NVIDIA Hopper H200 GPUs Supercharged With World's Fastest HBM3e Memory, Grace Hopper Superchips Power Jupiter Supercomputer

Photo of author
Written By Editor

Who keeps posting articles without emotional mental changes

NVIDIA has revealed its brand name brand-new H200 Hopper GPU which now comes geared up with the world's fastest HBM3e memory from Micron. In addition to the brand-new AI platforms, NVIDIA likewise revealed a significant supercomputer win with its Grace Hopper Superchips that now power the Exaflop Jupiter supercomputer.

NVIDIA Continues To Build AI Momentum With Upgraded Hopper GPUs, Grace Hopper Superchips & Supercomputer Wins

NVIDIA's H100 GPUs are the most extremely required AI chips in the market up until now however the green group wishes to provide a lot more efficiency to its consumers. Get In, HGX H200, the current HPC & computing platform for AI which is powered by H200 Tensor Core GPUs. These GPUs include the current Hopper optimizations on both software and hardware & while providing the world's fastest memory option to date.

Associated Story NVIDIA Blackwell B100 GPUs To More Than Double The Performance of Hopper H200 GPUs In 2024

The NVIDIA H200 GPUs are geared up with Micron's HBM3e option with memory capabilities of approximately 141 GB and approximately 4.8 TB/s of bandwidth which is 2.4 x more bandwidth and double the capability versus the NVIDIA A100. This brand-new memory option permits NVIDIA to almost double the AI reasoning efficiency versus its H100 GPUs in applications such as Llama 2 (70 Billion specification LLM). The current improvements in the TensorRT-LLM suite have likewise led to substantial efficiency gains in a large variety of AI applications.

In regards to options, the NVIDIA H200 GPUs will be offered in a large range of HGX H200 servers with 4 and 8-way GPU setups. An 8-way setup of H200 GPUs in an HGX system will offer as much as 32 PetaFLOPs of FP8 calculate efficiency and 1.1 TB of memory capabilities.

NVIDIA H200 GPU: Supercharged With HBM3e Memory, Available In Q2 2024

The GPUs will likewise work with the existing HGX H100 systems, making it simpler for clients to update their platforms. NVIDIA partners such as ASUS, ASRock Rack, Dell, Eviden, GIGABYTE, Hewlett Packard Enterprise, Ingrasys, Lenovo, QCT, Wiwynn, Supermicro, and Wistron, will use upgraded services when the H200 GPUs appear in the 2nd quarter of 2024.

NVIDIA Grace Hopper Superchips Power 1-Exaflop Jupiter Supercomputer In addition to the H200 GPU statement, NVIDIA has actually likewise revealed a significant supercomputer win powered by its Grace Hopper Superchips (GH200). The Supercomputer is referred to as Jupiter and will be found at the Forschungszentrum Jülich center in Germany as a part of the EuroHPC Joint Undertaking and contracted to Eviden and ParTec. The supercomputer will be utilized for Material Science, Climate Research, Drug Discovery, and More. This is likewise the 2nd supercomputer that NVIDIA revealed in November with the previous one being the Isambard-AI, providing to 21 Exaflops of AI efficiency.

In regards to setup, the Jupiter Supercomputer is based upon Eviden's BullSequana XH3000 that makes usage of a totally liquid-cooled architecture. It boasts an overall of 24,000 NVIDIA GH200 Grace Hopper Superchips which are adjoined utilizing the business's Quantum-2 Infiniband. Thinking about that each Grace CPU loads 288 Neoverse cores, we are taking a look at practically 7 Million ARM cores on the CPU side alone for Jupiter (6,912,000 to be precise).

Performance metrics consist of 90 Exaflops of AI training & 1 Exaflop of high-performance calculate. The supercomputer is anticipated to be set up in 2024. In general, these are some significant updates by NVIDIA as it continues to lead the charge of the AI world with its effective software and hardware innovations.

NVIDIA HPC/ AI GPUs NVIDIA Tesla Graphics Card NVIDIA H200 (SXM5) NVIDIA H100 (SMX5) NVIDIA H100 (PCIe) NVIDIA A100 (SXM4) NVIDIA A100 (PCIe4) Tesla V100S (PCIe) Tesla V100 (SXM2) Tesla P100 (SXM2) Tesla P100
(PCI-Express) Tesla M40
(PCI-Express) Tesla K40
(PCI-Express) GPU GH200 (Hopper) GH100 (Hopper) GH100 (Hopper) GA100 (Ampere) GA100 (Ampere) GV100 (Volta) GV100 (Volta) GP100 (Pascal) GP100 (Pascal) GM200 (Maxwell) GK110 (Kepler) Process Node 4nm 4nm 4nm 7nm 7nm 12nm 12nm 16nm 16nm 28nm 28nm Transistors 80 Billion 80 Billion 80 Billion 54.2 Billion 54.2 Billion 21.1 Billion 21.1 Billion 15.3 Billion 15.3 Billion 8 Billion 7.1 Billion GPU Die Size 814mm2 814mm2 814mm2 826mm2 826mm2 815mm2 815mm2 610 mm2 610 mm2 601 mm2 551 mm2 SMs 132 132 114 108 108 80 80 56 56 24 15 TPCs 66 66 57 54 54 40 40 28 28 24 15 L2 Cache Size 51200 KB 51200 KB 51200 KB 40960 KB 40960 KB 6144 KB 6144 KB 4096 KB 4096 KB 3072 KB 1536 KB FP32 CUDA Cores Per SM 128 128 128 64 64 64 64 64 64 128 192 FP64 CUDA Cores/ SM 128 128 128 32 32 32 32 32 32 4 64 FP32 CUDA Cores 16896 16896 14592 6912 6912 5120 5120 3584 3584 3072 2880 FP64 CUDA Cores 16896 16896 14592 3456 3456 2560 2560 1792 1792 96 960 Tensor Cores 528 528 456 432 432 640 640 N/A N/A N/A N/A Texture Units 528 528 456 432 432 320 320 224 224 192 240 Increase Clock ~ 1850 MHz ~ 1850 MHz ~ 1650 MHz 1410 MHz 1410 MHz 1601 MHz 1530 MHz 1480 MHz 1329MHz 1114 MHz 875 MHz TOPs (DNN/AI) 3958 TOPs 3958 TOPs 3200 TOPs 2496 TOPs 2496 TOPs 130 TOPs 125 TOPs N/A N/A N/A N/A FP16 Compute 1979 TFLOPs 1979 TFLOPs 1600 TFLOPs 624 TFLOPs 624 TFLOPs 32.8 TFLOPs 30.4 TFLOPs 21.2 TFLOPs 18.7 TFLOPs N/A N/A FP32 Compute 67 TFLOPs 67 TFLOPs 800 TFLOPs 156 TFLOPs
(19.5 TFLOPs basic) 156 TFLOPs
(19.5 TFLOPs basic) 16.4 TFLOPs 15.7 TFLOPs 10.6 TFLOPs 10.0 TFLOPs 6.8 TFLOPs 5.04 TFLOPs FP64 Compute 34 TFLOPs 34 TFLOPs 48 TFLOPs 19.5 TFLOPs
(9.7 TFLOPs basic) 19.5 TFLOPs
(9.7 TFLOPs basic) 8.2 TFLOPs 7.80 TFLOPs 5.30 TFLOPs 4.7 TFLOPs 0.2 TFLOPs 1.68 TFLOPs Memory Interface 5120-bit HBM3e 5120-bit HBM3 5120-bit HBM2e 6144-bit HBM2e 6144-bit HBM2e 4096-bit HBM2 4096-bit HBM2 4096-bit HBM2 4096-bit HBM2 384-bit GDDR5 384-bit GDDR5 Memory Size As much as 141 GB HBM3e @ 6.5 Gbps As much as 80 GB HBM3 @ 5.2 Gbps As much as 80 GB HBM2e @ 2.0 Gbps Approximately 40 GB HBM2 @ 1.6 TB/s
As much as 80 GB HBM2 @ 1.6 TB/s Approximately 40 GB HBM2 @ 1.6 TB/s
Approximately 80 GB HBM2 @ 2.0 TB/s 16 GB HBM2 @ 1134 GB/s 16 GB HBM2 @ 900 GB/s 16 GB HBM2 @ 732 GB/s 16 GB HBM2 @ 732 GB/s
12 GB HBM2 @ 549 GB/s 24 GB GDDR5 @ 288 GB/s 12 GB GDDR5 @ 288 GB/s TDP 700W 700W 350W 400W 250W 250W 300W 300W 250W 250W 235W

Categories PC

Leave a Comment