The DC-ROMA RISC-V Mainboard II (also called the DC-ROMA RISC-V AI PC) is the second-generation RISC-V mainboard for the Framework Laptop 13 by DeepComputing. Powered by an ESWIN EIC7702X SoC with a dual-die 8-core SiFive P550 CPU and a 40 TOPS NPU, it is the world's first RISC-V AI PC capable of running large language models locally.
The DC-ROMA RISC-V Mainboard III is the third-generation RISC-V mainboard for the Framework Laptop 13 by DeepComputing. Powered by the SpacemiT K3 SoC — the world's first RISC-V SoC supporting the RVA23 profile — it features an 8-core CPU up to 2.5 GHz, a 30 TOPS NPU, and up to 60 TOPS of combined general-purpose AI compute. It officially launched on May 12, 2026, with pre-orders opening at $699.
The Framework Desktop is a Mini-ITX workstation announced at the 2nd Gen event in February 2025. It features an AMD Strix Halo APU and is designed to be modular, containing two front-side expansion card slots as seen in Framework's laptop models. The front panel has slots for 21 interchangeable tiles for user customization. Pre-orders opened February 25, 2025 with shipping planned for early Q3 2025. After missing that target by ~6 weeks, Batch 1 units shipped the week of August 11, 2025 alongside the review embargo lift. Memory is soldered LPDDR5x (non-upgradeable) — Framework worked with AMD for months but concluded it was not technically feasible to land modular memory at high throughput with the 256-bit memory bus.
framework-rocm is a set of reproducible ROCm containers for the Framework Desktop (AMD Ryzen AI MAX / Strix Halo / Radeon 8060S, gfx1151). Developed by community member geoff-davis, it provides pinned PyTorch and JAX Docker images with GPU smoke tests and measured attention/training performance findings.
The MetaComputing AI PC Mainboard is the first Arm-based third-party mainboard for the Framework Laptop 13. Developed by Switzerland-based MetaComputing in collaboration with CIX Technology and announced in December 2025, it is powered by the CIX CP8180 (CIX P1) SoC with a 12-core Arm CPU and a 45 TOPS NPU.
The Plugable TBT5-AI is a Thunderbolt 5 eGPU enclosure marketed for local AI inference, pairing an 850W ATX 3.1 power supply (80+ Gold, 600W dedicated to the GPU) with a PCIe 4.0 x4 slot and a built-in Thunderbolt 5 dock. It provides 96W host charging, a downstream Thunderbolt 5 port, 10Gbps USB-A/C ports, and 2.5GbE Ethernet. It is TAA compliant and supports Thunderbolt 5, Thunderbolt 4, and USB4 (eGPU-capable) Windows 11 hosts with Intel 11th-gen+ or AMD Zen 4+ CPUs.
Strix Halo Guide is a comprehensive, practical guide to running large language models locally on AMD Strix Halo / Ryzen AI MAX+ 395 systems with Radeon 8060S (gfx1151) and 96GB/128GB unified memory. Maintained by community member hogeheer499-commits, it covers BIOS configuration, Ubuntu 24.04/kernel setup, Ollama, llama.cpp Vulkan/RADV, ROCm/HIP experiments, vLLM, benchmarks, raw logs, and reproducibility checks.
strix-llm is a known-good local LLM inference configuration tool for the Framework Desktop (Ryzen AI MAX+ 395 / Strix Halo). Maintained by community member Alberto Migliorato (GitHub: AlbeMiglio), it provides a tested, reproducible setup for running large language models locally on Framework Desktop hardware using ROCm and llama.cpp.
The XPG Mars 970 Plus (ADATA part AMTG970P) is a PCIe Gen5 x4 NVMe 2.0 SSD in the M.2 2280 form factor, built on the Silicon Motion SM2508 controller (TSMC 6nm) and 232-layer 3D TLC NAND. It delivers sequential reads up to 11,000 MB/s and writes up to 10,000 MB/s in a single-sided, laptop-friendly design. It is now offered directly by Framework on the Framework Laptop 13 Pro DIY Edition, where Framework qualified it for better performance, efficiency, and long-term reliability.