AMD Ryzen AI 300 vs Qualcomm Snapdragon X chips

Which is faster, the AMD Ryzen AI 300 or the Qualcomm Snapdragon Elite? That’s the key question I answer here, having tested four Snapdragon Elite laptops over the past month and a Ryzen AI 300 laptop since it arrived during the week.

To be clear on exactly what I’m comparing, AMD sent me an Asus Zenbook S16 laptop with the AMD Ryzen AI 9 HX 370 inside. And the four Qualcomm laptops I’ve tested are the Microsoft Surface Laptop 15, the Asus Zenbook S15, Lenovo ThinkPad T14s and Samsung Galaxy Book4 Edge 16in.

That means I’ve tested a wide variety of Qualcomm Snapdragon X chips, so I can give you a good idea of how the Ryzen AI 9 HX 370 inside the Zenbook S16 matches up across a series of tests. I’ve also thrown in one laptop powered by an Intel Core Ultra chip and another using AMD’s Ryzen 7840U CPU.

Note: this article was updated on retesting the Asus Zenbook S16 after updates on 29 July 2024.

Meet the Ryzen AI 300 family

Although AMD has doubled down on AI with this series of chips, both in terms of the name and the features, what most people should care about right now is that it’s based on AMD’s new Zen 5 architecture.

This is Zen 5’s debut in laptops, and AMD claims that it’s 16% faster (in terms of “instructions per clock”) than the equivalent Zen 4-based processors. They should also be more energy-efficient than before, thanks to a smaller 4nm die, and AMD has boosted its graphics skills, too.

Then there’s the NPU, or neural processing unit. Depending on which processor you buy, this will pack up to 55 TOPS of AI power. That’s 5 TOPS more than Qualcomm’s Snapdragon X family, and 10 TOPS more than is required to meet Microsoft’s Copilot+ PCs criteria.

Meet the Snapdragon X family

Qualcomm’s Snapdragon X family currently consists of four processors, which compare as follows:

The main takeaways:

• buying a Plus-powered laptop means you only have 10 cores to play with
• all the Snapdragon X chips include NPUs rated at 45 TOPS
• the top two chips include notable dual-core boost speeds
• if you want it all, you need to buy a laptop with an X1E-84-100 inside

Ryzen AI 300 vs Snapdragon X: the key differences

The most important difference between the Ryzen AI 300 family and the Snapdragon X is one of architecture. AMD’s chip uses the familiar x86 architecture that has been dominant on Windows for decades, while Qualcomm’s chip uses the Arm architecture.

I won’t go in-depth into the philosophical differences here (Red Hat includes a detailed article on the subject), but x86 chips have long been considered more powerful and Arm chips more efficient. Apple ripped up that belief system by designing its own Arm-based chip, the M series, which was both incredibly powerful and efficient.

By working closely together, Microsoft and Qualcomm have produced a new version of Windows for Arm that very nearly achieves what Apple did. As mentioned above, I’ve tested four Snapdragon-powered laptops now and they’re all fast and boast excellent battery life. In one case, over 24 hours.

Previously, I would have advised most people to steer clear of Arm-based Windows laptops. That’s purely down to software support: the vast majority of Windows software was designed for x86 chips. While Microsoft did provide what’s called a “translation layer” to convert x86 software so that it runs on Arm hardware, in practice it was often slow and buggy.

Now, however, Microsoft has released Prism emulation, and in general it’s worked extremely well in my tests. The only big area where it performs poorly is in games.

The memory question

There is one other area to note when comparing Ryzen AI 300 processors with Snapdragon X chips. Technically, the Snapdragon series are considered to a system-on-chip, or SoC. That means almost everything is built directly onto the die, including the computer’s system memory (aka RAM).

This means you can’t upgrade the memory at a later date.

With the Ryzen AI series, the memory still sits apart from the main CPU. So long as the memory isn’t integrated onto the laptop’s motherboard (which happens more often than I’d like), that means you can upgrade the memory or replace it if it becomes faulty.

How I tested

You can’t isolate a laptop’s processor, so I’ve had to rely on my mix of ever-so-slightly different laptops to give you an idea of their power. Below, I give a summary of each laptop’s main components and when I performed my tests. The latter is relevant because laptop makers update firmware when they discover bugs or potential enhancements, so I would generally expect more recent laptops to perform slightly better.

In all cases, I made sure that each laptop was in its peak performance mode, both in Windows settings and in the manufacturer’s settings. For example, Asus offers a Performance mode where fans activate more readily.

I tested all the laptops across a range of tests, including Geekbench 6.3, Cinebench R23, Cinebench 2024, 3DMark Time Spy and two games: Dirt 5 and Shadow of the Tomb Raider.

I also tested them using PCMark, but its tests are not comparable across x86 and Arm architectures so I don’t include them here.

Ryzen AI 370 vs Snapdragon X: Cinebench

Cinebench is a CPU-intensive test that renders an image multiple times across a ten-minute period. It’s useful because it tests the processor rather than whole system (such as the graphics chip). It also puts the CPU under heavy load for at least ten minutes.

Cinebench R23 has been around for a few years, while Cinebench 2024 is new this year. It introduces a GPU test as well, but none of the machines here meet its requirements so we’ll ignore that for now.

Also note that R23 and 2024 results are not comparable with each other.

A quick word here about benchmark results. Most of the results you see above are medians – so the best representation across a number of tests, although I strip out weird results that might be due to a background process kicking into action.

Benchmarks give a good guide to a laptop’s speed, but they also represent a point in time. I saw this with the Ryzen AI 370 HX in Cinebench’s multicore test. In Cinebench 2024, repeated tests gave scores around the 800 mark. However, a reboot and update the following day saw this jump up to 915.

This effect was even more pronounced in Cinebench R23. Initial testing gave results under 10,000 in the multicore benchmark, but this jumped to 14,462 after the update.

Both these scores are up compared to the Framework Laptop, which uses a Zen 4-based processor. Likewise the single-core results. But not by the 16%: in the multicore tests, it’s up by 3% (R23) and 13% (Cinebench 2024). In the single-core tests, that’s 11.5% (R23) and 5% (Cinebench 2024).

It’s a mixed result compared to the Snapdragon X chips. In Cinebench 2024, the is roughly level with the X1E-80-100, but substantially up in the older R23 benchmark.

For now, let’s not draw any big conclusions and move onto Geekbench.

Ryzen AI 370 vs Snapdragon X: Geekbench 6

Geekbench 6 takes a different approach to Cinebench. Rather than rendering a scene, it sets up a group of CPU-related tasks (such as encryption and unzipping a compressed file) and asks the computer to complete them as quickly as possible. The quicker it completes the set of tasks, the higher its score.

Like Cinebench, Geekbench provides single-core and multicore results, but typically the test lasts around four minutes so it’s not as exhaustive. I tested using Geekbench 6.3, which is the latest version.

This is a neater set of results in that it tells us the Snapdragon chips have a speed advantage in the multicore tests. For single-core power, the two Qualcomm chips that can boost up to 4.2GHz (the 84-100 and 80-100) clearly benefit, putting them on level pegging with AMD’s new processor.

Ryzen AI 370 vs Snapdragon X: AI performance

It’s early days for the neural processing units built into chips like the Snapdragon and Ryzen AI 300 series. Windows has a handful of AI features that you may or may not use – better blurring in video calls remains the most prominent – and I’m yet to be blown away by third-party software.

In truth, I haven’t yet found a reliable way to benchmark AI chips too. We have the TOPS ratings from each chip manufacturer, but it boils down to how the software makers make use of them. Purely for information, I will share how each laptop performed in the early AI benchmark tool Geekbench ML 0.6.

Note that I returned the Surface Laptop and Vivobook S15 to Microsoft and Asus before I could run these tests on them.

Ryzen AI 370 vs Snapdragon X: 3DMark Time Spy

In short, the Ryzen whips the Snapdragon chips every which way. And it also delivers a significant upgrade over Intel’s current rival and the previous-generation AMD graphics chip.

However, 3DMark Time Spy is a synthetic test, so let’s see how the Ryzen AI 300 series compares with Snapdragon X chips in real games…

Ryzen AI 370 vs Snapdragon X: real games

When I was testing my first Snapdragon X-based laptop, the Asus Vivobook S15, I had real trouble with games. It refused to even load up Shadow of the Tomb Raider. Things have improved since, with all three of the other Snapdragon-powered laptops running it fine – I get a display driver error each time the game loads, but it still runs.

But, as you shall soon see, these are not gaming laptops. That’s partly because they’re having to run using the Prism translation layer I mentioned above, but I think it’s also due to the Adreno 3D chip. This is fine for less demanding games on phones, but PC gaming requires a step up. I think we will have to wait until the next iteration of Snapdragon chips before we see enough horsepower to produce high frame rates in demanding PC games.

A quick note on why I chose the games. I know from experience that Dirt 5 can run on a laptop CPU’s integrated graphics (as opposed to discrete graphics, when a laptop maker slaps in (normally) a separate Nvidia graphics chip), with racing games typically less demanding than games such as Metro or Borderlands. Likewise Shadow of the Tomb Raider, which has plush graphical effects but is now a few years’ old.

Notably, of all the Snapdragon X laptops I’ve tested, only the Samsung Book4 Edge has gone above 29fps in any of these tests. As a rule of thumb, games are playable if they’re around 35fps on average.

While the AMD-powered Zenbook struggled in Dirt 5 at High settings, and only just breached that 35fps marker in Shadow of the Tomb Raider, you can see that games are playable on this laptop if you’re willing to drop down in quality settings.

AMD Ryzen AI 300 vs Qualcomm Snapdragon X: conclusion

I’ll be honest: I expected the Ryzen AI 300 series to blow the Snapdragon X away in the majority of these tests. Not only does it benefit from 12 performance cores, and 24 threads, but also the latest Zen 5 architecture.

I have come to expect great things from AMD’s laptop CPUs in the past few years, so thought Zen 5 would again be a hefty chunk faster than Zen 4 overall.

Instead, what we can see is that the top-end Snapdragon X Elite, the X1E-84-100, is faster than AMD’s Ryzen AI 370 HX in everything except for Cinebench R23 where it’s much slower – but then it’s faster in the newer Cinebench 2024. And in single-core tests, there’s no clear-cut advantage: the two top Snapdragon chips are the Ryzen AI 370 HX’s equal.

The sole area where the Ryzen reigns supreme is when playing games. As for AI… I’m holding out for better benchmarks.

I haven’t completed my battery tests yet, but I can confidently say that we are not going to see the 24-hour battery life that Lenovo managed with the ThinkPad 14s on any Ryzen AI 300-powered laptop.

This gives prospective buyers of Copilot+ PCs clear dividing lines. If you want to play it safe in terms of compatibility, and like playing games after hours, the Ryzen AI 300 series is the best choice around.

However, if you’re looking for snappy performance and exceptional battery life, then the Snapdragon-powered laptops are your friends.

Related articles:

• AMD Ryzen AI could push people from the cloud to local machines
• Qualcomm CMO on AI and cloud: “Hybrid approach will win the day”
• AMD launches Ryzen Pro 7040 series: an AI-enhanced Core i7 killer for professional laptops