| HGX B200 | HGX B100 | |
|---|---|---|
| HGX B200 8-GPU | HGX B100 8-GPU | GPUs |
| 8x NVIDIA B200 SXM | 8x NVIDIA B100 SXM | Form factor |
| 144 PFLOPS | 112 PFLOPS | FP4 Tensor Core |
| 72 PFLOPS | 56 PFLOPS | FP8/FP6 Tensor Core |
| 72 POPS | 56 POPS | INT8 Tensor Core |
| 36 PFLOPS | 28 PFLOPS | FP16/BF16 Tensor Core |
| 18 PFLOPS | 14 PFLOPS | TF32 Tensor Core |
| 640 TFLOPS | 480 TFLOPS | FP32 |
| 320 TFLOPS | 240 TFLOPS | FP64 |
| 320 TFLOPS | 240 TFLOPS | FP64 Tensor Core |
| Up to 1.5TB | Up to 1.5TB | Memory |
| Fifth generation | Fifth generation | NVIDIA NVLink |
| Fourth generation | Fourth generation | NVIDIA NVSwitch™ |
| 1.8TB/s | 1.8TB/s | NVSwitch GPU-to-GPU bandwidth |
| 14.4TB/s | 14.4TB/s | Total aggregate bandwidth |
| HGX H200 | ||
|---|---|---|
| 4-GPU | 8-GPU | |
| HGX H200 4-GPU | HGX H200 8-GPU | GPUs |
| 4x NVIDIA H200 SXM | 8x NVIDIA H200 SXM | Form factor |
| 16 PFLOPS | 32 PFLOPS | FP8/FP6 Tensor Core |
| 16 POPS | 32 POPS | INT8 Tensor Core |
| 8 PFLOPS | 16 PFLOPS | FP16/BFLOAT16 Tensor Core |
| 4 PFLOPS | 8 PFLOPS | TF32 Tensor Core |
| 270 TFLOPS | 540 TFLOPS | FP32 |
| 140 TFLOPS | 270 TFLOPS | FP64 |
| 270 TFLOPS | 540 TFLOPS | FP64 Tensor Core |
| Up to 564GB | Up to 1.1TB | Memory |
| Fourth generation | Fourth generation | NVLink |
| N/A | Third generation | NVSwitch |
| N/A | 900GB/s | NVSwitch GPU-to-GPU bandwidth |
| 3.6TB/s | 7.2TB/s | Total aggregate bandwidth |
| HGX H100 | ||
|---|---|---|
| 4-GPU | 8-GPU | |
| HGX H100 4-GPU | HGX H100 8-GPU | GPUs |
| 4x NVIDIA H100 SXM | 8x NVIDIA H100 SXM | Form factor |
| 268TF/4PF/8PF/16PF/16 POPS | 535TF/8PF/16PF/32PF/32 POPS | HPC and AI compute (FP64/TF32/FP16/FP8/INT8)* |
| 16 PFLOPS | 32 PFLOPS | FP8/FP6 Tensor Core |
| 16 POPS | 32 POPS | INT8 Tensor Core |
| 8 PFLOPS | 16 PFLOPS | FP16/BFLOAT16 Tensor Core |
| 4 PFLOPS | 8 PFLOPS | TF32 Tensor Core |
| 270 TFLOPS | 540 TFLOPS | FP32 |
| 140 TFLOPS | 270 TFLOPS | FP64 |
| 270 TFLOPS | 540 TFLOPS | FP64 Tensor Core |
| Up to 320GB | Up to 640GB | Memory |
| Fourth generation | Fourth generation | NVLink |
| N/A | Third generation | NVSwitch |
| N/A | 900GB/s | NVSwitch GPU-to-GPU bandwidth |
| 3.6TB/s | 7.2TB/s | Total aggregate bandwidth |
شرکت دانش بنیان نوران ارتباطات پایدار در سال ۱۳۸۷ فعالیت خود را به عنوان یکی از برترین ارائه دهندگان تجهیزات شبکه و اجرای شبکه های مبتنی بر فناوری اطلاعات و انجام فرآیند های مشاوره، طراحی، نصب و راه اندازی انواع شبکه‌های اینترنتی با بهره گیری از آخرین تکنولوژی های روز دنیا آغاز نمود.
