DC-ROMA RISC-V Mainboard III for Framework Laptop 13

Framework Laptop 13 with RISC-V Mainboard III 

DeepComputing reveals its latest RISC-V single board computer (SBC) and third-generation product for the Framework Laptop 13, the DC-ROMA RISC-V Mainboard III, into Early Access Program. Designed specifically to engage developers and hardware enthusiasts, this exciting release showcases the power of the SpacemiT K3 RISC-V AI SoC and marks the first addition to an immature production-level environment that utilizes the newly defined RVA23 instruction set profile. The DC-ROMA RISC-V Mainboard III is the first modular mainboard for laptops globally that is built with the RVA23 instruction set profile capability.

The DC-ROMA RISC-V Mainboard III provides unrivalled performance and AI functionality to open desktop-class RISC-V systems and enables users to modify their modular laptops from a fixed architecture to an open RISC-V architecture, while retaining the existing 2.2-kilowatt power supply and shell used in the current PC model recess. At its core, the SpacemIT K3 system-on-chip contains 16 cores, eight of which adhere to the RISC-V RVA23 specification with the remaining cores developed to offload specific workloads and conform to the RVA22 architecture. The DC-ROMA RISC-V Mainboard III retains full functionality of the Framework expansion card system by incorporating four USB-C connections. Two of these four connections permit 65-watt PD 3.0 charging capabilities as well as DisplayPort 1.4 video output support for driving 4K monitors.

We’ve previously covered several RISC-V SBCs, including VisionFive2 Lite, ESWIN EBC77 Series, PINE64 StarPro64, Alibaba T-Head RVB-ICE, and more. Feel free to explore those articles if you’re interested in these platforms.

DC-ROMA RISC-V Mainboard III

DC-ROMA RISC-V Mainboard III for Framework Laptop 13 Specifications:

• SoC: SpacemiT K3
• CPU and AI Subsystem:
  • 8× 64-bit RISC-V X100 high-performance cores up to 2.5 GHz (RVA23 compliant), ~130 KDMIPS
  • 8× RISC-V A100 AI accelerator cores with up to 1024-bit RVV 1.0 vector processing (RVA22 compliant), optimized for matrix workloads
  • AI compute capability up to 60 TOPS (INT4) via dedicated TCM memory and DMA acceleration
• System Memory:
  • 16 GB or 32 GB LPDDR5 (dual-package)
  • 6400 MT/s data rate (~51 GB/s bandwidth)
• Storage:
  • M.2 2280 Key-M slot with SATA or NVMe support
  • MicroSD (TF) card slot
• Display:
  • 40-pin IPEX display connector
  • Compatible with Framework 13-inch 2.2K panel
  • 100% sRGB color gamut support
• Camera: Integrated 1080p60 webcam
• Audio:
  • Dual 2 W stereo speakers
  • Dual MEMS microphones with hardware privacy mute switch
  • 3.5 mm combo audio jack (headphone/microphone)
• USB and I/O Expansion:
  • 1× USB-C (USB 3.0/2.0, DisplayPort 1.4 Alt Mode, USB-PD 3.0 up to 65 W)
  • 1× USB-C (USB 3.0/2.0, USB-PD 3.0 up to 65 W)
  • 2× USB-C (USB 3.0/2.0)
• Wireless Connectivity:
  • M.2 Key-E 2230 slot (PCIe x1 + USB 2.0)
  • Supports Intel AX210, Intel AX200, and MediaTek RZ616 Wi-Fi modules
• User Interface:
  • Full-size QWERTY keyboard support
  • Precision touchpad support
• Security: Fingerprint reader interface
• Power Supply: 55Wh battery support
• Mechanical Characteristics:
  • Dimensions: 296.63 mm × 228.98 mm × 15.85 mm
  • Weight: 1.3 kg
• Mechanical Construction:
  • CNC-machined aluminum top cover
  • Formed aluminum input cover and bottom cover
  • 3.5 kg-rated display hinges

SpacemIT K3 block diagram

Ubuntu 24.10 is the main supported OS (operating system) for the hardware of DC-ROMA RISC-V Mainboard III. It offers developers a reliable platform to develop code using RVA23 toolchains and upstream Linux packages. 

DeepComputing plans to start shipping the boards by May and has opened pre-orders for a $99.99 reservation. The Early Access Program has a $99.99 fully refundable deposit for reserving one of the first units in the first batch. Details on pricing, availability and shipping will be provided soon in the news page section. Users can check the DeepComputing website for more information.

Images used courtesy of deepcomputing