Skip to main content

Framework 16 to MXM GPU Adapter

The Framework 16 to MXM GPU Adapter is an open-source community project by Joe Shapiro (GitHub: JJ-Shapiro) that aims to create a custom PCB adapter allowing standard MXM mobile GPU modules to be used in the Framework Laptop 16's expansion bay. If successful, it would enable users to install a wide range of mobile GPUs — including Nvidia MXM modules such as the mobile RTX 4090 — into their Framework 161.

The project has progressed through multiple PCB revisions, with the V0.5 design finalized and sent off for manufacturing in March 202623. Earlier revisions (V0.2, August 2025; V0.3, October 2025) iteratively refined the EEPROM circuitry, differential pair routing, and MXM connector placement.

Prototype / In Development

This project is in the prototype phase. V0.5 PCBs were sent for manufacturing in March 2026, but no working hardware has been publicly demonstrated yet.

Design Overview​

MXM (Mobile PCI Express Module) is a standardised form factor for mobile GPUs that packages the GPU die and VRAM onto a small module with a PCIe connector, similar in concept to a desktop PCIe slot but in a compact laptop-friendly form factor14. The adapter board connects the Framework 16's expansion bay interposer to an MXM card edge connector, routing the eight PCIe lanes provided by the Framework 16 to the MXM module1.

V0.2 Improvements​

The V0.2 revision, released in August 2025, introduced several improvements over the initial design2:

  • Integrated EEPROM controller based on Framework's reference dual M.2 expansion bay design, required for the Framework EC to identify the expansion module
  • ID lines pulled to ground with 330k resistors
  • Fan headers wired to the interconnect
  • 0.1 µF capacitors on eDP lines
  • PCIe REFCLK+ and REFCLK- reference clock lines
  • Panel and backlight controls connected to the interconnect
  • Corrected SCL/SDA lines for EEPROM communication

V0.3–V0.5 Progress (October 2025 – March 2026)​

In October 2025, Joe Shapiro began working on writing the EEPROM firmware required for the Framework EC to recognize the MXM module as a GPU. After struggling with the Framework framework_tool command-line utility, he successfully wrote the EEPROM using a CH341A programmer, duplicating the configuration from a Framework 7700S GPU module5.

The V0.5 revision, committed to GitHub on March 7, 2026 and finalized for manufacturing on March 22, 2026, features fully routed differential pairs, production-ready Gerber files, and a Bill of Materials (BOM)3. The design uses a 6-layer PCB (increased from the earlier 4-layer prototypes) to improve PCIe signal integrity at MXM speeds.

Key Challenges​

Mechanical Constraints​

Framework VP of Engineering nrp noted that the main limitation is mechanical — recent MXM modules are larger than the original MXM specification and require larger thermal solutions6. Finding a low-profile MXM connector that keeps the card at a height compatible with a cooling solution within the expansion bay shell has been a significant design challenge2.

Power Management​

The Framework 16 expansion bay interposer can deliver up to approximately 200 W (at 20 V). Power management for the GPU — ensuring it does not exceed the interposer's power delivery limits — remains an open question7.

Cooling​

A universal cooling solution for different MXM modules is challenging, as GPU die placement varies between modules. Possible approaches include per-module cooling solutions or an inverted board layout where the MXM connector and interposer pads are on opposite sides of the PCB78.

Embedded Controller Integration​

The expansion bay module requires an EEPROM chip that the Framework EC reads to determine the board type and configure the system accordingly. The project uses Framework's open-source ExpansionBayConfigurationGenerator to generate EEPROM contents. Additional EC firmware changes may be needed to fully support a new expansion module type9.

Technical Specifications​

SpecificationDetails
Target PlatformFramework Laptop 16
PCIe Lanes8 (max provided by FW16 expansion bay)
GPU Form FactorMXM (Type A and Type B)
Power DeliveryUp to ~200 W via interposer
EEPROMIntegrated, based on FW reference design
PCB Layers6-layer (V0.5); earlier revisions used 4-layer
GitHub Stars83
SourceGitHub — JJ-Shapiro/Framework-GPU-to-MXM

Footnotes​

  1. Framework 16 to MXM Gpu - V0.1 Prototype design — Framework Community (Joe_Shapiro, August 2025) ↩ ↩2 ↩3

  2. V0.2 update in the same thread — Framework Community (Joe_Shapiro, August 2025) ↩ ↩2 ↩3

  3. committing final version sent off for manufacturing — GitHub (JJ-Shapiro, March 22, 2026) ↩ ↩2 ↩3

  4. MXM: Powerful, Misused, Hackable — Hackaday — Hackaday (April 2024) ↩

  5. Framework 16 to MXM Gpu - V0.1 Prototype design — Page 5 — Framework Community (Joe_Shapiro, October 2025) ↩

  6. nrp comment on mechanical limitations — Framework Community (nrp, August 2025) ↩

  7. Gmanny comment on power and cooling — Framework Community (Gmanny, August 2025) ↩ ↩2

  8. Gmanny inverted board suggestion — Framework Community (Gmanny, August 2025) ↩

  9. Gmanny comment on EEPROM and EC — Framework Community (Gmanny, August 2025) ↩