Intel Tuesday launched its second-generation Habana AI deep learning processors, which deliver high performance and efficiency, at its Vision 2022 event.
The new chips, which use 7-nanometer technology, are the Habana Gaudi2 and Habana Greco. Intel paid $2 billion for Habana Labs in 2019.
The introduction of Habana’s new deep learning processors is a prime example of Intel executing its AI strategy to provide customers with a wide range of solution options—from cloud to edge—addressing the growing number and complexity of AI workloads, said Sandra Rivera, Intel’s executive vice president and general manager of the data center and AI group, in a statement.
Because of complex datasets and AI workloads, training deep learning models is an expensive and time-consuming endeavor for data center professionals. According to the company, Gaudi2 was created to improve deep learning performance and efficiency in cloud and on-premises systems.
According to IDC, 74% of machine learning practitioners polled in 2020 run five to ten training iterations of their models, more than 50% rebuild models weekly or more frequently, and 26% rebuild models daily or even hourly. The cost was cited as the most significant barrier to their businesses utilizing the improvements that AI can provide in handling those workloads.
When compared to the A100 GPU, which is implemented in the same process node and roughly the same die size, Gaudi2 delivers clear leadership training performance as demonstrated by apples-to-apples comparison on key workloads, said Eitan Medina, COO of Habana Labs. This deep-learning acceleration architecture is fundamentally more efficient and is supported by a solid road map.
Intel’s partners praised Habana Labs’ progress.
We’re excited to bring to market our next-generation AI deep learning server with the high-performance 7 nm Gaudi2 processor, which will enable our customers to achieve faster time-to-train advantages while maintaining competency and expanding on the scalability of first-generation Gaudi, said Charles Liang, CEO of Supermicro.
Channel partners were also present to test the new processing power. In an interview with CRN, Accenture’s Jason Mitchell, managing director, and Intel client account lead, stated, “Understanding how those technologies are brought to bear is important to our applied intelligence, artificial intelligence, and analytics.” This is a significant investment area. So, we’re excited to see what’s next and how their software platform acts as a bridge between their silicon and our software. It’s always exciting to see what’s next in the underlying hardware.
