Why Choose NVIDIA HGX B300 NVL16
The HGX B300 NVL16 platform is a major leap in AI infrastructure, engineered
for trillion-parameter model training, large-scale inference, and reasoning workloads.
1.8TB/s
GPU-to-GPU Bandwidth
Key Benefits of HGX B300 NVL16 GPU Servers
Train trillion-parameter models faster
with Blackwell Ultra GPUs and NVLink NVL16 fabric
that delivers record intra-node bandwidth.
Serve AI at scale
with architecture optimized for reasoning,
inference, and high-throughput deployment in production clusters.
Future-proof your data center
with NVIDIA’s latest HGX™ reference design, ready
for InfiniBand or Ethernet fabrics and scalable to multi-rack AI factories.
Ideal Workloads for HGX B300 NVL16 Servers
LLM Training at Scale
Fifth gen NVLink fabric removes bottlenecks for multi-trillion
parameter pre-training and continual training, delivering faster time-to-accuracy for
models.
Inference and RAG
Blackwell Ultra GPUs enable low-latency, large-scale inference,
powering retrieval-augmented generation and production AI deployments.
Agentic AI and Reasoning
Optimized for reasoning-heavy pipelines, the HGX™ B300 NVL16
supports autonomous agents and advanced decision-making models at scale.
NVIDIA HGX B300 NVL16 Server Options
Blackwell Cloud offers customizable server configurations, featuring up to 8x
NVIDIA HGX™ B300 NVL16 GPUs, manufactured by various OEMs.
NVIDIA HGX B300 NVL16 8U
Supermicro SYS-822GS-NB3RT
NVIDIA HGX B300 NVL16 Specifications
| Form Factor |
8x NVIDIA Blackwell Ultra SXM |
| FP4 Tensor Core |
144 PFLOPS | 105 PFLOPS |
| FP8/FP6 Tensor Core |
72 PFLOPS |
| INT8 Tensor Core |
2 POPS |
| FP16/BF16 Tensor Core |
36 PFLOPS |
| TF32 Tensor Core |
18 PFLOPS |
| FP32 |
600 TFLOPS |
| FP64/FP64 Tensor Core |
10 TFLOPS |
| Total Memory |
Up to 2.3 TB |
| NVLink |
Fifth generation |
| NVIDIA NVSwitch™ |
NVLink 5 Switch |
| NVSwitch GPU-to-GPU Bandwidth |
1.8 TB/s |
| Total NVLink Bandwidth |
14.4 TB/s |
| Networking Bandwidth |
1.6 TB/s |
| Attention Performance |
2X |
|
|
|
|
|
|
|
|
|
|
Explore Our High-Performance NVIDIA GPU Servers
NVIDIA HGX B300 NVL16 Servers
Build AI factories that train faster and serve smarter with
the next generation of NVIDIA HGX™ systems, powered by Blackwell Ultra accelerators and fifth
generation NVLink technology.
NVIDIA HGX B200 Servers
Leverage the power of NVIDIA Blackwell GPUs for accelerated AI
and HPC workloads, offering 15x faster inference and 12x lower energy consumption.
NVIDIA RTX PRO 6000 Servers
Unleash Blackwell architecture in your data center with RTX
PRO 6000 Server Edition. Perfect for demanding AI visualization, digital twins, and 3D content
creation workloads.
NVIDIA HGX H200 Servers
Experience enhanced memory capacity and bandwidth over H100,
ideal for large-scale AI model training.
NVIDIA HGX H100 Servers
Optimize your infrastructure with NVIDIA 's best-selling enterprise GPU, delivering
unparalleled performance and scalability.
