5 min

Today, Intel unveils its new generation of laptop chips. Codenamed Meteor Lake, this generation is meant to unleash the AI revolution to consumers. That revolution starts with so-called Intel Core Ultra processors, aimed at midrange and high-end laptops.

This year, almost all imposing AI applications run in the cloud, with only some simple applications computed locally on a PC. Things like background blurring in a webcam image or noise removal from speech don’t really capture the imagination. Now, Intel hopes to enable advanced solutions on local PC hardware as well, such as chatbots, AI-generated images and all sorts of applications that Intel says we are currently unaware of. This will require not only software, but also hardware. The first step towards that is Meteor Lake, and branded as Intel Core Ultra on laptops. The AI revolution is so important to the company that it is restarted the count all over again: this successor series to the 13th generation Intel Core chips will be known to the wider world as Generation 1.

The promise is one of AI assistants that “know everything about you,” as Intel puts it. That requires the use of proprietary, privacy-sensitive information, so that an application will be able to peer into your calendar, edit your photos and know your browser history. That may sound scary, the company admits, but the software achievable with AI should be worth it. After all, everything stays safe on your own system.

The debut of the AI PC

First of all, Intel talks about the introduction of “AI PC,” referring to the specific hardware capable of running AI. A Meteor Lake chip contains not only a CPU and GPU but also an NPU (Neural Processing Unit), which can handle AI workloads with highly efficient power consumption. The end result is impressive: every AI calculation is now up to 2.5 times as efficient as before.

Jan van Offeren, Industry Technical Specialist at Intel, explains that Meteor Lake is built with several “tiles,” basically several small chips that collectively make up the processor. It’s a philosophy shared by its server-bound Xeon chips. The goal is to create the ideal performance-per-watt. That sounds highly logical, especially for the laptop form-factor, since cooling and power consumption options are both limited. However, Intel took its clock frequencies skywards with the past 13th generation chips, with all the consequences that entails. The end result was disappointing battery life and meteoric temperatures, but that should now be a thing of the past with Intel Core Ultra products. The new tiles can be turned off when not needed.

Intel Core Ultra5 125U5 135U7 155U7 165U
Cores (P-/E-/LP E-cores)12 (2+8+2)12 (2+8+2)12 (2+8+2)12 (2+8+2)
Threads14141414
Intel Smart Cache12 MB12 MB12 MB12MB
P-core max turbofreq.4.3 GHz4.4 GHz4.8 GHz4.9 GHz
E-core max. turbofreq.3.6 GHz3.6 GHz3.8 GHz3.8 GHz
Graphics max. freq.1.85 GHz1.9 GHz1.95 GHz2.0 GHz
Xe-cores4444

Specifically, four tiles are involved: compute, graphics, I/O and the SoC. The first two are CPU and GPU, respectively, with integrated graphics skills expected to be twice as powerful as before. It is optimized for 3D performance, Van Offeren tells us. The implicit goal seems to be to eventually say farewell to the need for discrete GPUs with the proprietary Arc graphics architecture. Don’t let Nvidia hear about it though. The SoC-tile is entirely new. That is precisely where they focus on efficiency, so when little is expected of the laptop in question in terms of performance, the compute and graphics tiles turn off. The SoC features a new kind of Low-power Efficiency (LP E) core, which reduces power consumption even further than the existing Efficiency (E) cores. From now on, those watching Netflix, for example, will rely only on the very economical LP E cores.

To make development as easy as possible, Intel relies on open standards. For example, there’s OpenVINO, a toolkit that allows developers to build applications hardware agnostic. Specifically, Intel dynamically controls which hardware resources are appropriate for a workload. The focus is ultimately on the software, as Intel VP & GM of Client AI John Rayfield also made clear earlier this year.

Not better than before on all fronts

Interestingly, Intel’s latest chips are not always faster than their predecessors. Compared to the 13th generation, single-threaded performance is sometimes degraded, mainly due to the decreased clock frequency (a maximum of 4.9 GHz versus the 5.6 GHz of before). In terms of gaming, though, performance is significantly stronger than the previous Intel generation.

Obviously, the new Meteor Lake chips are positioned differently than Intel did with previous generations. It focuses on AI, efficiency and emphasizes the importance of software. In any case, the company is already working with more than 100 software vendors to get AI-powered applications onto the PC.

At Intel’s presentation, we also got to hear from Sales & Marketing Director at Acer Oscar Brouwers. The Acer Swift Go 14 was the only actual product that accompanied the unveiling of Meteor Lake, which will go into full production in 2024. Acer has always been a major Intel OEM partner. For OEMs, the heyday of PC sales during the COVID pandemic is a thing of the past. For that reason, Intel now seems to be plunging fully into AI and efficiency to ensure that consumers and businesses have a reason to buy PCs en masse again. Either way, according to Brouwers, it’s important to get back to telling a story with your product to convince people to actually make a purchase. The new Intel Core Ultra laptops will therefore have to impress not only in terms of raw numbers, but also convince people to dive into this new type of AI hardware immediately.

The problem for Intel will be that it cannot control how fast the AI revolution proceeds. However, it does do well to at least provide AI-capable hardware so that all the promises of ubiquitous artificial intelligence should actually be possible. As a consumer, however, you may wonder what software will already be in place next year that takes advantage of it, and whether you can wait a little longer. After all, the lesson from previous revolutionary designs is that their second and third versions advance by leaps and bounds over the product that debuted with one. Just look at Nvidia’s RT cores to enable ray-tracing on a local device, which took two generations to really mature.

Newborn process

Maurits Tichelman, VP & GM EMEA Partner Sales, puts Meteor Lake’s development in the broader context of Intel’s roadmaps. He reminds us of Intel’s “5 nodes in 4 years” plan. With this, it aims to regain the lead as a chip manufacturer after years of being behind the times. This catch-up already seems to be paying off, as Nvidia is apparently also open to using Intel Foundry Services. Tichelman does state that we won’t always hear who is using these services (he didn’t state Nvidia’s interest either, Nvidia’s CFO did). That’s ultimately a story for the longer term, for now this feat is especially important to make Intel’s own chips competitive again on all fronts.

The Meteor Lake chips are partially produced in Ireland, where Intel has been operating for thirty years. There, for now, sits the only facility where the brand new Intel 4 manufacturing process is operational, made possible by ASML’s EUV scanners. The smaller process node is a key reason for the efficiency gains previously promised.

Also read: 5th generation Intel Xeons: powerful AI performance, sizable cost savings