The EdgeXpert is MSI's variant of NVIDIA's DGX Spark platform, which is a compact system for entry-level AI development. Its most notable feature is a huge 128GB of shared memory space, available to both sides of the Grace Blackwell GB10 superchip. The CPU side is made up of 20 Arm cores, 10 each of two versions, while the GPU side is a Blackwell-based design roughly equivalent to the GeForce RTX 5070.
Four USB-C ports are provided on the back, three of which support DisplayPort output - along with a single HDMI port. High-speed networking options are also built-in, with 10Gb Ethernet, WiFi, Bluetooth, and dual NVIDIA ConnectX-7 200G QSFP ports. Those support NVLink-C2C, allowing two of these to be paired together and combine their compute capabilities.
In terms of software, the EdgeXpert comes with an NVIDIA-customized version of Ubuntu Linux, branded as DGX OS, as well as NVIDIA development tools pre-installed.
Four USB-C ports are provided on the back, three of which support DisplayPort output - along with a single HDMI port. High-speed networking options are also built-in, with 10Gb Ethernet, WiFi, Bluetooth, and dual NVIDIA ConnectX-7 200G QSFP ports. Those support NVLink-C2C, allowing two of these to be paired together and combine their compute capabilities.
In terms of software, the EdgeXpert comes with an NVIDIA-customized version of Ubuntu Linux, branded as DGX OS, as well as NVIDIA development tools pre-installed.
William George (Go-To-Market (GTM) Manager) Says:
This system effectively pairs mid-range GPU compute with a huge amount of usable memory space, which is a unique combination and impressive for the price point. It is a solid option for getting started with AI development, allowing users to work with larger models than a traditional desktop PC, but is limited by the speed of compute. The next step up would be to a workstation with a few RTX PRO Blackwell GPUs or the DGX Station with an even larger shared memory space.
This system effectively pairs mid-range GPU compute with a huge amount of usable memory space, which is a unique combination and impressive for the price point. It is a solid option for getting started with AI development, allowing users to work with larger models than a traditional desktop PC, but is limited by the speed of compute. The next step up would be to a workstation with a few RTX PRO Blackwell GPUs or the DGX Station with an even larger shared memory space.
John Dinning (Tier 2 R&D Engineer) Says:
While the EdgeXpert does have a WiFi7 controller, the effective range is extremely limited due to the lack of an external antenna. We strongly recommend using one of the wired network interfaces instead, which will also provide substantially higher peak performance.
While the EdgeXpert does have a WiFi7 controller, the effective range is extremely limited due to the lack of an external antenna. We strongly recommend using one of the wired network interfaces instead, which will also provide substantially higher peak performance.
Model: MSI EdgeXpert-13S
Specifications
| Chipset Manufacturer | NVIDIA |
| Wired Network Connectivity | 10G Ethernet + Dual NVIDIA ConnectX-7 200G QSFP |
| Wireless Network Connectivity | WiFi 7 |
| Cooling System | Heatsink and Fan |
| Number of Fans | 1 |
| Color | Black |
| Windows Support | No |
| Linux Support | Yes |
Memory Support | |
| Type | LPDDR5x |
| Included Capacity | 128GB |
Onboard Video | |
| Chipset Manufacturer | NVIDIA |
| Processor Type | GB10 |
| HDMI Version | 2.1 |
| Notes: | 3x DisplayPort 1.4a via USB-C |
External Connections | |
| USB 3.2 Type C | 4 |
| LAN (RJ45) | 1 |
Internal Connections | |
| M.2 Connector | 1 |
| Notes: | 4TB M.2 Pre-Installed |
Additional Information | |
| Bluetooth | Yes (5.4) |
Dimensions | |
| Height | 50.5 mm (2.0 in) |
| Width | 150 mm (5.9 in) |
| Depth | 150 mm (5.9 in) |
| Net Weight | 1.26 kg (2.8 lbs) |
Power Supply | |
| Input Voltage | 100-240 Vac |
| Wattage at 120V | 240 W |
| Wattage at 240V | 240 W |
AC Adapter Output | |
| DC Voltage | 48 V |
| Amperage | 5 A |
Integrated CPU | |
| Total Number of Cores | 20 |
| Notes | 10 Arm Cortex-X925 + 10 Arm Cortex-A725 |