Intel To Sunset First-Gen Max Series GPU To Focus On Gaudi, Falcon Shores Chips

A source confirms to CRN key details of a report stating that Intel has started to sunset its first-gen Max Series GPU so that the semiconductor giant can focus on selling its Gaudi accelerator chips and developing the next-gen Falcon Shores GPU.

Intel is sunsetting its first-generation Max Series GPU, previously code-named Ponte Vecchio, to focus on selling its Gaudi AI accelerator chips and developing its next-gen Falcon Shores GPU.

This is according to a Tuesday report by industry publication Serve The Home and a source familiar with the matter who confirmed key details of the report to CRN.

[Related: Analysis: How Nvidia Surpassed Intel In Annual Revenue And Won The AI Crown]

An Intel spokesperson on Wednesday confirmed that the company is focusing on selling its Gaudi chips and developing Falcon Shores, which is expected to launch in 2025 and will combine Gaudi technology with the Xe GPU architecture that was used in Ponte Vecchio.

The Santa Clara, Calif.-based company is gearing up to launch Gaudi 3 later this year as a more powerful and efficient solution over Gaudi 2, which debuted in 2023.

“To meet the growing demand for enterprise AI, we are focused on fulfilling the rapid expansion of the Intel Gaudi AI accelerator—capitalizing on its proven performance edge and competitive pricing,” the representative said in a statement.

However, the Intel spokesperson didn’t directly address the status of the first-gen Max Series GPU, which debuted a little more than a year ago and had been hyped for several years as a central part of the company’s strategy to compete with Nvidia in the accelerator chip market.

Instead, the Intel representative referenced the availability of the Max GPU on Intel Developer Cloud, which was recently renamed to Intel Tiber Developer Cloud as part of a new brand push.

“Intel Xeon remains the host CPU choice for [high-performance computing] solutions along with our Max GPU available on Intel Developer Cloud,” the spokesperson said.

“This approach will pave the path for developer and ecosystem readiness for Falcon Shores, our next-generation GPU for AI and HPC which will leverage the Xe IP architecture that is foundational to the Max Series GPU,” the Intel representative added.

Serve The Home reported that Intel has stopped selling Ponte Vecchio into new server clusters and is only fulfilling orders for existing clusters, including for Intel Tiber Developer Cloud, its cloud platform for letting developers test and build AI solutions using Intel chips.

The Rocky Past, Present And Future Of Intel’s Data Center GPUs

Intel launched Ponte Vecchio as the Max Series GPU for data centers running HPC and AI workloads at the beginning of last year.

But it only arrived after roughly two years of delays, caused in part by issues with Intel’s 7-nanometer manufacturing process that emerged in 2020.

This, in turn, helped delay the delivery of the Aurora exascale supercomputer to the U.S. Department of Energy’s Argonne National Laboratory. The lab had chosen the Max Series GPU and Intel’s Xeon Max Series CPU, which comes with high-bandwidth memory baked into the chip, as the supercomputer’s engines after Intel cancelled Knights Hill, its high-end Xeon Phi accelerator that was meant for the original version of Aurora that was conceived in 2015.

After getting outfitted last year with more than 10,000 server blades of Max Series GPUs and CPUs from Intel, Aurora’s supercomputer managed to recently surpass 1 exaflop of performance for the high-performance LINPACK benchmark, as confirmed this week by Top500, the independent organization that tracks the world’s fastest supercomputers.

While Intel originally planned to start sampling a GPU successor to Ponte Vecchio called Rialto Bridge with customers in mid-2023, the company announced in March of that year that it had decided to cancel Rialto Bridge to focus on Falcon Shores for a 2025 launch.

The semiconductor giant initially planned a 2024 release window for Falcon Shores, which was conceived as a new “flexible chiplet-based architecture” that would include versions with CPU cores in addition to GPU cores.

The chipmaker announced further changes to its plan for Falcon Shores two months later, in May of 2023, saying that it no longer planned a variant with CPU cores and that it now planned to incorporate technology from its Gaudi accelerator chips into the Falcon Shores GPU.

Intel said it was doing this as part of a decision to converge the Gaudi chip and Max Series GPU road maps. This means Falcon Shores will serve as the successor to both Ponte Vecchio and Gaudi 3. The company has not yet said what brand it will use for Falcon Shores when it arrives.

All Eyes On Gaudi 3—For Now

At the Intel Vision event in March, the semiconductor giant hyped up the upcoming arrival of Gaudi 3 and said it can best Nvidia’s powerful H100 GPU for training large language models and offer similar or, in some cases, better performance than the rival’s memory-rich H200 for large language model inferencing. The chip is set to debut in the third quarter, with support planned from Dell Technologies, Hewlett Packard Enterprise, Lenovo and Supermicro.

During its latest earnings call in April, Intel CEO Pat Gelsinger said the company expects to make more than $500 million in revenue from Gaudi 3 sales in the second half of the year. Two months earlier, Gelsinger said that the company had built a sales pipeline of more than $2 billion for its accelerator chips, which included Gaudi 2 and 3 as well as the Max and Flex Series GPUs.

These figures pale in comparison to the $47.5 billion in revenue Nvidia generated for its data center business last year, largely thanks to fast growth of powerful GPUs like the H100. AMD, on the other hand, expects to make $4 billion from data center GPUs this year, thanks to its recently launched Instinct MI300 accelerator chips, which CEO Lisa Su recently called its “fastest-ramping product” ever.

An executive at a system integrator that partners with all three chip companies told CRN that Intel’s challenges in the accelerator chip market speak to Nvidia’s dominance in the space.

“One of the hills to climb there is the amount of development work that Nvidia has been able to do over the years with CUDA,” the company’s parallel computing platform and programming model that has made its GPUs popular with developers, said Dominic Daninger, vice president of engineering at Burnsville, Minn.-based Nor-Tech.

While Intel could fight through the noise with its Gaudi chips, Daninger said, he hasn’t seen any demand from customers yet for the company’s accelerator chips.

“We're just not seeing much interest,” he said.