Microsoft has announced plans to provide its cloud computing customers with AI chips developed by AMD to meet the high demand for high-performance computing (HPC). Additionally, Microsoft will showcase its new custom Cobalt 100 processors, which offer a 40% performance improvement over other ARM-designed competitors. The company also introduced the Copilot+ PC during an opening speech, setting hardware requirements for systems to qualify as “AI-first” PCs. These PCs will be able to handle certain AI-accelerated tasks locally, without relying on cloud services.
To execute these tasks efficiently, Microsoft has partnered with Intel, AMD, and Qualcomm to build neural processing units (NPUs) that allow AI models to run locally on PCs. Microsoft has set specific hardware requirements for Qualcomm systems, requiring a minimum of 16GB of RAM and 256GB of SSD for systems to meet the memory and disk storage requirements for LLMs like Microsoft’s Phi-3. PCs with Snapdragon X Plus and X Elite chips will have Copilot+ features pre-installed and will begin shipping on June 18.
An important new requirement is the need for an integrated NPU with a performance rating of 40 trillion operations per second (TOPS). Current-generation chips from Intel and AMD do not meet this requirement, but both companies have future products in development to improve NPU performance. Microsoft’s Copilot+ PC requirement is currently only met by Qualcomm’s Snapdragon X Elite and X Plus chips, which feature an NPU with a capacity for 45 TOPS.
Intel has announced plans to release its Lunar Lake client processors in the third quarter of 2024, featuring NPUs with over 40 TOPS and GPUs with over 60 TOPS. AMD is also introducing its Ryzen 8050 APU with a performance of around 50 TOPS, expected to launch in the second half of 2024. NVIDIA has criticized Microsoft’s AI PC requirements, claiming its GPUs offer better performance than any NPU on the market, with GPUs reaching between 100 and 1300 TOPS.
In summary, Microsoft’s partnership with AMD to provide AI chips for cloud customers, the introduction of Copilot+ PCs with specific hardware requirements, and the development of NPUs by Intel, AMD, and Qualcomm signify a shift towards local AI processing on PCs. While current-generation chips do not meet Microsoft’s NPU performance requirements, upcoming releases from Intel, AMD, and Qualcomm are expected to address this gap and compete with NVIDIA’s GPU offerings in the AI PC market.
Article Source
https://analyticsindiamag.com/microsoft-shows-love-for-intel-and-amd-but-chooses-qualcomm-for-copilot-pcs/