6 takeaways from AMD’s consumer chip announcements at CES

AMD CEO Lisa Su kicked off CES with a keynote on AMD’s chip announcements, from Ryzen 7000 series of mobile processors to its mobile GPUs from the RX 7000 series to the high-powered X3D class of desktop GPUs. Some key themes are the growing role of AI in everyday computing (as well as traditional roles) and the impact (and needs) of hybrid working.

Su also touched on some of the more commercial aspects of his partners’ products, such as relatively detailed discussions about the role of robotics and Magic Leap’s AR in healthcare (the latter just won Attestation 60601). We were treated to an astronaut talking about AMD’s research and adaptive computing, such as the Perseverance Mars Rover and simulating (wo)manned missions to the moon. AMD isn’t just in your console: it’s in space.

He concluded by discussing sustainability in its components and technologies and how this translates to large data centers and high performance computing. It also launched the AMD Instinct MI300, a GPU/CPU combo designed for data center-scale AI training and services. It comes in the second half of this year.

Note that some of this is taken from separate background documents in addition to what was compressed during the main tone. You can follow more CES news on our live blog, as well as more details on partner commercial announcements.

On-chip adaptive AI is AMD’s most exciting mobile development

AMD partially announced its XDNA architecture to speed up AI inference, which could go up or down depending on how many processing units the hardware developer throws at it. (It came from the company’s acquisition of Xilinx.) Hence the “Adaptive AI” brand. Because it can pass data from array to array without having to interact with external memory or the CPU or GPU, AMD says this lowers latency and makes it a flexible architecture.

All of this is a prelude to the launch of the company’s next-generation HS mobile processors, which include a small built-in adaptive AI accelerator that can handle up to four simultaneous AI streams, for separate tasks or to execute a task with parallel threads. It’s optimized for power efficiency, and because it frees up CPU and GPU cycles that AI processing can do, AMD says it results in less performance degradation and better battery life.

Artificial intelligence algorithms are pervasive in computing these days, and mobile devices increasingly rely on them for even mundane operations, such as cleaning up video in video conferences and improving battery saving behavior. AMD claims it is up to 50% more efficient than the Apple M2 neural engine. Intel uses a separate chip for AI acceleration, called Movidius, which means its presence in a laptop depends on individual manufacturers.

Su also previewed the Alveo V70 Accelerator, a small, low-power (75-watt) AI inference accelerator board. It’s up for pre-order now and will ship in the spring.

HXs are faster, but HSs are more cutting edge

It’s a schism I don’t really enjoy. Intel and AMD use “HX” to refer to processors for maximum performance in gaming or creative laptops, which are often used by people looking for the latest technology (and usually willing to pay for it). But the aforementioned Adaptive AI accelerator is only in Ryzen 9 7940HS, Ryzen 7 7840HS and Ryzen 5 7640HS, the processors designed for thin but powerful models, and they also have the advantages of using the 4 nm manufacturing process , while HX is still at 5nm (which partially explains how they can safely achieve higher clock speeds).

A smaller process tends to be more energy efficient and faster. The HS needs it to hit the 35-45 watt power target for this class of laptops, but given the amount of AI used in creative apps and games, it would have been nice to get some power bonus on the smallest die . .

HS chips also use the newer RDNA3 architecture for their integrated graphics. This is not a rare practice; the reasoning is that the HX class of laptops will use discrete GPUs. But most laptops with discrete GPUs still rely on the integrated version for, for example, lower power consumption when discrete GPU speed is not needed.

On the other hand, the HX chips are the first mobile processors used by AMD chiplets. And the best Ryzen 9 has 16 cores/32 threads.

Gaming Laptops announced at the show that takes advantage of all that AMD has to offer is the Alienware M16 and M18, Asus TUF Gaming A16 Advantage Edition and a new Lenovo Legion Pro series. The first 7040 series Ryzen laptops will start shipping in March. HP CEO Enrique Lores came on stage to present the new Dragonfly Pro laptop.

USB 4 support is not enough

It’s bad enough that it’s been another year without Thunderbolt support on AMD laptops, but even USB 4 isn’t supported at the lower end of the CPU spectrum where laptops live. “thin and light mainstream” chips. And information from AMD seems to indicate that even the HX models may not have it since the USB 4 controller may be optional.

AMD is deliberately undercutting 65w desktop processors

And this is a good thing. The 65w class of desktop processors is aimed at low-power, small form factor, or low-cost PCs, and tends to be popular with upgraders. For this generation, AMD designed the processors to run at 65W, but will be able to squeeze nearly double that through overclocking, which the company says can deliver a performance boost beyond 65W. by 30%. Since so much of everyday computer use is for mundane tasks — at least when you consider this class of processor — it makes sense to default to a low power state.

Ryzen 9 joins the X3D generation

AMD’s cross-generation powerhouses use the company’s 3D V-Cache architecture to cram more cache into the chip to deliver better performance with essentially the same processor. This year, AMD is adding the Ryzen 9 7950X3D to the mix – there’s also some Ryzen 7 like before, the 7900X3D and 7800X3D. AMD will get together Star Wars Jedi: Survivoran anticipated game, with processors.

Mobile discrete GPUs are not very exciting

It might just be me, but there’s nothing really remarkable about the RX 7600M, RX 7600M XT, RX 7700S and RX 7600S. They seem like a nice step up from last year’s offerings, but they’re not as worthy of the HX class as, say, the RTX 4090 or RTX 4080 mobile. AMD GPUs in conjunction with CPUs, such as SmartShift Radeon Super Resolution upscaling, FidelityFX Super Resolution 3 (a better class of upscaling), and Hypr-RX mode (one-click frame rate maximization and frame rate minimization) latency using all available settings) are not here yet. In the first half of this year, sometimes, we are told.