Intel takes a swing at Qualcomm and AMD with Core Ultra Series 2 (Lunar Lake) superiority

By now, you should already be familiar with Intel's next-generation notebook CPUs, codenamed Lunar Lake, officially known as the Core Ultra 200V processors or collectively the Core Ultra Series 2. We've previously covered here, as well as during Computex 2024. Intel has just officially launched Lunar Lake as a precursor to IFA 2024 in Berlin, and we're finally able to share some of Intel's claimed performance advantages and some early power draw and benchmark numbers. For information about processor SKUs and availability, check out our launch news article here.

TL;DR: Intel claims the new Core Ultra Series 2 (Lunar Lake) processors will beat any thin-and-light laptop across all applications and use case scenarios with lower power consumption, better battery efficiency, and yet offer higher performance across CPU, GPU and NPU, with unmatched AI compute and compatibility.

Play

No Hyper-Threading? No Problem

Much has been said about Intel removing Hyper-Threading from Lunar Lake, but this is not the first time Intel has done so. During the shift from Pentium 4 to Core Duo architecture, Intel removed Hyper-Threading in favour of true dual-core computing. Intel then reinstated Hyper-Threading in a succeeding Core architecture starting with Nehalem. The reasoning for this apparent flip-flop is actually very simple. Intel has always tried to balance performance-per-watt. Usually, it’s either improving performance while maintaining the same package power or reduce power consumption while maintaining the same level of performance.

With Lunar Lake, Intel chose to increase performance and reduce power. They are claiming higher performance-per-watt-per-thread with just 8 cores vs. Meteor Lake’s 22 (with HT). In fact, the break even point is around 20-threads at about 20W. And if you look at the chart above, Lunar Lake falls slightly below Meteor lake at 23W with a full 22-thread workload. However, remember that Lunar Lake achieves this performance with just 8 cores. Intel did test out what it would be like to enable HT on Lunar Lake, and the trade-off for performance-per-watt wasn’t worth it. I’m quite confident that when a future architecture makes sense to do so again, Intel will re-enable HT once more.

Ultimate Efficiency

Intel claims to have reduced the overall package power of Lunar Lake by up to 50% over Meteor Lake in almost every benchmark and applications they’ve run, and this is with 32GB on onboard memory. And with the power reduction, battery life has massively increased to over 20 hours. I talked to a few Intel partners launching Lunar Lake notebooks such as Acer, Asus and Dell, and they’re all claiming continuous video playback battery life around the 25- to 27-hour mark.

In a live demo, a Lunar Lake notebook was only drawing around 5W of power decoding a 4K video, while the previous generation Meteor Lake was pulling upwards of 11-12W. We also saw a gaming comparison where the Core Ultra 9 outperformed AMD Strix Point and Qualcomm Snapdragon X Elite with a similar power draw, while the Core Ultra 7 matched AMD and outperformed Qualcomm with a much lower power draw. You can see this in the embedded video here.

Better Performance Across the Board

CPU, GPU, NPU, Intel is staking claim for all three with synthetic and real world benchmarks. Of course, do take these presentation charts with a grain of salt. We will definitely test them out ourselves when we get our hands on Lunar Lake notebooks for an actual review.

However, we were able to see and run some direct benchmarks with Procyon, Tomb Raider, Cyberpunk 2077, and 3D Mark on site on a mix of laptops. From what I can tell, they're all ASUS Zenbooks, but some of them are running Core Ultra 7 and some are running Core Ultra 9. For example, the Tomb Raider benchmark was run on a Core Ultra 9 288V while the CyberPunk 2077 benchmark was running on a Core ultra 7 258V.  so I’ll just leave these numbers here, make of it what you will. Again, check out the embedded video for some additional commentary and live demos that we saw as well.

AI isn’t just about the NPU

Intel themselves agreed that TOPS isn’t indicative of real-world performance, and I felt during the launch that they’ve deliberately steered clear from identifying the AI PC as a sole function of the NPU. Yes, there’s an NPU in Lunar Lake. Yes, it has up to 48 TOPS. But the majority of real-world AI workloads are still run on the GPU, and Intel made it a point to showcase the new Arc graphics with Xe2 architecture on Lunar Lake. Whether it’s XeSS for games, laptop ray-tracing, 8K 10-bit decoding, or GPU diffusion, Lunar Lake does it all. Intel claims to have the highest compatibility for an AI PC to run the most AI models and framework.