consumption, as well as increased noise levels).

 <img alt="Building Your Own PC" width="375" height="142" src="https://i.blogs.es/e38908/pc2023-ap/375_142.jpeg"/>

Could the Framework Desktop have been designed to be more modular and repairable? Absolutely. However, this mirrors the majority of current PCs. The manufacturer chose a different path, and this approach offers intriguing possibilities. Ultimately, it’s the users who will decide (and you will decide).



General News – 2