Intel Nova Lake-S chips complete tape-out, could be company's most ambitious desktop CPU yet

[Skye Jacobs]

Rumor mill: Intel may be edging closer to unveiling what could be its most powerful and innovative desktop CPUs yet. A new report from SemiAccurate indicates that the company's upcoming Nova Lake-S processors have completed the tape-out phase at TSMC's fabrication plants in Taiwan. This key step marks the transition from design to the first real silicon chips made for internal testing.

For years, speculation surrounded which manufacturing process Intel would use for Nova Lake-S, with many expecting a debut for the company's in-house 18A node. The latest developments, however, reveal that at least part of Nova Lake-S will rely on TSMC's cutting-edge N2 process, supporting the theory that Intel is embracing a hybrid foundry strategy.

This approach could be a response to concerns about production yields and readiness with the 18A node or a pre-emptive measure to ensure enough chips to meet demand by tapping proven external capacity.

The tape-out phase is far from the end of the road. Next, Intel must validate the chips through rigorous "power on" testing and laboratory qualification, a process likely to last several weeks to a month. Only after this stage, and once high-volume manufacturing is scheduled, will Nova Lake-S move closer to retail. Industry watchers expect the launch to come in the second half of 2026, with some sources pinpointing the third quarter as the most likely target for consumer availability.

The move to TSMC's advanced N2 process technology comes against a backdrop of ongoing struggles with Intel's 18A node, which is reportedly experiencing low yield rates and delays that have already impacted timelines for upcoming product families.

The SemiAccurate report confirms earlier speculation fueled by an Intel support document about the Time Coordinated Computing (TCC) platform, which referenced several unreleased Intel platforms expected to be announced soon, including Nova Lake-S for desktops.

Separately, according to reputable leakers @chi11eddog and @jaykihn0, Nova Lake-S is expected to launch in the second half of 2026 and will introduce a sweeping overhaul in both design and performance targets.

If these specifications prove accurate, Nova Lake-S will represent a significant shift in Intel's approach to processor design. The flagship desktop models are expected to feature a highly intricate, tiled architecture and up to 52 cores, built from a combination of 16 high-performance P-cores based on the new Coyote Cove design, 32 Arctic Wolf E-cores for efficient multi-threaded tasks, along with four ultra-low-power LPE-cores for background operations. Such a leap more than doubles the maximum core count seen in the previous Arrow Lake generation.

Supporting features are equally ambitious. The integrated memory controller reportedly supports speeds up to 8,800 MT/s, while graphics and media duties will be distributed across a new hybrid iGPU platform consisting of Xe3 "Celestial" and Xe4 "Druid" cores.

Intel's platform roadmap will also require users to adopt new hardware. Nova Lake-S is expected to introduce a new LGA 1854 socket and 900-series chipsets, necessitating motherboard upgrades but unlocking a suite of next-generation features tied to the new architecture.

While real-world performance benchmarks will have to wait, industry rumors suggest Nova Lake-S could deliver up to 25 percent higher speeds with as much as 36 percent greater energy efficiency compared to its predecessors. How these numbers will play out in actual computing tasks remains to be seen, but the signs point to Nova Lake-S as a potential landmark in Intel's client CPU history.

Permalink to story:

Intel Nova Lake-S chips complete tape-out, could be company's most ambitious desktop CPU yet

 

[Squid Surprise]

Here’s hoping it’s a success cause AMD needs some competition… North of 5k for a 9980x is ridiculous…

 

[HardReset]

If previous leaks hold true, Intel plans to release 16P+8E version with 125 TDP. And then put two of those on same package to get 32P+8E. But TDP only rises 25W to 150W. Not surprised if that gets cancelled.

Also good luck with sceduling. Two separate 16P+8E packages and 4 LPE cores

 

[rmcrys]

It's more probable that Nvidia releases a very high performer (ARM based) that beats Intel

 

[seeprime]

It's more probable that Nvidia releases a very high performer (ARM based) that beats Intel

What has happened that makes that "more probable"?

 

[Theinsanegamer]

What has happened that makes that "more probable"?

Intel's now long history of cancelling promising designs and fab updates springs to mind, a long with disappointing releases.

 

[DrSuess]

Let's hope it has been more thoroughly tested than some of their most recent releases.

 

[HardReset]

It's more probable that Nvidia releases a very high performer (ARM based) that beats Intel

For where? Nvidia "upcoming" CPU for notebooks/desktops is using ARM default CPU cores. Nothing groundbreaking.

As for servers, Nvidia rather promote GPUs rather than CPUs.

 

[rmcrys]

What has happened that makes that "more probable"?

ARM based architectures are improving speed much faster than x86, and an absurd if we talk about speed/TDP. Look at Apple M chips based on ARM's ISA. Nvidia has a lot of money and can secure enough engineers to make it faster than Intel.

For where? Nvidia "upcoming" CPU for notebooks/desktops is using ARM default CPU cores. Nothing groundbreaking.

As for servers, Nvidia rather promote GPUs rather than CPUs.

Much of the play with caches and even Co working with their GPU can accelerate processing a lot. Either way, they may not do much with their first generation (for this consumer space) but as Apple, again, in not to many years give it a huge boost

 

[HardReset]

ARM based architectures are improving speed much faster than x86, and an absurd if we talk about speed/TDP. Look at Apple M chips based on ARM's ISA. Nvidia has a lot of money and can secure enough engineers to make it faster than Intel.

And China designed CPUs are developing faster than ARM CPUs. As for Apple chips, main reason for speedup is usage of integrated DRAM. That gives more speed but no possibility to expand RAM.
As for TDP, ARM chips have much lower clock speeds that allow better efficiency on low clocks. Just look how much smaller are AMD c-chips vs normal. Only real difference is clock speed.

In short, ARM chips develop fast because starting point was so low. If ARM chips want more clock speed, they will encounter same problems as x86.

Much of the play with caches and even Co working with their GPU can accelerate processing a lot. Either way, they may not do much with their first generation (for this consumer space) but as Apple, again, in not to many years give it a huge boost

Accelerate CPU processing too? I doubt that.

Again, Nvidia wants to sell GPUs, not CPUs. Nvidia has so few experience developing CPUs that challenging Intel on CPUs is like Intel challenging Nvidia on high end GPUs. Possible but very unlikely.

 

[Anton Longshot]

Like many others I would like to see CPU producers/designers release desktop CPUs that can use 4-or-more channel memory.
As it is I may want to build what's technically mostly a server but components for such machines can get very expensive and super-rich I'm not.
Btw the companies involved are aware that many would like true multi-channel support.
As they obviously posses the necessary technology it's unclear why they're currently dragging their feet on this.

 

[hahahanoobs]

For where? Nvidia "upcoming" CPU for notebooks/desktops is using ARM default CPU cores. Nothing groundbreaking.

As for servers, Nvidia rather promote GPUs rather than CPUs.

Grace isn't exactly chopped liver.