کارت گرافیک انویدیا H100 NVL ظرفیت ۹۴ گیگابایت HBM3

بدون رابط

حرفه ای, دیتاسنتر, رندرینگ, علم داده (Data Science), کاربری صنعتی, محاسبه برنامه های کاربردی فشرده, هوش مصنوعی AI, ورک استیشن یا ایستگاه کاری, یادگیری عمیق DL, یادگیری ماشین ML

NVIDIA, NVIDIA Data Center GPUs

NVIDIA H100 94GB HBM3 Data Center PCIe GPU Accelerator

آکبند

Release Date: Mar 21st, 2023 Availability: 2023 Generation: Server Hopper (Hxx) Predecessor: Server Ada Successor: Server Blackwell Production: Active Bus Interface: PCIe 5.0 x16

GPU Name : GH100 Architecture : Hopper Foundry : TSMC Process Size : 5 nm Transistors : 80,000 million Density : 98.3M / mm² Die Size : 814 mm² GH100 GPU Notes: Tensor Cores: 4th Gen NVENC: No Support NVDEC: No Support PureVideo HD: VP11 VDPAU: Feature Set K

Shading Units : 16896 TMUs : 528 ROPs : 24 SM Count : 132 Tensor Cores: 528 L1 Cache : 256KB (per SM) L2 Cache : 50 MB

Memory Size : 94 GB Memory Type : HBM3 Memory Bus : 5120 bit Bandwidth : 3.9 TB/s

Base Clock : 1665 MHz Boost Clock : 1837 MHz Memory Clock : 1313MHz, 5.3 Gbps effective

No outputs

Pixel Rate: 44.09 GPixel/s Texture Rate: 969.9 GTexel/s FP16 (half): 248.3 TFLOPS (4:1) FP32 (float): 62.08 TFLOPS FP64 (double): 31.04 TFLOPS (1:2) FP64: 30 TFLOPS FP64 Tensor Core: 60 TFLOPS FP32: 60 TFLOPS TF32 Tensor Core: 835 TFLOPS* BFLOAT16 Tensor Core: 1,671 TFLOPS* FP16 Tensor Core: 1,671 TFLOPS* FP8 Tensor Core: 3,341 TFLOPS* INT8 Tensor Core: 3,341 TOPS*

DirectX : N/A OpenGL : N/A OpenCL : 3.0 Vulkan : N/A CUDA : 9.0 Shader Model : N/A

NVIDIA H100 GPUs feature fourth-generation Tensor Cores and theTransformer Engine with FP8 precision, further extending NVIDIA’smarket-leading AI leadership with up to 9X faster training and anincredible 30X inference speedup on large language models. Forhigh-performance computing (HPC) applications, H100 triples thefloating-point operations per second (FLOPS) of FP64 and addsdynamic programming (DPX) instructions to deliver up to 7X higherperformance. With second-generation Multi-Instance GPU (MIG),built-in NVIDIA confidential computing, and NVIDIA NVLink SwitchSystem, H100 securely accelerates all workloads for every data centerfrom enterprise to exascale.

Slot Width : Dual-slot TDP: 350-400 W Suggested PSU : 1100 W Outputs : No outputs Power Connectors : 8-pin EPS

Slot Width : Dual-slot Length : 268 mm - 10.6 inches Width : 111 mm - 4.4 inches