➀ Dr. Tahir Ghani discussed the impact of Moore's Law and the challenges of energy efficiency in semiconductor technology. ➁ He highlighted the evolution of transistor technology over six decades. ➂ He advocated for new transistor designs and collaborations to improve energy efficiency for AI computing.
Recent #AI Computing news in the semiconductor industry
➀ Elon Musk's xAI has built the Colossus AI supercomputer in just 122 days; ➁ The system is powered by 100,000 NVIDIA H100 AI GPUs; ➂ Supermicro liquid-cooled racks, each containing 64 GPUs, form mini clusters.
1. AMD launches Ryzen AI 300 series 'Strix Point' APU laptops with Zen 5 CPU, RDNA 3.5 GPU, and XDNA 2 NPU. 2. The new APUs offer flexible TDP settings ranging from 15W to 54W. 3. AMD's XDNA 2 NPU provides 50 TOPS, outperforming competitors in the AI PC market.
1. GIGABYTE introduces Radeon PRO W7900 Dual Slot AI TOP 48G and W7800 AI TOP graphics cards with 48GB and 32GB of VRAM respectively. 2. The W7900 Dual Slot AI TOP 48G features a Navi 31 GPU with 6144 shaders and supports ECC on a 384-bit memory bus. 3. The card supports up to 4x 4K 120Hz displays or dual 6K 60Hz displays, and is designed for AI computing with a 295W TBP.
➀ DeepSeek has partnered with numerous companies across various sectors, including chip enterprises, cloud service and AI computing companies, AI infrastructure enterprises, and application end enterprises.
➁ The chip enterprises include SiBase Flow, Huawei Ascend, Wallenius Technology, Muxi, TianShu Zhixin, etc., which are working on new adaptive AI chip solutions for edge computing scenarios.
➂ Cloud service and AI computing enterprises such as Huawei Cloud, Tianyi Cloud, Tencent Cloud, Alibaba Cloud, Baidu AI Cloud, and JD Cloud have integrated DeepSeek models into their platforms.
➃ AI infrastructure and application end enterprises like SiBase Flow, Wuwen Xingqiong, Lenovo, DaoTong Technology, and Sunyuan Technology are also collaborating with DeepSeek for various applications and optimizations.
➀ The rise of ASICs in the capital market is challenging the dominance of GPUs in AI computing;
➁ ASICs and GPUs are both semiconductor chips used for computing, with ASICs being dedicated to specific tasks;
➂ The performance and efficiency of ASICs are highly matched to the task algorithms, making them more reliable and energy-efficient than general-purpose chips like GPUs;
➃ NVIDIA's GPUs have a strong market position due to their performance, ecosystem, and integration capabilities, but the rise of ASICs presents opportunities for diversification in AI computing.
➀ HBM market is rapidly expanding due to its critical role in AI computing, with SK Hynix leading in supply and innovation. ➁ SK Hynix is collaborating with TSMC to maintain its lead in the next-generation HBM products, which are expected to see a 40% increase in transmission speed. ➂ Samsung is increasing its investment in HBM to catch up, with plans to quadruple its supply in 2024 and double it again in 2025.