
NVIDIA DGX B300 for Modern AI Factories
The NVIDIA® DGX™ B300 is the foundational building block for modern AI infrastructure, delivering hyperscale-class performance. AI factories and real-time intelligence bring massive challenges to the data center. Models demand unprecedented compute power, and data centers must be redesigned to handle AI workloads.
Powered by NVIDIA Blackwell Ultra architecture and paired with NVIDIA Mission Control software, NVIDIA DGX B300 provides the performance, efficiency, and simplified operations that enterprises need to run AI at scale.
As an NVIDIA Elite Partner and consecutive NVIDIA Partner Network Solution Integration Partner of the Year (2024 and 2025), Exxact is positioned to deliver the NVIDIA DGX B300, providing support for seamless integration, deployment, and optimization. Whether that’s across research labs, data centers, and enterprise AI environments, we aim to provide the hardware to enable your innovations and discoveries.
NVIDIA DGX B300 is available to order from Exxact and will start shipping later this year, Q4 2025.

Image Source: NVIDIA
NVIDIA DGX B300 Specifications Compared with NVIDIA DGX B200
System | NVIDIA DGX B300 | NVIDIA DGX B200 |
GPU | 8x NVIDIA Blackwell Ultra | 8x NVIDIA Blackwell |
Memory | 2.3TB HBM3e (8x 288GB) | 1.4TB HBM3e (8x 192GB) |
Memory Bandwidth | 64TB/s total bandwidth | 64TB/s total bandwidth |
NVSwitch Bandwidth | 1.8TB/s | 1.8TB/s |
Total NVLink Bandwidth | 14.4TB/s | 14.4TB/s |
Networking Bandwidth | 1.6TB/s | 0.8TB/s |
Attention Performance (versus NVIDIA Blackwell) | 2x | 1x |
CPU | 2x Intel Xeon 6776P (Total 128 Cores) |
2x Intel Xeon Platinum 8570 (Total 112 Cores) |
System Memory | Up to 4TB | 2TB (configurable up to 4TB) |
Networking | - 8x OSFP ports serving 8x single-port NVIDIA ConnectX-8 VPI (Up to 800Gb/s InfiniBand/Ethernet) - 2x dual-port QSFP112 NVIDIA BlueField-3 DPU (Up to 400Gb/s InfiniBand/Ethernet) |
- 4x OSFP ports serving 8x single-port NVIDIA ConnectX-7 VPI (Up to 400Gb/s InfiniBand/Ethernet) - 2x dual-port QSFP112 NVIDIA BlueField-3 DPU (Up to 400Gb/s InfiniBand/Ethernet) |
Management Network | 1GbE onboard NIC with RJ45 1GbE RJ45 Host BMC |
10Gb/s onboard NIC with RJ45 100b/s dual-port ethernet |
Storage | OS: 2x 1.9TB NVMe M.2 Storage: 8x 3.84TB NVMe E1.S |
OS: 2x 1.9TB NVMe M.2 Storage: 8x 3.84TB NVMe U.2 |
Rack Units | 10U | 10U |
System Power Usage | 14.3kW Max | 14.3kW Max |
GPU Baseboard Performance | — | — |
FP4 (Sparce | Dense) | 144 PFLOPS | 108 PFLOPS | 144 PFLOPS | 72 PFLOPS |
FP8/FP6 | 72 PFLOPS | 72 PFLOPS |
INT8 | 3 POPS | 72 POPS |
FP16/BF16 | 36 PFLOPS | 36 PFLOPS |
TF32 | 18 PFLOPs | 18 PFLOPs |
FP32 | 600 TFLOPS | 600 TFLOPS |
FP64 | 10 TFLOPS | 296 TFLOPS |
Compared with the NVIDIA DGX B200, the NVIDIA DGX B300 delivers a major leap forward in performance and scalability. Together, these advances make the DGX B300 a stronger foundation for enterprises building AI factories that demand higher throughput, larger model support, and greater operational efficiency.
- Its FP4 dense performance has been significantly increased, enabling faster, more efficient reasoning workloads.
- GPU memory has grown from 1.4TB to 2.4TB, allowing larger models to run in memory without partitioning or complex workarounds.
- Networking bandwidth has also expanded, ensuring faster data movement across nodes and more seamless scaling in multi-system deployments.
NVIDIA DGX B300 has lower INT8 and FP64 performance compared to the NVIDIA DGX B200, which will impact INT8 optimized inferencing and double-precision imperative scientific computing. This will not affect performance in extreme AI training, mixed-precision workloads, AI reasoning, and inferencing, where the NVIDIA DGX B300 shines.
Why the DGX B300 Matters
The NVIDIA DGX B300 is built for enterprises ready to scale AI from development to deployment. Exxact customers have deployed NVIDIA DGX systems to deliver the compute, efficiency, and simplicity needed to compete in the era of AI reasoning.
- Real-Time AI Powerhouse: With NVIDIA Blackwell Ultra GPUs, the DGX B300 combines high-performance training and real-time inference on a single platform. It enables organizations to run advanced generative AI and reasoning workloads at scale without compromise.
- Efficient and Sustainable Innovation: Designed with flexible power architecture, the system delivers higher performance per watt than previous generations, helping enterprises balance growth with sustainability goals while lowering operational costs.
- Blueprint for Modern Data Centers: The DGX B300 is an all-in-one system that is built to integrate into a variety of pre-qualified data center environments. Exxact is an NVIDIA Elite Partner for the NVIDIA DGX Platform; talk to an Exxact representative today for more information.
- Simplified AI Operations: With NVIDIA Mission Control software, the DGX B300 reduces operational complexity by automating cluster bring-up, workload management, and resiliency. IT teams can maximize utilization and uptime while developers stay focused on innovation.
The result is faster time-to-market, higher efficiency, and the ability to run AI at an unprecedented scale. For enterprises, the DGX B300 is a critical leap toward AI factories.
Why Partner with Exxact for DGX B300
Exxact helps enterprises get NVIDIA DGX systems up and running faster. As an NVIDIA Elite Partner, we deliver more than the system itself: we provide the expertise to integrate DGX B300 into your environment, streamline deployment, and ensure long-term operational success.
From research teams to enterprise-scale AI factories, Exxact enables customers to unlock the full performance and value of their DGX platform investment. We have installed DGX systems at multiple research and business institutions to empower their research goals.
The DGX B300 is engineered to meet the demands of AI reasoning at scale. With Exxact as your partner, you can accelerate innovation, reduce infrastructure complexity, and stay ahead in the age of generative AI. Contact Exxact today to learn how the DGX B300 can accelerate your AI strategy.

Accelerate AI Training an NVIDIA DGX
Training AI models on massive datasets can be accelerated exponentially with the right system. It's not just a high-performance computer, but a tool to propel and accelerate your research. Deploy multiple NVIDIA DGX nodes for increased scalability. DGX B200 is available today. DGX B300 is shipping soon!
Get a Quote Today
Exxact Offers NVIDIA DGX B300 Shipping Soon
NVIDIA DGX B300 for Modern AI Factories
The NVIDIA® DGX™ B300 is the foundational building block for modern AI infrastructure, delivering hyperscale-class performance. AI factories and real-time intelligence bring massive challenges to the data center. Models demand unprecedented compute power, and data centers must be redesigned to handle AI workloads.
Powered by NVIDIA Blackwell Ultra architecture and paired with NVIDIA Mission Control software, NVIDIA DGX B300 provides the performance, efficiency, and simplified operations that enterprises need to run AI at scale.
As an NVIDIA Elite Partner and consecutive NVIDIA Partner Network Solution Integration Partner of the Year (2024 and 2025), Exxact is positioned to deliver the NVIDIA DGX B300, providing support for seamless integration, deployment, and optimization. Whether that’s across research labs, data centers, and enterprise AI environments, we aim to provide the hardware to enable your innovations and discoveries.
NVIDIA DGX B300 is available to order from Exxact and will start shipping later this year, Q4 2025.

Image Source: NVIDIA
NVIDIA DGX B300 Specifications Compared with NVIDIA DGX B200
System | NVIDIA DGX B300 | NVIDIA DGX B200 |
GPU | 8x NVIDIA Blackwell Ultra | 8x NVIDIA Blackwell |
Memory | 2.3TB HBM3e (8x 288GB) | 1.4TB HBM3e (8x 192GB) |
Memory Bandwidth | 64TB/s total bandwidth | 64TB/s total bandwidth |
NVSwitch Bandwidth | 1.8TB/s | 1.8TB/s |
Total NVLink Bandwidth | 14.4TB/s | 14.4TB/s |
Networking Bandwidth | 1.6TB/s | 0.8TB/s |
Attention Performance (versus NVIDIA Blackwell) | 2x | 1x |
CPU | 2x Intel Xeon 6776P (Total 128 Cores) |
2x Intel Xeon Platinum 8570 (Total 112 Cores) |
System Memory | Up to 4TB | 2TB (configurable up to 4TB) |
Networking | - 8x OSFP ports serving 8x single-port NVIDIA ConnectX-8 VPI (Up to 800Gb/s InfiniBand/Ethernet) - 2x dual-port QSFP112 NVIDIA BlueField-3 DPU (Up to 400Gb/s InfiniBand/Ethernet) |
- 4x OSFP ports serving 8x single-port NVIDIA ConnectX-7 VPI (Up to 400Gb/s InfiniBand/Ethernet) - 2x dual-port QSFP112 NVIDIA BlueField-3 DPU (Up to 400Gb/s InfiniBand/Ethernet) |
Management Network | 1GbE onboard NIC with RJ45 1GbE RJ45 Host BMC |
10Gb/s onboard NIC with RJ45 100b/s dual-port ethernet |
Storage | OS: 2x 1.9TB NVMe M.2 Storage: 8x 3.84TB NVMe E1.S |
OS: 2x 1.9TB NVMe M.2 Storage: 8x 3.84TB NVMe U.2 |
Rack Units | 10U | 10U |
System Power Usage | 14.3kW Max | 14.3kW Max |
GPU Baseboard Performance | — | — |
FP4 (Sparce | Dense) | 144 PFLOPS | 108 PFLOPS | 144 PFLOPS | 72 PFLOPS |
FP8/FP6 | 72 PFLOPS | 72 PFLOPS |
INT8 | 3 POPS | 72 POPS |
FP16/BF16 | 36 PFLOPS | 36 PFLOPS |
TF32 | 18 PFLOPs | 18 PFLOPs |
FP32 | 600 TFLOPS | 600 TFLOPS |
FP64 | 10 TFLOPS | 296 TFLOPS |
Compared with the NVIDIA DGX B200, the NVIDIA DGX B300 delivers a major leap forward in performance and scalability. Together, these advances make the DGX B300 a stronger foundation for enterprises building AI factories that demand higher throughput, larger model support, and greater operational efficiency.
- Its FP4 dense performance has been significantly increased, enabling faster, more efficient reasoning workloads.
- GPU memory has grown from 1.4TB to 2.4TB, allowing larger models to run in memory without partitioning or complex workarounds.
- Networking bandwidth has also expanded, ensuring faster data movement across nodes and more seamless scaling in multi-system deployments.
NVIDIA DGX B300 has lower INT8 and FP64 performance compared to the NVIDIA DGX B200, which will impact INT8 optimized inferencing and double-precision imperative scientific computing. This will not affect performance in extreme AI training, mixed-precision workloads, AI reasoning, and inferencing, where the NVIDIA DGX B300 shines.
Why the DGX B300 Matters
The NVIDIA DGX B300 is built for enterprises ready to scale AI from development to deployment. Exxact customers have deployed NVIDIA DGX systems to deliver the compute, efficiency, and simplicity needed to compete in the era of AI reasoning.
- Real-Time AI Powerhouse: With NVIDIA Blackwell Ultra GPUs, the DGX B300 combines high-performance training and real-time inference on a single platform. It enables organizations to run advanced generative AI and reasoning workloads at scale without compromise.
- Efficient and Sustainable Innovation: Designed with flexible power architecture, the system delivers higher performance per watt than previous generations, helping enterprises balance growth with sustainability goals while lowering operational costs.
- Blueprint for Modern Data Centers: The DGX B300 is an all-in-one system that is built to integrate into a variety of pre-qualified data center environments. Exxact is an NVIDIA Elite Partner for the NVIDIA DGX Platform; talk to an Exxact representative today for more information.
- Simplified AI Operations: With NVIDIA Mission Control software, the DGX B300 reduces operational complexity by automating cluster bring-up, workload management, and resiliency. IT teams can maximize utilization and uptime while developers stay focused on innovation.
The result is faster time-to-market, higher efficiency, and the ability to run AI at an unprecedented scale. For enterprises, the DGX B300 is a critical leap toward AI factories.
Why Partner with Exxact for DGX B300
Exxact helps enterprises get NVIDIA DGX systems up and running faster. As an NVIDIA Elite Partner, we deliver more than the system itself: we provide the expertise to integrate DGX B300 into your environment, streamline deployment, and ensure long-term operational success.
From research teams to enterprise-scale AI factories, Exxact enables customers to unlock the full performance and value of their DGX platform investment. We have installed DGX systems at multiple research and business institutions to empower their research goals.
The DGX B300 is engineered to meet the demands of AI reasoning at scale. With Exxact as your partner, you can accelerate innovation, reduce infrastructure complexity, and stay ahead in the age of generative AI. Contact Exxact today to learn how the DGX B300 can accelerate your AI strategy.

Accelerate AI Training an NVIDIA DGX
Training AI models on massive datasets can be accelerated exponentially with the right system. It's not just a high-performance computer, but a tool to propel and accelerate your research. Deploy multiple NVIDIA DGX nodes for increased scalability. DGX B200 is available today. DGX B300 is shipping soon!
Get a Quote Today