NVIDIA Unveils World’s First Data Center Accelerated CPU, the Grace Hopper Superchip

The computing universe took a giant leap forward today as NVIDIA, the renowned technology company, unveiled the GH200 Grace Hopper Superchip, the world’s first data center accelerated CPU specifically architected for large-scale artificial intelligence (AI) and high-performance computing (HPC) applications.

Named after the famous computer scientist Grace Hopper, this innovative CPU superchip is set to revolutionize the technology landscape by delivering up to 10 times higher performance for applications running on terabytes of data. This could fast track scientific advancements and bring unprecedented solutions to some of the world’s most complex problems.

CPU+GPU Synergy: The Foundation of the GH200 Grace Hopper Superchip

The NVIDIA Grace Hopper Superchip uniquely blends the Grace and Hopper architectures using NVIDIA’s NVLink-C2C, creating a CPU+GPU coherent memory model that dramatically enhances computing efficiency. This integration allows the CPU and GPU to work in harmony, optimizing AI and HPC applications on a giant scale.

Grace CPU: A Symphony of Power and Efficiency

The Grace CPU incorporated in the superchip delivers twice the performance per watt compared to standard x86-64 platforms, earning it the title of the world’s fastest Arm data center CPU. This breakthrough was made possible by a unique design featuring 72 Neoverse V2 Armv9 cores combined with up to 480GB of server-class LPDDR5X memory with ECC.

Unparalleled Performance with the Hopper H100 GPU

The superchip also incorporates the Hopper H100 GPU, NVIDIA’s ninth-generation data center GPU. This GPU heralds a leap in performance for large-scale AI and HPC over its predecessor, the NVIDIA A100 Tensor Core GPU. The H100 flaunts multiple innovations, including new fourth-generation Tensor Cores for enhanced matrix computations and a novel Transformer Engine for improved AI training and inference.

Coherent Memory: The Driving Force of The Superchip

At the heart of the Grace Hopper Superchip is the NVLink-C2C, a memory-coherent, high-bandwidth, and low-latency interconnect. It delivers up to 900 GB/s of total bandwidth, a rate seven times higher than the PCIe Gen5 lanes typically used in accelerated systems. This architectural marvel enables the GPU to directly utilize the NVIDIA Grace CPU’s memory at high bandwidth, while the NVIDIA NVLink Switch System provides all GPU threads running on up to 256 NVLink-connected GPUs access to up to 150 terabytes of memory.

A New Era of High-Performance Computing

The unveiling of the Grace Hopper Superchip marks a new era in high-performance computing. By bridging the strengths of GPUs and CPUs, the superchip provides a heterogenous accelerated platform for HPC and AI workloads. This simplified and productive programming model allows engineers and scientists to focus on resolving some of the world’s most critical issues.

Full NVIDIA Platform Support and Extended Ecosystem

The NVIDIA Grace Hopper Superchip is a significant addition to the diverse ecosystem of 64-bit Arm processors. It offers full NVIDIA platform support and runs the same containers, application binaries, and operating systems as other Arm products but at greater speed. NVIDIA’s software expertise is also baked into the superchip, supported by the full NVIDIA software stack, including the NVIDIA HPC, NVIDIA AI, and NVIDIA Omniverse™ platform.

Closing Note

With its revolutionary design and unprecedented power, the NVIDIA Grace Hopper Superchip is set to redefine the future of computing. Its introduction not only marks a major stride for NVIDIA but also a significant leap for the tech industry as a whole. It is, without doubt, a critical catalyst for future advancements in AI, HPC, and beyond.

The Grace Hopper Superchip is expected to lay the foundation for the next generation of HPC supercomputers and AI factories, enabling customers to take on larger datasets, more complex models, and new workloads.

https://www.nvidia.com/en-us/data-center/grace-hopper-superchip/

{
"seed": 313571826,
"used_random_seed": true,
"negative_prompt": "",
"num_outputs": 1,
"num_inference_steps": 50,
"guidance_scale": 7.5,
"width": 512,
"height": 512,
"vram_usage_level": "balanced",
"sampler_name": "euler",
"use_stable_diffusion_model": "Dreamshaper_3.32_baked_vae_clip_fix",
"clip_skip": false,
"tiling": "none",
"use_vae_model": "vae-ft-mse-840000-ema-pruned",
"stream_progress_updates": true,
"stream_image_progress": false,
"show_only_filtered_image": true,
"block_nsfw": false,
"output_format": "jpeg",
"output_quality": 75,
"output_lossless": false,
"metadata_output_format": "json",
"original_prompt": "NVIDIA Unveils World's First Data Center Accelerated CPU - the Grace Hopper Superchip, deepleaps.com",
"active_tags": [
"Fantasy",
"Realistic",
"Photo",
"Surrealist",
"Unreal Engine"
],
"inactive_tags": [],
"use_upscale": "RealESRGAN_x4plus",
"upscale_amount": "4",
"prompt": "NVIDIA Unveils World's First Data Center Accelerated CPU - the Grace Hopper Superchip, deepleaps.com, Fantasy, Realistic, Photo, Surrealist, Unreal Engine",
"use_cpu": false
}