HGX™ Server powered by 8x B300 SXM6 GPUs
Get ahead of the competition!
A complete powerhouse AI server with 8x B300 SXM6 GPUs
Meet our latest B300 server, powered by an HGX™ platform with 8 B300 SXM6 GPUs and 2 Intel® Xeon® 6th Generation.
This B300 server is designed for the AI experts in search for power, quickness, and reliability.
Order now to be ahead of the competition with B300 servers!
Godì 1.8E2D-NV8: B300 server
19"
Based
on Intel® Xeon® 6th generation and NVIDIA SXM6 with 8 B300!
Get in touch
Attention: NVIDIA B300 GPUs are under embargo from the United States Department of Commerce towards multiple countries across the world. 2CRSi group and all its subsidiaries are following US trade legislation and will ensure through compliance process is complete before all shipments of its GPU systems. Feel free to ask any question regarding the compliance process to your 2CRSi Sales agent or contacting the U.S. Department of Commerce.
HGX™ 8 NVIDIA B300 270GB
The B300 server with 8 GPUs utilizes Transformer Engine, NVLink, and 2 160GB HBM3e memory to provide optimum performance and easy scaling across any data center, bringing LLMs to the mainstream.
| GPU Memory |
270GB |
| GPU Memory Bandwidth |
7.7TB/s |
| FP8 Tensor Core Performance |
9 PetaFLOPS |
| Form Factor |
SXM |
Dual CPU socket for Intel® Xeon® 6 Processors
AI workloads are more demanding than ever.
Our HGX B300 server, the Godì 1.8E2D-NV8, harnesses the power of next-generation computing with the Intel® Xeon® 6 processors.
With its innovative modular x86 architecture, the Intel® Xeon® 6 processor family empowers data center architects to design and deploy infrastructures purpose-built for diverse workloads across private, public, and hybrid clouds.
This flexibility is made possible through two distinct CPU microarchitectures:
- Performance-cores (P-cores) optimized for high-performance, compute-intensive tasks.
- Efficient-cores (E-cores) tailored for high-density, scale-out, and energy-efficient workloads.
8x 800 Gb/s via onboard NVIDIA ConnectX®-8 SuperNIC™
Experience up to 800 Gb/s of fabric bandwidth and ultra-low-latency connectivity, accelerating model training, scaling inference, and keeping clusters in perfect sync. The ConnectX-8 SuperNIC supports NVIDIA Quantum-X800 InfiniBand and Spectrum-X™ Ethernet natively, and features a built-in PCIe Gen6 switch, eliminating bottlenecks and simplifying your architecture.
Get the most of your HGX B300 by enabling ConnectX-8 features!
Image courtesy of NVIDIA. 
NVLink 5ᵗʰ Generation: Redefining Multi-GPU Performance
NVIDIA’s 5ᵗʰ gen NVLink sets a new standard for GPU connectivity, delivering up to 1.8 TB/s bidirectional bandwidth per GPU, 14× faster than PCIe Gen-5.
With 18 NVLink connections per GPU, data moves effortlessly, powering the most demanding AI training and HPC workloads.
At rack scale, the NVLink Switch unlocks 14.4 TB/s switching capacity, enabling up to 576 GPUs to function as a single, massive accelerator.
This architecture doubles throughput over the previous generation. It ensures ultra-low latency, full memory coherence, and unmatched scalability, perfect for trillion-parameter AI, advanced scientific simulations, and next-generation data centers.
NVIDIA B300 vs H200
Performance
H200 GPU: this graphic cards improves performance significantly for LLMs with up to 2 times faster inference compared to the H100 thanks to the increase in memory capacity.
B300 GPU: performance improves up to 11 times faster inference compared to the H100.
Memory & bandwidth
H200 GPU: features 141GB of HBM3e memory and a bandwidth of 4.8 TB/S, which improves data handling capabilities.
B300 GPU: features 270GB of HBM3e memory and a bandwidth of 7.7TB/S, which improves data handling capabilities.