Intel announced support for Microsoft's Phi-3 AI models across its entire range of CPU and GPU product portfolio.
Before we discuss the details, let's examine Microsoft's Phi-3 family of open models. The firm's new GenAI models focus on providing capabilities on a smaller and more efficient level, deviating from industry norms. With Phi-3, we could see the emergence of Small Language Models (SLMs), ideal for low-power devices with limited processing power.
Fortunately, Intel has become one of the first hardware manufacturers to bring in support for the SLM within its product lineup, and now Intel's Gaudi AI accelerators, Xeon and Core Ultra CPUs, and Intel's Arc GPUs (discrete and integrated) are optimized to support Phi-3 models from Microsoft.
Microsoft has released multiple Phi-3 GenAI models, categorizing them within their respective parameters. These include mini, small, and medium models; interestingly, Intel has showcased the performance of its products in all open models. Starting with Intel's 6th Gen Xeon CPUs (Granite Rapids), the firm benchmarked Phi-3 medium 4k and 128k variants, and the platform offered 2X more performance compared to 5th Gen Xeon (Emerald Rapids) counterparts.
Moving on to Intel's Core Ultra CPU platform, the firm says that Phi-3 models will catalyze the edge AI process and will offer several advantages such as "personalization and responsiveness." With that, the company tested the Phi-3 mini model with Intel's Core Ultra 7 165H, with the Arc A770 as the discrete GPU onboard, and the benchmarks did show impressive performance and token generation latencies.
The debut of Microsoft's Phi-3 AI models will indeed open new doorways in the era of personalized AI computing, and with the performance modern-day processors hold,
Read more on wccftech.com