- GB10 Grace Blackwell Superchip (1petaFLOP AI performance Blackwell GPU, 20-core Arm CPU)
- Workloads 128GB coherent unified system memory
- ~100GB of system memory for user workloads
- Cluster up to two EdgeXpert systems together to work with models of up to 405 billion parameters
- DGX OS & NVIDIA AI software stack
FEATURES
Description
Edge AI Computing Power. Scaled Down.
The MSI EdgeXpert by BRESSNER Technology delivers powerful AI performance at the edge. It brings supercomputer-level capabilities into a compact, desk-side system. As a result, it is ideal for modern edge computing environments.
The system is built on the NVIDIA® DGX™ Spark platform. It is powered by the advanced Grace Blackwell architecture. This enables up to 1 petaFLOP of AI performance. Therefore, it supports demanding workloads such as deep learning, data science, and visualization.
EdgeXpert features a powerful GB10 GPU and a 20-core Arm CPU. It can handle AI models with up to 200 billion parameters. In addition, it offers 128GB of unified memory for efficient data processing. High-speed ConnectX-7 networking allows multiple systems to be connected. This makes it easy to scale performance for larger workloads.
The system offers high flexibility. It can be used as an AI workstation or as a personal AI cloud. At the same time, it integrates easily into existing infrastructures. With support for NVIDIA DGX OS and the AI software stack, workloads can be moved seamlessly between edge, data center, and cloud environments.
Despite its high performance, EdgeXpert remains compact and energy-efficient. It is also easy to deploy. Therefore, it is an ideal solution for organizations that require scalable AI capabilities at the edge.
Specifications
| Product: MSI EdgeXpert | |
|---|---|
| SYSTEM | |
| CPU | NVIDIA® GB10 Grace Blackwell Superchip (Blackwell GPU + 20-core Arm CPU: 10x Cortex-X925 + 10x Cortex-A725) |
| Chipset | NVIDIA® Grace Blackwell Architecture |
| GPU | NVIDIA® GB10 (Blackwell Generation, 5th Gen Tensor Cores, 4th Gen RT Cores) |
| RAM | 128GB LPDDR5x unified system memory |
| Storage | 1TB or 4TB NVMe M.2 (self-encrypting) |
| Expansion | ConnectX-7 Smart NIC (QSFP 200GbE support for system clustering) |
| Operating System | NVIDIA® DGX™ OS |
| AI Performance | Up to 1 petaFLOP (FP4, Sparse) |
| INTERFACE | |
| Ethernet | 1x RJ-45 (10GbE) |
| Networking | ConnectX-7 Smart NIC (high-speed interconnect, up to 2 systems cluster) |
| USB | 4x USB 3.2 Type-C (up to 20Gb/s) |
| Video | 1x HDMI 2.1, 4x DisplayPort 1.4a via USB-C |
| Audio | HDMI multichannel audio output |
| Wireless | Wi-Fi 7*, Bluetooth 5.4 |
| Other | NVENC/NVDEC: 1x / 1x |
| ENVIRONMENTAL | |
| Power Supply | Standard wall outlet (power-efficient design) |
| Operating Temperature | 0 ~ 35°C |
| Storage Temperature | 0 ~ 60°C |
| Vibration / Shock Resistance | – |
| Humidity | 10 ~ 90%, non-condensing |
| MECHANICAL | |
| Dimensions (W x H x D) | 151 x 151 x 52 mm |
| Weight | 1.2 kg |
| Mounting | Desktop (compact design) |
| ADDITIONAL FEATURES | |
| Memory Bandwidth | 273 GB/s |
| Supported Data Formats | TF32, FP16, BF16, INT8, FP8, FP6, FP4 |
| Use Case | AI workstation or personal AI cloud |
| Scalability | Cluster up to 2 systems (up to 405B parameter models) |
