intel The company on Tuesday announced its latest artificial intelligence chip, called Gaudi 3, as chipmakers rush to produce semiconductors that can train and deploy large-scale AI models like the ones that power OpenAI’s ChatGPT.
Intel says the new Gaudi 3 chips are more than twice as power efficient and can run AI models 1.5 times faster. Nvidia’s H100GPU. It also offers a variety of configurations, including bundles of his eight Gaudi 3 chips on one motherboard and cards that can be slotted into existing systems.
Intel tested the chip on models like meta Falcon, backed by open source Rama and Abu Dhabi. Gaudi 3 is said to be useful for training and deploying models such as Stable Diffusion for speech recognition and OpenAI’s Whisper model.
Intel says its chips use less power than Nvidia’s.
Nvidia has Estimation Eighty percent of the AI chip market is made up of graphics processors, known as GPUs, which have been the high-end chips of choice for AI builders over the past year.
Intel said its new Gaudi 3 chips will be available to customers in the third quarter. Dell, hewlett packard enterpriseand super micro Build a system using chips. Intel did not reveal the price range for Gaudi 3.
“We expect it to be very competitive” with Nvidia’s latest chips, Das Kamhout, Intel’s vice president of Xeon software, said on a call with reporters. “With competitive pricing, a distinctive open integrated network-on-a-chip, we use industry standard Ethernet, and we believe this is a powerful product.”
The data center AI market is also expected to grow as cloud providers and enterprises build infrastructure to deploy AI software, and even if Nvidia continues to manufacture the majority of AI chips, other competitors This suggests that there is room for companies to enter the market.
Running generative AI and purchasing Nvidia GPUs can be expensive, and companies are looking for additional suppliers to help reduce costs.
The AI boom has caused Nvidia’s stock price to more than triple in the past year. Intel’s stock price has risen only 18% over the same period.
AMD is also aiming to expand and sell AI chips for servers. Last year, the company introduced a new data center GPU called his MI300X and already counts Meta and Microsoft as customers.
Earlier this year, Nvidia announced its B100 and B200 GPUs. These are the successors to his H100 and also promise improved performance. These chips are expected to begin shipping later this year.
Nvidia owes its success to a powerful proprietary software suite called CUDA that gives AI scientists access to all of the GPU’s hardware capabilities. Intel partners with other chip and software giants, including: Google, Qualcomm and arm It’s about building open software that allows software companies to easily switch chip providers, rather than proprietary.
“We are working with the software ecosystem to build open reference software and building blocks that allow people to piece together the solutions they need, rather than being forced to buy them. ,” said Sachin Katti, senior vice president of Intel Networking. the group said in a call with reporters.
Gaudi 3 is built on a 5-nanometer process, a relatively recent manufacturing technology, suggesting the company is using an external foundry to manufacture the chip. In addition to designing Gaudi 3, Intel also plans to manufacture AI chips, potentially for outside companies, at a new factory in Ohio expected to open in 2027 or 2028, CEO Patrick Gelsinger said. told reporters last month.