NVIDIA B100 Blackwell AI GPU

Type: B100 Discrete Accelerator

Memory Clock: 8Gbps HBM3E

Memory Bus Width: 2×4096-bit

Memory Bandwidth: 8TB/sec

VRAM: 192GB (2x96GB)

FP4 Dense Tensor: 7 PFLOPS

INT8/FP8 Dense Tensor: 3.5 P(FL)OPS

FP16 Dense Tensor: 1.8 PFLOPS

TF32 Dense Tensor: 0.9 PFLOPS

FP64 Dense Tensor: 30 TFLOPS

Interconnects: NVLink 5 (1800GB/sec) + PCIe 6.0 (256GB/sec)

GPU: “Blackwell GPU”

GPU Transistor Count: 208B (2x104B)

TDP: 700W

Manufacturing Process: TSMC 4NP

Interface: SXM-DELTA NEXT 6

Architecture: Blackwell


Pricing and availability will be released soon, expected in Q3 2024.

Categories: ,
  • B100 Discrete Accelerator: Revolutionizing Computational Power

    The B100 Discrete Accelerator marks a significant milestone in the evolution of computing hardware, engineered to cater to the demanding needs of high-performance computing and artificial intelligence. At the heart of its design lies the state-of-the-art 8Gbps HBM3E memory, paired with a dual 4096-bit memory bus, facilitating an unprecedented memory bandwidth of 8TB/sec and a generous 192GB of VRAM. Its computational capabilities are equally impressive, with the accelerator capable of achieving up to 7 PFLOPS in FP4 Dense Tensor operations, 3.5 P(FL)OPS in INT8/FP8, and extending its prowess to 1.8 PFLOPS and 0.9 PFLOPS in FP16 and TF32 tensor operations, respectively, culminating in a robust 30 TFLOPS for FP64 Dense Tensor computations. Integration with high-speed NVLink 5 and PCIe 6.0 interfaces ensures seamless communication at speeds up to 1800GB/sec and 256GB/sec, respectively. The heart of this computational giant is the "Blackwell GPU," boasting a transistor count of 208 billion, all within a surprisingly efficient 700W thermal design power (TDP). Manufactured using TSMC's 4NP process and potentially featuring the next-generation SXM interface, the B100 Discrete Accelerator embodies the pinnacle of the Blackwell architecture's engineering excellence, setting a new benchmark for computational efficiency and performance in the realm of discrete accelerators.

Attribute Specification
Type B100 Discrete Accelerator
Memory Clock 8Gbps HBM3E
Memory Bus Width 2×4096-bit
Memory Bandwidth 8TB/sec
VRAM 192GB (2x96GB)
FP4 Dense Tensor 7 PFLOPS
INT8/FP8 Dense Tensor 3.5 P(FL)OPS
FP16 Dense Tensor 1.8 PFLOPS
TF32 Dense Tensor 0.9 PFLOPS
FP64 Dense Tensor 30 TFLOPS
Interconnects NVLink 5 (1800GB/sec) + PCIe 6.0 (256GB/sec)
GPU “Blackwell GPU”
GPU Transistor Count 208B (2x104B)
TDP 700W
Manufacturing Process TSMC 4NP
Interface SXM-Next?
Architecture Blackwell



    There are no reviews yet.

    Only logged in customers who have purchased this product may leave a review.

    You may also like…

    • NVIDIA DGX B200 Blackwell 1,440GB 4TB AI Supercomputer

    • NVIDIA GB200 NVL72