Intel Reveals “Jaguar Shores” As Its Next-Gen AI Chips For AI Inference At SC24

Intel has never talked about Jaguar Shores before, and it will probably bring together both Gaudi ASICs and Xe-HPC GPUs.

Intel Jaguar Shores is a next-gen AI GPU, aimed at faster AI inference and will probably utilize the in-house 18A node

Intel may not look in a good position right now, but it is continuing its plans for the client and server platforms. The company has been working on the next-gen Intel Battlemage GPUs, which has been leaked recently, but Battlemage isn’t the only GPU it is planning to launch shortly.

As per the report from HPC Wire, the company has revealed its plans for next-gen AI chips. The company usually brings high-performance chips on the Intel Gaudi platform from Intel’s Habana Labs division and while Gaudi 3 will be shipping in volume next year, the latest addition to its lineup won’t be just limited to Gaudi. Intel revealed that it is working on a new high-performance chip for powerful computing, known by the code name “Jaguar Shores”.

This was revealed at the SC2024 conference where Intel shared its roadmap comprising its high-performance chips for AI workloads. The Jaguar Shores is the successor to the Falcon Shores, the latter being planned for a 2025 release. Similar to Falcon Shores, the Jaguar Shores is likely to be a GPU rather than a processor for the Intel Gaudi Accelerator.

Falcon Shores was initially designed to be a hybrid GPU-CPU solution, but now it is going to be a dedicated GPU for AI inference. The Jaguar Shores is said to be in a similar situation but might bring a GPU-ASIC design for delivering enhanced AI Inferencing. With Intel keeping its eye on the next-gen 18A manufacturing node, Intel will likely go with in-house 18A for both Falcon Shores and Jaguar Shores to leverage the power of technologies like RibbonFET and backside power delivery for improved efficiency.

Our AI investments will complement and leverage our x86 franchise – with a focus on enterprise, cost-efficient inferencing. Our roadmap for Falcon Shores remains,

– Intel Spokesperson

Intel already outsources some of its chips from TSMC like Lunar and Arrow Lake, but this time it wants to have an advantage with the 18A, which will also be used in Clearwater Forest next year. A few years later, Intel also plans to utilize the in-house 14A process node in 2027.

Intel Jaguar Shores will go against giants like NVIDIA and AMD, which already have a consistent roadmap of high-performance GPUs through 2027. Both companies dominate the AI training and inference markets, which is why Intel is targeting mostly the inference market with the upcoming chips.

We continuously evaluate our roadmap to ensure it aligns with the evolving needs of our customers. While we don’t have any new updates to share, we are committed to providing superior enterprise AI solutions across our CPU and accelerator/GPU portfolio.

– Intel Spokesperson

Intel has been struggling with its roadmap, which led to the cancellation of the Rialto Bridge GPU last year. The lack of foresight can be seen with the Falcon Shores as well, which transitioned from CPU+GPU to only discrete GPU. Furthermore, Intel also canceled the Ponte Vecchio GPU after the supercomputer Aurora went online.

Intel has to carefully strategize and execute its plans amid the leadership turmoil, which saw frequent changes in the GPU leadership and organization in general. It has already slowed down drastically in the client CPU market, and it can’t afford to lag in the AI market as well.

News Source: HPCWire