AMD tuyên bố vi xử lý AI hàng đầu Ryzen nhanh hơn chip M3 Pro của Apple

2024 will go down in tech history as the year Microsoft was finally able to make Windows laptops into serious competitors to the MacBook. So far, that’s thanks to Qualcomm’s new Snapdragon chips, which switched to a homogeneous chip architecture, increased clock speeds, and caught up to Apple’s speedy and power-efficient processors. But now, AMD says it has chips that can take on the MacBook, too — and keep the company’s processors in the mix.

Last week, AMD held a two-day event in Los Angeles to reveal in-depth information about its new Strix Point Ryzen AI chips built on its brand-new Zen 5 architecture. Zen 5 is supposed to be a major leap from AMD’s last-gen chip architecture, delivering more instructions per clock cycle and higher gaming frame rates with just 15W of power. 

At that event, I heard AMD brag about beating the MacBook more than I’ve ever heard a company directly target a competitor before. AMD claimed its new Ryzen chip “exceeds the performance of what MacBook Air has to offer in multitasking, image processing, 3D rendering, and gaming”; “is 15 percent faster than the M3 Pro” in Cinebench; and is capable of powering up to four displays, “unlike the MacBook Air, which limits you to two displays only.”

But not only did AMD tell reporters its upcoming Ryzen AI chips are faster than Apple’s M3 and M3 Pro, it also said its new integrated graphics beat Qualcomm’s current-gen and Intel’s last-gen while pointedly remarking that it can power “triple-A games in full HD,” including titles that simply “don’t work on some of our competitors.” AMD also claimed its NPU performs 50 trillion floating point operations per second, more than any of its Microsoft Copilot Plus laptop competitors will offer this year.

But if they’re faster, AMD has yet to really prove it.

Games that AMD said ran faster on its new iGPU were not available for me to test at the event. Most of AMD’s AI demos weren’t actually running on AMD’s NPU, and the ones that were weren’t responsive. The most interesting of AMD’s AI demos — Asus’ automatic file consolidation and organization program — wasn’t available to try at all, and AMD’s most powerful gaming laptop on display was running its games on Nvidia graphics with Nvidia upscaling, not its own integrated graphics. 

Various AMD spokespeople gave me various answers as to why none of that was available: the demo laptops are not representative of the final product; Asus might be working on some final BIOS adjustments; the hotel Wi-Fi is too slow to install other games; they were not sure why some of the AI-powered apps were not running on the NPU.

While I wasn’t able to actually see these Ryzen AI chips in action, here’s what I was able to find out at the event.

It does sound like Ryzen AI could be notably faster than AMD’s previous generation of laptop chips. AMD says the new Zen 5 CPU architecture delivers 16 percent more instructions per clock cycle (IPC) on average, performing tasks that much faster without having to increase the chip’s clock speed.

And while its CPU cores only offer 10 percent IPC uplift in an example game (Far Cry 6), AMD says its new RDNA 3.5 GPU architecture gives these chips between 19 percent and 32 percent more graphics performance per watt at 15 watts, which is the wattage that the thinnest laptops and handheld gaming systems typically use by default. Compared to the previous generation, the integrated graphics should theoretically be able to generate more frames per second or use less power or some of each.

Even though AMD’s chips are theoretically more efficient than before, it wouldn’t confirm if these machines will have any further improvement in battery life. During the event, AMD would only say that battery life would last “all day,” which the company defines as “eight hours or more.” I was unable to speak to an Asus representative at the event to get an actual number for the laptops it demoed, and AMD seemed hesitant about giving me an answer beyond “checking with my Asus representative.”

Thin and light productivity laptops have been able to get well over eight hours of battery life for a while. So, it stands to reason that these Ryzen AI laptops could probably do the same based on the numerous improvements AMD has made to its architecture. But laptop manufacturers generally overpromise yet underdeliver on battery life. An Asus representative on this Best Buy page said the Ryzen AI Zenbook S16 gets about 12 hours of battery life, while the Qualcomm Vivobook S15 gets 18 hours — meaning Asus’ flagship AMD laptop offers six hours fewer than its Qualcomm flagship. Twelve hours is also about six hours less than the MacBook Air M3 achieved in my testing.

“I wanted to build notebooks that are faster than MacBook Pro, thinner and lighter than MacBook Air,” said AMD’s Jack Huynh, senior vice president and GM of computing and graphics, as he introduced Asus’ Zenbook S16 onstage. But when I finally got to hold the Zenbook S16 in the demo area, it didn’t feel lighter and thinner, because it apparently wasn’t.

According to the companies’ spec sheets, the 16-inch Zenbook is the same weight and thickness as the 15-inch Air: 3.3lbs (1.5kg or 1.51kg) and 0.43in or 0.45in (1.1cm or 1.15cm) thick. (It’s also 0.52 inches wider and 0.32 inches longer.) That’s not to say it wasn’t impressively thin and lightweight, because it was — but for me to be truly impressed, I will need to see Ryzen AI beat its competitors with my own eyes.

During one of the two-hour general presentations, AMD bragged that its 50 TOPS NPU is over five times faster than Intel’s Meteor Lake. (Never mind that Intel’s Lunar Lake NPU, coming this fall, offers up to 48 TOPS.) But I couldn’t get a good sense of how much faster AMD’s NPU really was compared to its competitors’ chips I’ve tested previously because the available demos weren’t running on the NPU. 

I demoed two AI programs that generated images from typed prompts on the Zenbook S16 and the 13-inch ProArt, but neither of those programs were using AMD’s NPU to run the applications. Windows Task Manager showed either the CPU or the integrated graphics doing most or all of the image generation.

There was also an MSI Prestige laptop demoing webcam effects like background blur and automatic emoji, using 51 percent of the NPU in the process, but AMD’s CPU was still being taxed — 78 percent of it, along with nearly half of the laptop’s 32GB of memory. It also couldn’t reliably generate an onscreen emoji based on my facial expression, and when it did, it took several seconds. 

One of the most interesting AI apps AMD talked up was Story Cube, Asus’ AI-powered app that comes with its upcoming ProArt series laptops; the company says it can automatically retrieve, label, and sort photos and videos according to who or what is in the photo or where the photo was taken, using local on-device AI processing.

It also appeared to be available to test in the demo area — but when I asked an AMD representative to show me it in action, they said they couldn’t. Instead, I was shown the program sitting in an idle state, as if it had already finished organizing photos.

During both its general sessions, AMD claimed its Radeon 890M iGPU could generate 52fps in Cyberpunk 2077 and Red Dead Redemption 2, and 72fps in Forza Horizon 5. With AMD Fluid Motion Frames turned on, AMD said it could get 93fps in Cyberpunk 2077, 90fps in Red Dead Redemption 2, and 148fps in Forza Horizon 5.

The company did not specify onstage (nor on some slides) the graphics settings and display resolution until it got to its comparison between the Radeon 890M plus AFMF and Nvidia’s mobile RTX 2050: full HD (1080p) on medium graphics. When I asked an AMD representative in the demo area if the same settings were used with AFMF turned off, they said yes.

AMD also claims its new iGPU runs faster in gaming than the Intel Core Ultra 9 185H and Qualcomm Snapdragon X Elite X1E-84-100: 1.65 times faster in Cyberpunk 2077 and 1.36 times faster in Shadow of the Tomb Raider. The company would not say what in-game settings it used here. They were not called out on the slide or in the footnotes at the end of the slide deck.

None of those games were available to demo with Radeon 890M, so I couldn’t verify any of AMD’s claims. In their place were Fallout 4 and Lies of Pi, but I wasn’t able to verify if either of those games was running at 1080p like AMD said they were or check out the graphics presets; they didn’t appear where they normally would be in the settings menu. I did notice that Lies of Pi’s frame rate was locked at 60fps when I enabled Steam’s in-game fps counter; Fallout 4 was running between 75fps and 95fps depending on what I was doing in-game.

I asked an AMD representative if it would be possible to run Cyberpunk 2077, Shadow of the Tomb Raider, or any other game it bragged about on the Radeon 890M but was told it would be best to wait for a review unit since performance is not indicative of the final, off-the-shelf version and that they would take too long to download because the hotel’s Wi-Fi was too slow.

The first laptops with AMD’s Strix Point chips — Asus’ Zenbook S 16, ProArt P16, and ProArt PX13 — will hit shelves on July 28th. With MacBooks and Snapdragon laptops already taking up space on those shelves, this is a crucial moment for AMD to prove that its x86 Zen 5 architecture can be just as fast — or faster — than the Arm architecture of its competitors.

If AMD succeeds, that puts even more pressure on Intel ahead of its Lunar Lake release, especially since Intel also wants to prove its new x86 chips can beat Arm. If AMD doesn’t succeed, that puts the pressure back on Intel to show the old guard of PC chips can still keep up.