MSI XpertStation WS300 NVIDIA DGX Station AI Supercomputer
PRODUCT DESCRIPTION
MSI XpertStation WS300 NVIDIA DGX Station with Grace Blackwell Ultra, 748GB Coherent Memory, 1.92TB RAID 1 & Dual 400G ConnectX-8 Networking - Local Warranty
MSI XpertStation WS300 — Datacentre-Class AI in a Workstation
The MSI XpertStation WS300 is an enterprise AI supercomputer built on the NVIDIA DGX Station GB300 platform. The NVIDIA Grace Blackwell Ultra Desktop Superchip delivers up to 20 petaFLOPS of FP4 AI performance with 748 GB of coherent unified memory — enough to work with AI models up to 1 trillion parameters — without a specialised liquid-cooled rack installation.
748 GB Coherent Memory: One Memory Space for Giant Models
252 GB of HBM3e GPU memory at 7.1 TB/s pairs with 496 GB of LPDDR5X CPU memory over NVLink-C2C at 900 GB/s, so models and datasets live in a single 748 GB coherent memory space — accelerating LLM fine-tuning, RAG pipelines, simulations and multi-team AI workloads.
Enterprise Storage and 800G-Class Networking
1.92 TB of enterprise NVMe storage in RAID 1 protects your work from day one, with expansion slots for growth. Dual 400G QSFP112 ports on the NVIDIA ConnectX-8 SuperNIC (up to 800 Gb/s) connect the WS300 into private AI clusters and distributed training environments.
Available in Singapore Through SourceIT
Enterprise procurement with GST invoice and local warranty support. For desktop-scale AI, see the NVIDIA DGX Spark; for an alternative GB300 workstation, see the HP ZGX Fury G1n.
MSI XpertStation WS300 Datasheet — Technical Specifications
General
| Model | MSI XpertStation WS300 (NVIDIA DGX Station GB300 Platform) |
| Product Type | Enterprise Desktop AI Supercomputer / AI Workstation |
| Warranty | MSI Enterprise Warranty, Local Support via SourceIT (Configuration Dependent) |
Processor — NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip
| GPU | 1 × NVIDIA Blackwell Ultra |
| CPU | 1 × NVIDIA Grace — 72-Core Arm Neoverse V2 |
| Interconnect | NVLink-C2C, 900 GB/s |
| AI Performance | Up to 20 PFLOPS FP4 Tensor Core |
| Model Support | AI Models Up to 1 Trillion Parameters |
| GPU Partitioning | Multi-Instance GPU (MIG) — Up to 7 Instances |
Memory
| GPU Memory | 252 GB HBM3e, 7.1 TB/s |
| CPU Memory | 496 GB LPDDR5X, 396 GB/s |
| Total Coherent Memory | 748 GB Unified |
Storage
| Installed Storage | 1.92 TB Enterprise NVMe (2 × M.2 2280 PCIe 5.0) in RAID 1 |
| Expansion | 2 × PCIe 6.0 x4 M.2 2280 Slots Available (4 × M.2 Gen 5 Platform Slots) |
Networking
| SuperNIC | NVIDIA ConnectX-8 — Up to 800 Gb/s Ethernet |
| High-Speed Ports | 2 × QSFP112 (400 Gb/s Each) |
| Standard Ethernet | 10 GbE + 1 GbE Management |
Software & Power
| Operating System | Ubuntu with NVIDIA AI Developer Tools (DGX Software Stack) |
| AI Stack | CUDA, NVIDIA AI Enterprise, TensorRT, NVIDIA NIM, PyTorch/TensorFlow |
| Power | 1,600 W Enterprise High-Efficiency Power Supply |
| Cooling | Enterprise Workstation Cooling Architecture — No Liquid-Cooled Rack Required |
| Optional | Supports Additional NVIDIA RTX PRO Blackwell GPU |
Ordering Information — AI Supercomputers at SourceIT
| MSI XpertStation WS300 (DGX Station GB300) | This Product — In Stock |
| HP ZGX Fury G1n GB300 AI Workstation | Enquire |
| NVIDIA DGX Spark 4TB (Desktop, 1 PFLOP) | 940-54242-0007-000 — In Stock |
| ASUS Ascent GX10 (GB10 Platform, from 1TB) | In Stock |
Datasheet & Downloads
Frequently Asked Questions — MSI XpertStation WS300
How does the WS300 compare to the NVIDIA DGX Spark?
Different class entirely: the WS300 delivers up to 20 PFLOPS FP4 and 748 GB coherent memory for models up to 1 trillion parameters, versus the DGX Spark's 1 PFLOP and 128 GB (200B parameters). Choose the Spark for individual developers; the WS300 for teams and production-scale AI.
Can multiple team members share one WS300?
Yes — Multi-Instance GPU (MIG) partitions the Blackwell Ultra GPU into up to 7 isolated instances, so several users or workloads run concurrently with guaranteed resources.
Does it need a server room?
No liquid-cooled rack is required — the WS300 uses enterprise workstation cooling and a 1,600 W power supply. Plan a dedicated power circuit as you would for a high-end workstation.
What storage does it ship with?
1.92 TB enterprise NVMe in RAID 1 for data protection, with two additional PCIe 6.0 M.2 slots free for expansion.
How do I purchase in Singapore?
SourceIT supplies the WS300 with GST invoice, local warranty support and enterprise procurement assistance — contact us for configuration options and volume pricing.
