Intel Pushes "Clearwater Forest" Xeon CPU Series Launch into 2026

[T0@st]

Intel has officially announced that its "Clearwater Forest" Xeon processor family will be arriving somewhere in the first half of 2026. During a recent earnings call, interim co-CEO—Michelle Johnston Holthaus—discussed Team Blue's product roadmap for 2025 and beyond: "this year is all about improving Intel Xeon's competitive position as we fight harder to close the gap to the competition. The ramp of Granite Rapids has been a good first step. We are also making good progress on Clearwater Forest, our first Intel 18A server product that we plan to launch in the first half of next year." Press outlets have (correctly) pointed out that Intel's "Clearwater Forest" Xeon processors were originally slated for release in 2025, so the company's executive branch has seemingly admitted—in a low-key manner—that their next-gen series is delayed. Industry whispers from last autumn posit that Team Blue foundries were struggling with their proprietary 18A (1.8 nm) node process—at the time, watchdogs predicted a postponement of "Clearwater Forest" server processors.

The original timetable had "Clearwater Forest" server CPUs arriving not long after the launch of Intel's latest line of "Sierra Forest" products—288-core models from the Xeon 6-series. The delay into 2026 could be beneficial—The Register proposes that "Xeons bristling with E-cores" have not found a large enough audience. Holthaus disclosed a similar sentiment (in the Q4 earnings call): "what we've seen is that's more of a niche market, and we haven't seen volume materialize there as fast as we expected." Despite rumors swirling around complications affecting chip manufacturing volumes, Intel's temporary co-leaders believe that things are going well. David Zinsner—Team Blue's CFO—stated: "18A has been an area of good progress...Like any new process, there have been ups and downs along the way, but overall, we are confident that we are delivering a competitive process." His colleague added: "as the first volume customer of Intel 18A, I see the progress that Intel Foundry is making on performance and yield, and I look forward to being in production in the second half, as we demonstrate the benefits of our world-class design."



View at TechPowerUp Main Site | Source

 

[Daven]

When you try to fool customers that you have a competitive number of enterprise level cores but really they are gimped E-cores barely suited for desktop tasks, don’t be surprised when you get called out.

That and the fact that AMD came upon a surprising good solution with its dense cores.

 

[efikkan]

A very niche product approaching being a gimmick.

I would much rather they made an "XL" core; basically doubling the execution units and running at lower clocks, for server and workstation. For computationally dense server workloads this wouldn't require a massive extension of the CPU front-end to extract good performance gains.

 

[dyonoctis]

When you try to fool customers that you have a competitive number of enterprise level cores but really they are gimped E-cores barely suited for desktop tasks, don’t be surprised when you get called out.

That and the fact that AMD came upon a surprising good solution with its dense cores.

While Skymont's IPC is lower than Zen 5's, we are still talking about RPL level of IPC. Darkmont, used here, is apparently Skymont on steroids with support for AVX. But customers are probably a bit wary of that first gen product and want to see how it performs first. Although I would have thought that concerned customers had the opportunity to get a decent preview of those chips

 

[Daven]

While Skymont's IPC is lower than Zen 5's, we are still talking about RPL level of IPC. Darkmont, used here, is apparently Skymont on steroids with support for AVX. But customers are probably a bit wary of that first gen product and want to see how it performs first. Although I would have thought that concerned customers had the opportunity to get a decent preview of those chips

Xeon 6780E 144 E cores (144 threads) 330W for $11,350

or

Xeon 6980P 128 P cores (256 threads) 500W for $13,000

Which would you pick?

 

[dyonoctis]

Xeon 6780E 144 E cores (144 threads) 330W for $11,350

or

Xeon 6980P 128 P cores (256 threads) 500W for $13,000

Which would you pick?

I'm pretty sure the 330w matter more than the absolute performance figures for the market they target. There's also Ampere, AWS and ARM Neoverse who aren't exactly competing for top HPC performance either. But Ampere offers low-power 192 cores CPUs at half the cost of AMD and Intel. Bergamo and the XEON e-core were made because there was a demand for low power datacenters that would be cheaper to run in applications that don't need High performance.

The classic Epyc was already so far ahead of Intel when it comes to efficiency, that there wasn't a need to make a "cut down'" zen core if there wasn't a demand for an even more efficiency-focused CPU. That's the angle that ARM for datacenter has been focused on : fast enough for things like the clouds while using less power than X86 based solution, thus offering a better ROI.

From Phoronix :

The Xeon 6700E processors are targeted for cloud native applications, databases, micro-servers, some CI/CD deployments, and related tasks where E cores can compete sufficiently. For those more interested in HPC and AI workloads, it will be very interesting to see how the Intel Xeon 6900P Granite Rapids processors are competing in performance and power efficiency against AMD's current and upcoming EPYC processors

AMD EPYC Bergamo is an extremely promising new option for cloud service providers or those wanting dense container/VM deployments while interested in reducing operational costs with leading power efficiency