2024 CPU review: a hard year

Competition breeds innovation – or so they say. Following this, Qualcomm entering the market as another CPU player should be a good thing, no? But despite wielding competitive computing power for the first time, the new model hasn’t yet been able to establish itself significantly. 2024 would’ve been a great year for this at that, since former giant Intel is slacking. A look back at the past CPU year.

Innovation in the mobile domain

Microsoft has been trying to make Windows on Arm work for years. The first devices date back to a time when Apple still relied on Intel computing units. However, the performance was so poor that they had no chance against notebooks with Intel or AMD processors. With the M1 in 2020, Apple showed that Arm architecture can do much more, offering outstanding performance and amazing efficiency.

Qualcomm shakes up the notebook market with Snapdragon X

With this year’s Snapdragon X SoCs, chip designer Qualcomm has delivered a platform to seriously compete with AMD and Intel. The chips can’t keep up with Apple’s performance, but they’re top-notch in terms of efficiency. Regarding graphics performance, however, the chips are still clearly lagging behind – they definitely aren’t suitable for gaming.

Qualcomm has an agreement with Microsoft to exclusively develop Arm CPUs for Windows until the end of 2024. But the company will soon have competition. AMD and Nvidia have already announced they also want to develop Arm SoCs.

Despite the solid Snapdragon X performance, Qualcomm is struggling. In the third quarter of 2024, the market share of Snapdragon X devices was just 0.8 per cent. It seems the architecture hasn’t yet become established in Windows notebooks.

I think this share of Arm notebooks will grow as soon as there’s some competition. AMD and Nvidia may be new to Windows on Arm, but both have experience with the architecture and still build good chips in other respects. It’s why they enjoy the trust of notebook manufacturers and customers. This new competition is likely to further boost business.

Massive increase in efficiency with x86

Mind you, x86 architecture has also made a big leap in terms of efficiency this year. AMD has shown this with the Ryzen AI 300, as has Intel with Lunar Lake. These AMD devices stand out in particular. In addition to a longer battery life, they also take a step forward performance-wise. Intel’s Lunar Lake is less advanced in terms of performance, but it’s still significantly more efficient than its predecessor. Something that also applies to Intel’s desktop CPUs.

Ailing Intel

Intel has had a horrendous year, plain and simple. In quarter three, the company posted a loss of 16.6 billion US dollars. Most recently, CEO Pat Gelsinger resigned – probably under pressure from the Board of Directors. He hasn’t succeeded in implementing his four-year turnaround plan – mind you, the four years weren’t even over yet. Gelsinger already inherited a shambles when he took over as CEO in 2021.

Intel’s biggest issues

Intel’s problems go back pretty far and are complex at that. On the one hand, the company has failed to produce competitive chips for smartphones. This is doubly confusing since the chip giant recognised the potential of these devices even before the first iPhone. But products like the Intel Atom couldn’t prevail against Arm superiority. On top of this, the company continued to stubbornly focus on x86 architecture.

What’s probably even worse today is Intel’s insignificance in the second major development of the last 20 years: the AI boom. The company simply doesn’t offer any competitive products. There have been attempts (article in German), mind you. Intel acquired AI startup Nervana in 2016 and chip manufacturer Habana Labs in 2019. Gaudi 3 is currently Intel’s AI accelerator flagship. The company wanted to sell 500 million US dollars worth of those. It’s already obvious this target won’t be met. For comparison, AMD wants to generate three billion US dollars with AI GPUs for data centres. Nvidia doesn’t disclose any figures, but is expecting 80 to 90 billion US dollars.

As if that weren’t enough, Intel has been struggling with production for years. In particular, the transition from the 14 nm to the 10 nm technology node was delayed by years. The upcoming Intel 18A manufacturing generation also isn’t fully developed yet (article in German). Intel is also lagging behind competition from TSMC. What’s more, the company is now dependent on the Taiwanese manufacturer, since the latest Core Ultra 200S desktop generation is manufactured by TSMC.

New desktop CPU generation unimpressive, old one is faulty

As if manufacturing with the competition wasn’t enough of a slap in the face for proud Intel, the Core Ultra 200S chips are also unpolished. Their performance lags behind the previous generation in gaming, and results are mixed in other applications. At least the chips make a leap forward in terms of energy efficiency. Intel thought it found the reasons for this poor performance and subsequently provided various fixes – but the chips still aren’t significantly better. As if that wasn’t enough, they’re only at the level they should originally have been at two months after release.

Intel urgently needed positive news. Sadly, reports in July only set the ball rolling for an even worse year. Apparently, a bunch of CPUs from the 13th and 14th Core i generation were giving up the ghost. The reason? A microcode problem leading to excessive voltages. Intel has supposedly released a fix. But since the problem only emerges over time, it’ll take a while before it can be definitively described as solved.

But the damage has already been done. Not only from the error, but above all because of the communication and the way customers were treated. Intel initially didn’t want to be held liable for damages. The manufacturer only relented after those affected put up a fight. There goes your customer trust.

This is doubly clear from the sales figures. For example, the 13th and 14th Core i generations were barely selling (article in German) shortly before the release of the Core Ultra 200S. The latest generation was unable to change this. AMD also currently accounts for the majority of desktop CPU sales in this country. Mind you, they weren’t flawless either.

Unrefined AMD devices

AMD’s Ryzen 9000 really should’ve been released at the end of July. Shortly before this, however, the company bit the bullet and postponed the launch by two weeks. The reason? Unspecified quality problems. Maybe AMD would’ve done well to postpone the launch even further, as both the «small» and «large» models failed to impress in initial reviews.

Various teething troubles, such as faulty operating system settings and latency problems, prevented the CPUs from reaching their full potential. Only following several fixes – over a month after release – were the problems resolved, and the processors are finally working to their full potential.

AMD then ended the 2024 CPU year with a bang. The Ryzen 7 9800X3D didn’t just become the new best gaming CPU, it also solved the problems of its predecessors. They were always slower in applications than their non-X3D counterparts, but that’s a thing of the past now. The 3D-V cache is now located below the actual computing unit, leading to lower temperatures and higher clock frequencies as a result. The X3D processors can finally unleash their full potential. It’s why I think the 9800X3D is the best gaming chip of the decade so far.

What I’m taking away from 2024

2024 was an exciting year in terms of CPUs. Thanks to Qualcomm and the Snapdragon X, something is finally happening in terms of efficiency for non-Apple products. With the Ryzen AI 300, AMD has proven that x86 architecture isn’t completely left in the dust by Arm in this respect either. If other players with Arm SoCs for notebooks come onto the scene in 2025, things are likely to get really exciting.

But 2024 has also provided warning signs: manufacturers are increasingly rushing their products onto the market. The latest desktop generations from AMD and Intel simply weren’t ready yet – it’s not surprising that customers feel like beta testers. Fixes have been released or are still being released at least, but I’d still prefer if manufacturers took more time.

It’s also clear that leaps in processor performance will tend to be smaller in the future. Physically, manufacturing technologies are increasingly reaching their limits – Intel Foundry and the resulting problems for the company are the best example here. Manufacturers have to invest more and more in software and operating system implementation. This year, AMD announced that it’ll develop more in the direction of software. Something that Nvidia, for example, has been doing for years.