Intel buries news of GPU cuts and delays in low-key Friday post
Falcon Shore will be late. Rialto Bridge and Lancaster Sound cancelled. Will Chipzilla ever nail its XPU plan?
Intel has used a quiet Friday announcement to reveal delays and deletions to its GPU and high-performance computing roadmap.
Intel's document opens with cheery chat about GPUs. But well below the fold are several humbling pieces of news for Chipzilla. The first two are the dumping of the Rialto Bridge generation of GPUs and postponing the Falcon Shores GPU architecture – to 2025.
As The Register reported in May 2022, when Intel offered an update to its HPC roadmap at the International Supercomputing Conference, Intel planned to start sampling Rialto Bridge in mid-2023. At the time, Rialto Bridge was billed as a worthy successor to Ponte Vecchio GPU, and capable of delivering a 30 percent performance boost. Falcon Shores – a platform with x86 CPU cores and Xe GPU cores in a single package – was billed as an "XPU" and mentioned as likely to debut in 2024.
Intel's latest announcement reframes Rialto Bridge as a source of mere "incremental improvements" – which has therefore been discontinued.
Falcon Shores, meanwhile, is now "Targeted for introduction in 2025." Intel argued that rescheduling – and a move to a two-year release cadence for the HPC-grade "Max" GPU range – "matches customer expectations on new product introductions and allows time to develop their ecosystems."
The post also mentions discontinuation of the datacenter-grade Lancaster Sound GPU architecture used in the current "Flex" GPU range – again on grounds it offered only incremental performance increases.
Doing so "allows us to accelerate development on Melville Sound, which will be a significant architectural leap from the current generation in terms of performance, features and the workloads it will enable."
No date for Melville Sound's debut is offered, but the post implies that the changes outlined will result in a two-yearly release cadence for the Flex range.
Intel's news isn't all glum on the GPU front: the chip shop has promised "continuous updates for our Max Series and Flex Series products, with performance improvements, new features, expanded operating systems support and new use cases to broaden the benefits of these products."
Detail is, again, not available. And with Intel's business-focused GPU roadmap having been elongated, customers will probably need to wait at least a year before they get to see if "continuous delivery" offers anything worthy.
Assuming continuous delivery even works for these products, given Intel's remarks hat HPC customers need longer lead times to prepare their environments for this sort of hardware.
There's no way to sugarcoat Intel's status as a GPU laggard: Ponte Vecchio was years late and the Arc range of standalone consumer GPUs debuted only last year.
Nvidia had standalone product in the market for both punters and pros for more than a decade, while Intel focused on integrated graphics.
Intel started a long way behind on GPUs, and will be even further behind when – or if – Falcon Shores debuts in 2024.
And if the troubled giant manages to ship the architecture in that year, it will still have missed deadlines it set itself – adding to a long list of products that ended up arriving late and in configurations that rivals had already surpassed.
- Years late and 36 cores short of AMD, who are Intel’s 4th-gen Xeons even for?
- Intel cuts some workers’ pay to fund its future
- Intel inside a world of pain as revenue plunges by a third
- Intel kills $700M liquid cooling lab amid chip slump
Intel CEO Pat Gelsinger has been in the job for a little over two years. One of the first strategies he personally announced was "Integrated Device Manufacturing 2.0" (IDM 2.0) – a plan to have foundries other than Intel itself make some components of its products, with Chipzilla doing final integration and packaging.
Falcon Shores will, the once-dominant giant promised, be a product of IDM 2.0 at work. By the time it debuts, Gelsinger will have had at least four years in the job (fingers crossed) – making the GPU very much a product he'll have to own.
Hardly anyone, however, seems to have wanted to own Intel's GPU roadmap update. While Chipzilla's previous HPC GPU roadmap was delivered at a big international gabfest, this update was squeezed out without fanfare on a Friday – a time when even members of the press may not be at their sharpest as minds turn to family and leisure pursuits.
Political journalists have a phrase for that sort of behavior: they call it "taking out the trash." If Intel can't deliver on this roadmap, it will be in danger of delivering a steaming pile of irrelevance. ®
For more analysis and commentary on Intel's roadmap changes, don't miss Timothy Prickett Morgan in The Next Platform.