Tests reveal: Intel Meteor Lake CPUs still deliver 93% of performance on Linux after 2 years

A battery of tests carried out this week by Phoronix brought surprising (and worrying) results for car owners. notebooks equipped with processors Intel Meteor Lake with Linux operating system.

The detailed analysis published by the portal reveals that, theafter two years of kernel and driver updates on Linuxthe performance of this architecture suffered a measurable regression.

Contrary to the expectation of continuous optimization (the famous “fine wine” effect), the performance of Meteor Lake no Linux fell to one average of 93% compared to the numbers recorded at launch, in December 2023.

The test: Core Ultra 7 155H two years later

To reach this conclusion, the same hardware reference of the original reviews: one Acer Swift Go 14 equipped with the Intel Core Ultra 7 155H chip.

The comparison brought the 2023 software scenario face to face (UUbuntu 23.10, Linux 6.7) versus the current end-2025 environment (Ubuntu 26.04 Daily, Linux 6.18 e GCC 15.2).

When running more than 200 benchmarks diferentes, the geometric mean of the results indicated that the current system is approximately 7% slower than the original factory setting.

This behavior is atypical in the open source ecosystem, where kernel updates generally bring refinements in task scheduler and better power management for processors Intel hybrids.

The data becomes more alarming when placed alongside contemporary or successor architectures. In the same year-end testing period, other platforms showed consistent gains:

• Intel Lunar Lake: recorded an average gain of 6% in performance after one year.
• Intel Arrow Lake: increased around 9% in desktop performance, with reduced consumption.
• AMD Strix Point e Krackan Point: Rival chips from AMD have seen improvements of between 5% and 8% thanks to the maturation of graphics and chipset drivers.

While the rest of the industry and even Intel’s own other products have advanced, Meteor Lake appears to have suffered under the weight of security updates and changes to the operating system structure.

The result of security fixes

Although the initial report does not pinpoint a single cause, experts point to the cumulative impact of mitigations (security fixes) via software and microcode. Specifically, mitigation for Branch History Injection (BHI), known as BHIDISS, is enabled by default for Meteor Lake in newer kernels.

The extra layer of protection, necessary to close speculative vulnerability gaps, usually takes its toll in processing cycles.

This would explain why, even with Mesa 25.2 graphics drivers more modern and optimized compilers (GCC 15), the CPU cannot maintain the score it had when it arrived on the market.

Power consumption (SoC Power) remained practically unchangedindicating that the loss of efficiency was not accompanied by a proportional battery saving.

Also read:

Perspectives for the user

The performance regression in Meteor Lake serves as a stark reminder of the balance between security and speed in the hardware world.

For the end user of a Core Ultra 100 notebook, the 7% difference may not be noticeable in everyday browsing or streaming tasks, but it impacts code compilation, rendering and heavy workloads.

With the imminent arrival of Panther Lake, the question remains whether Intel will be able to reverse this situation via specific optimizations in the Linux kernel or whether Meteor Lake will be marked as a generation that, ironically, ran better the day it came out of the box than years later.

Fonte(s): Phoronix