A Few Quick Notes About Meteor Lake

I wanted to add just a few quick notes to the Meteor Lake conversation based largely on a discussion I had with an old friend from Intel.

Laurent covered the Intel Core Ultra “Meteor Lake” chipset launch earlier today, but I wanted to add just a few quick notes to the conversation based largely on a discussion I had earlier this week with an old friend from Intel.

As Laurent writes in his post, the Meteor Lake generation of chipsets is the first to integrate a Neural Processing Unit (NPU) into the CPU and GPU, and the latter is also getting a big upgrade, with Intel replacing the Iris Xe integrated graphics with a new integrated Arc GPU.

This is all great news: Thanks to its new GPU and NPU, the Meteor Lake chipsets will offload even more from the microprocessor and will offer dramatically better performance in AI and graphics workloads. Intel notes that the NPU, called AI Boost, is about two and a half times as efficient its non-integrated predecessor while the Arc GPU is about twice as fast as its predecessor and offers hardware-accelerates ray tracing, mesh shading, and AV1 encode and decode.

But in speaking with Intel this week, I learned a few things that aren’t really mentioned in its announcement and related documentation.

The first involves the schedule. I had noted earlier that the Meteor Lake release schedule deviated from the typical Intel processor release schedule of the past several years. That is, Intel normally launches the desktop versions of its new processor generations in the fall and then follows that up with the mobile versions early the following year so that PC makers can announce new laptops at CES.

Intel confirmed this was deliberate, as it wanted to get NPUs into as many customer hands as possible as quickly as possible, and so it made some changes to the schedule. First, Meteor Lake will only be available in mobile chipset form, meaning it won’t release desktop Meteor Lake chipsets. And so it moved up the release schedule so that the first Meteor Lake PCs could ship before the end of the year. I was told that this was meant to signal how important this release was to the industry.

Only the H-series Meteor Lake chips are available now: PCs based on Meteor Lake U-series chips will ship in early 2024, and you can expect a lot more PC maker announcements at CES, as always.

Related to this, the Meteor Lake schedule changes are a one-off: In Fall 2024, Intel will release the desktop versions of its next-generation Core Ultra chipsets and will follow them up with the mobile versions in early 2025. So it will be back to the usual schedule.

What’s more, there won’t be a P-series family of Meteor Lake chips, as we saw with Intel’s 12th and 13th Gen Core chipsets. This is based on several factors, but the two biggest were customer confusion and feedback from its PC maker partners.

For those unfamiliar, the U-series chipsets have typically targeted thin and light PCs, whereas H-series were for high-performance PCs like those used for gaming or workstations. The P-series line was kind of a weird in-between product line from power and efficiency standpoints, but Intel has come to believe that it’s no longer necessary. So now we will see U-series designs that come in at the low end of what the P-series was while low-end H-series chips come in at what was the high end of the P-series.

Also, Intel continues to market its not-well-understand Evo brand, which it introduced in late 2020 to counter the momentum of the M1 Apple Silicon chipsets. Evo is basically a set of specifications that PC makers can try to meet that includes such things as 10 or more hours of battery life, instant wake, fast charging, cool and quiet operation, and the like. But it’s not clear how much this spec has changed. I was told it hadn’t, but looking at the Intel documentation, I see it mentions that “the newest Intel Evo edition laptops” will include Windows Studio Effects support, which requires an NPU and is thus new. So perhaps there are further enhancements to this spec, I’ll try to find out.