Highlights

Pegatron has announced its participation in this year’s Paris GTC conference, where it will showcase its latest innovations in AI infrastructure. The company is unveiling a new range of next-generation GPU-accelerated server platforms featuring NVIDIA's Blackwell architecture, tailored for AI training and inference, along with diversified needs of data centers and enterprise deployments. This strategic advancement emphasizes Pegatron's commitment to blend high-performance and energy-efficient solutions.

Sentiment Analysis

• Pegatron's announcement has generated a positive reception, highlighting the integration of cutting-edge NVIDIA technology to meet growing AI demands.
• The sentiment is largely optimistic due to the focus on performance and energy efficiency.
• This positions Pegatron as a competitive player in the AI server market.

Article Text

Pegatron (4938-TW) has announced its participation at the upcoming Paris GTC conference, showcasing significant advancements in AI server technology. Utilizing the powerful NVIDIA Blackwell architecture, Pegatron plans to introduce a series of cutting-edge GPU-accelerated server platforms. These are specially developed to meet the evolving requirements for AI training, inference computation, and the diverse needs of expanding data centers and enterprise deployments.

According to Mei-Hui Wang, Vice President of Pegatron, the continuous transformation in the industrial landscape due to AI mandates the urgent need for efficient, scalable, and energy-efficient infrastructures. Pegatron's new platform combining NVIDIA Blackwell Ultra and RTX PRO architectures is meticulously designed for next-level data centers and AI enterprises, promising to enhance performance, stability, and productivity.

The exhibition will feature NVIDIA GB300 NVL72, HGX B300 liquid and air-cooled solutions, and RTX PRO servers. Pegatron explains that the NVIDIA GB300 NVL72 platform, integrating 72 NVIDIA Blackwell Ultra GPUs and 36 NVIDIA Grace CPUs, elevates inference performance by fifty-fold, significantly boosting AI production efficiency. The system comes equipped with Pegatron's proprietary cooling distribution unit (CDU) capable of handling 310kW cooling capacity to support high-performance computing demands.

In terms of NVIDIA HGX B300 platforms, both liquid and air cooling options are offered. This scalable system can expand up to a 48U NVIDIA MGX rack supporting 128 GPUs and 32 CPUs configurations with high-performance direct liquid cooling architecture. Additionally, for environments lacking liquid cooling infrastructure, an 8U air-cooled GPU server is available, accommodating smart agent systems, AI inference, and image generation workloads efficiently due to its foundation on the Blackwell Ultra architecture.

NVIDIA RTX PRO servers cater to enterprise-scale applications and AI factories, providing solutions for diverse accelerated workloads such as multimodal inference, physical AI, design engineering, scientific computing, and multimedia applications.

Key Insights Table