NVIDIA to Launch New Data-Center GPU for China in Response to US Restrictions

The H20 GPU stands out with an impressive 96GB of HBM3 memory and a remarkable memory bandwidth of 4.0 TB/s, notably surpassing the 'global' H100, which boasts 3.6 TB/s.

Introduction:

In the ever-evolving landscape of GPU technology, regulatory changes can redefine the game. NVIDIA, a stalwart in innovation, has recently showcased its adaptability with the introduction of the HGX H20 and L20 series for China.

Crafted in response to updated U.S. restrictions on high-performance GPUs, these cutting-edge solutions not only ensure compliance but also position NVIDIA as a pioneer in addressing the challenges posed by evolving regulations.

Join Our WhatsApp News

The Regulatory Landscape for Nvidia GPU in China:

The recent shift in U.S. restrictions puts a spotlight on compute power over interconnect communications, affecting not only data-center GPUs but also gaming cards, including the highly anticipated RTX 4090.

NVIDIA’s strategic response to this regulatory shift reflects the company’s proactive stance, navigating complexities to maintain technological innovation.

Read More: What are the Key Regulatory Changes by US in the Expanded Ban to China?

NVIDIA Adaptive GPU for China Approach :

In the face of regulatory changes, NVIDIA sees opportunities for innovation. The HGX H20 and L20 series epitomize this adaptive approach, showcasing NVIDIA’s commitment to meeting market demands while aligning with regulatory requirements.

This strategic shift reinforces NVIDIA’s position as a dynamic industry leader capable of navigating complex regulatory landscapes without compromising technological innovation.

Read More: What is the Progress of Semiconductor Ecosystem in India in Last 2 years?

HGX H20: Redefining Data-Center GPU Clusters:

The star of NVIDIA’s response is the HGX H20, a data-center GPU optimized for clusters of 8 GPUs. With NVLink connectivity and a demand for 400W of power, the H20 sets a new standard for data-center GPU clusters, ensuring seamless data exchange within high-performance computing environments.

Read More: How Huawei Could be a Winner in US Ban of Nvidia GPUs?

L20 and L2 GPUs: Meeting Diverse Memory Needs:

Complementing the HGX H20 are the L20 and L2 GPUs, operating on PCIe interfaces. These GPUs intentionally limit compute capabilities to adhere to the newly imposed restrictions. With memory options of 48GB and 24GB, respectively, the L20 and L2 cater to diverse applications, addressing the unique memory needs of data-center tasks and gaming experiences.

Technological Advancements for Nvidia GPU in China:

NVIDIA has officially launched the HGX H20 and L20 series for Chinese customers. The H20 GPU intentionally reduces compute power compared to its predecessors. It stands out with 96GB of HBM3 memory and an impressive 4.0 TB/s memory bandwidth, surpassing the global H100’s 3.6 TB/s.

Conversely, the L20 GPU, built on the AD102 GPU architecture, features a configuration with 48GB of GDDR6 memory. These solutions intentionally limit compute power to keep GPUs below the 4800 TOPS performance threshold. This strategic decision aligns with specific performance constraints, ensuring optimal adherence to regulations.

Sampling and Order Commencement:

NVIDIA’s strategic move includes chip sampling to partners, generating anticipation in the market. Orders for the HGX H20 and L20 series start on November 16th, strategically coinciding with the full implementation of U.S. restrictions. The projected first shipment in December underlines NVIDIA’s commitment to timely delivery despite regulatory challenges.

Here’s a table comparing the key specifications of the new NVIDIA GPUs: HGX H20, L20, and L2.

SpecificationHGX H20L20L2
Compute PowerOptimized for clusters of 8 GPUsLimited compute capabilitiesLimited compute capabilities
ConnectivityNVLinkPCIePCIe
Power Demand400WVariesVaries
Memory96GB HBM348GB GDDR624GB GDDR6
Memory Bandwidth4.0 TB/sVariesVaries
Sampling InitiatedYesYesYes
Order CommencementNovember 16thNovember 16thNovember 16th
Projected ShipmentDecemberDecemberDecember

Read More: China: Too large, Too important & Too strategic to walk away from

Conclusion:

NVIDIA’s HGX H20 and L20 series reveal a strategic pivot, showcasing resilience and innovation. This places NVIDIA as a frontrunner in the industry, committed to pushing technological boundaries.

Their forward-thinking approach ensures they stay a driving force in shaping the future of GPU technology, even amid regulatory complexities.

Kumar Priyadarshi
Kumar Priyadarshi

Kumar Priyadarshi is a prominent figure in the world of technology and semiconductors. With a deep passion for innovation and a keen understanding of the intricacies of the semiconductor industry, Kumar has established himself as a thought leader and expert in the field. He is the founder of Techovedas, India’s first semiconductor and AI tech media company, where he shares insights, analysis, and trends related to the semiconductor and AI industries.

Kumar Joined IISER Pune after qualifying IIT-JEE in 2012. In his 5th year, he travelled to Singapore for his master’s thesis which yielded a Research Paper in ACS Nano. Kumar Joined Global Foundries as a process Engineer in Singapore working at 40 nm Process node. He couldn’t find joy working in the fab and moved to India. Working as a scientist at IIT Bombay as Senior Scientist, Kumar Led the team which built India’s 1st Memory Chip with Semiconductor Lab (SCL)

Articles: 2224