Intel’s latest public server GPU product roadmap, released back during SC22 in May 2022, outlined that Intel intended to release server GPU products at a fairly rapid clip. This was (in part) to make up for lost time from the many delays involving Ponte Vecchio, Intel’s first HPC-class GPU now sold under the Data Center GPU Max family. At that event, Intel announced the Ponte Vecchio’s successor, the Rialto Bridge, which was to be an evolution of the Ponte and to be attempted in mid-2023. Following Rialto would be Falcon Shores in 2024, which, along with using an even newer version of Intel’s Xe GPU architecture, would be Intel’s first combined CPU+GPU (XPU) product and should be the ultimate evolution of both Intel’s HPC CPU and GPU lines.
However, citing “a goal to maximize return on investment for customers”, Intel has significantly changed their server GPU roadmap and canceled their previously planned 2023/2024 products. The Rialto Bridge, announced less than a year ago, has been scrapped. Instead, Intel will go straight to Falcon Shores for their HPC GPU needs.
These cancellations also affect Intel’s Flex line of server GPUs for cloud gaming and media encoding, with Intel ditching Lancaster Sound (aka “Next Sound”) in favor of moving on to its successor, Melville Sound. Melville Sound’s development, on the other hand, will be accelerated, although Intel has not attached a hard date to it. Previously, it was expected at roughly the same time as Falcon Shores.
Intel didn’t provide a new image for the new roadmap, but I’ve gone ahead and put together something based on their SC22 slide to show where things stand and what products have been canceled
According to Intel, these product changes are designed to allow the company to adjust to a two-year product release cadence for server GPUs. Competitors NVIDIA and AMD have operated at a similar cadence for the past several years, so this would bring Intel’s product cycles roughly in line with the competition. Which, as Intel puts it, “matches customer expectations for new product introductions and allows time to develop their ecosystems.”
Reading between the lines, the implication is that Intel wasn’t confident in their ability to sell the Rialto and Lancaster to their core customer base. If it is exclusively due to the rapid product release cycles, or if there was something more to it, is a detail that will remain inside Intel’s halls for now. But regardless, the company is essentially scrapping plans to release new server GPU products this year to focus on their successors later.
And if there’s any doubt as to whether this is a good or bad development for Intel, I’ll note that this information was released after 17 Eastern on a Friday. So Intel is very clearly trying to bury bad news here by releasing it at the end of the week.
Falcon Shores will go from an XPU in 2024 to a GPU in 2025
Besides canceling the Rialto Bridge and Lancaster Sound, the other major update to come from Intel’s letter is the revelation that Intel has significantly changed their plans for Falcon Shore. What was (and perhaps still is) Intel’s first combined CPU+GPU product for HPC has now been retooled to serve as Intel’s next generation HPC GPU. Which in turn has significant consequences for Intel’s product range and their competitive positioning.
Falcon Shores was another 2022 Intel announcement that was revealed back at the company’s 2022 investor meeting in February. At a high level, Falcon Shores was designed to be Intel’s first XPU product – a chip that uses a variety of computing architectures to best meet the execution needs of a single workload. This would be achieved by using separate processor chips, and while in practice this primarily meant putting Xe GPU chips and x86 CPU chips together on a single chip, the chip approach left the door open to adding other types of accelerators as well. Intel already makes tiled server chips with products like Ponte Vecchio and the recently launched Sapphire Rapids XCC, so Falcon Shores should be the next step in putting more than just one type of accelerator on a chip.
At the time, Falcon Shores was slated for release in 2024, to be built on an “angstrom-era process”. As planned, Falcon Shores would have been Intel’s power play for the HPC industry, allowing them to deliver a combined GPU+CPU product based on their latest architectures and built on their cutting-edge 20A/18A process node, giving them an advantage in HPC market, putting them ahead of traditional GPU-based accelerators. And while Falcon Shores may yet meet those goals, it won’t in 2024. Or even 2025, for that matter.
Instead, Falcon Shores has been reconfigured to act as an HPC GPU part. Intel’s brief does not go into the technical changes, but the implications are that instead of mixing CPU and GPU chips on a single chip, Intel will devote itself to building a product from only GPU chips, reducing , what should be. intel’s XPU to a more straightforward GPU.
With the cancellation of the Rialto Bridge, it goes without saying that Intel needs a new HPC-class GPU product if they want to stay in the market. And while there’s currently no reason to believe that Falcon Shores won’t be a good product, the fact that Intel has delayed it until 2025 isn’t a promising sign. Ponte Vecchio is already an older design than its age would allow – it was originally intended to launch in 2021 and compete with the HPC GPUs of that era – so by not having anything newer to offer until 2025, Intel is breaking away in the essential HPC GPU market in the next two years, outside of their supercomputer wins. Intel’s long-term plans still call for the company to take a significant slice of the highly profitable server GPU market — to push for a piece of what has largely been NVIDIA’s pie — so this development essentially puts those plans on hold in another two years.
And what could turn out to be worse, this puts Intel even further behind its competitors in shipping an XPU/APU-like product. As confirmed by Serve The Home, while Intel hasn’t given up on shipping a Falcon Shores XPU, the prioritization of GPUs means that such a product wouldn’t launch in 2025. This means that an Intel server-class XPU is now a 2026 product at best – and even longer at worst.
Meanwhile, AMD will ship a similar server APU later this year with the Instinct MI300. That part will use multiple chiplets/tiles on a single chip to offer HPC-class CPU and GPU performance on a single chip, achieving a long-awaited goal for AMD. And while NVIDIA is a bit further behind in integration with their Grace Hopper superchip – it’s essentially two tightly coupled chips on a single board, rather than being chiplets on a single chip – it’s still ahead of where Intel is today . And worryingly for Intel, the next-gen successor to that part is almost certain to launch before a potential Falcon Shores XPU hits the market.
In other words, by pushing back their server GPU schedule, all of Intel’s products that depended on it are now correspondingly delayed. That leaves Intel’s server CPUs to hold the line alone for the next few years.
A history of server GPU cancellations as Intel keeps trying
Ultimately, the cancellation of the Rialto Bridge makes it the latest in what has become a surprisingly long line of canceled GPUs and GPU-like accelerators at Intel. The company has been trying to break into the accelerator market in one form or another since the late 00s with projects like Larrabee. And now, 15 years later, Intel still isn’t as far as it wants to be, and is still canceling chips to get there.
As the latest in Intel server GPU damage, the Rialto Bridge joins not only Larrabee, but also Intel’s ill-fated Xeon Phi lineup. And even then, the Rialto Bridge can’t claim to be the first Xe architecture part to be preserved—that honor goes to the Xe-HP, which was canceled in favor of using Intel’s mixed computing and graphics Xe-HPG silicon.
At least the advantage of this situation is that Intel is not canceling an entire architecture here. The Xe architecture has firmly established itself as Intel’s internal GPU architecture, and the fact that it ships in everything from SoCs to video cards to HPC accelerators underscores its viability and importance at Intel. The cancellation of the Rialto Bridge combined with the delay of Falcon Shores is still a significant setback for Intel, but at the end of the day it’s just shelving one iteration of the broader Xe architecture for another.
In the meantime, Intel will continue to work on expanding their GPU offerings and to break into the HPC GPU market in a bigger way. It’s a market that’s simply too big and too profitable for Intel to ignore—at the very least, it’s eating into their server CPU sales even more—so they’ll have to keep at it until they find the success they need. they search.
Source: Intel