As expected, AMD took advantage of CES 2025 to unveil its new processors and graphics circuits for PCs and portable consoles. A host of next-generation chips to tackle an increasingly crowded market.
Like all major manufacturers in the computer world, AMD held a conference at the Consumer Electronic Show (CES) in Las Vegas this Monday, January 6, during which the company unveiled its technological roadmap and its main new products for the year 2025. Processors for desktop computers, processors for laptops, graphics cards and processors for portable console PCs, AMD’s four main product families are enriched with new references. The result is no major revolution in terms of performance, but rather gains in small steps, and above all an increasing complexity of the manufacturer’s ranges, reinforced by the total absence of price announcements.
Desktop processors: two new models with 3D V-Cache technology
In terms of CPUs for desktop computers, a segment in which AMD has been celebrating for several years, two new models are entering the Ryzen 9000 series, the Ryzen 9900X3D and 9950X3D. These are variants of the 9900X and 9950X CPUs released last July, but this time equipped with AMD’s famous “3D V-Cache” technology, which offer an appreciable performance gain in demanding workloads and especially video games.
The Ryzen 9900X3D has 12 physical cores and 24 logical cores, displays a maximum clock frequency of 5.5 GHz in boost, a thermal envelope of 120 W and a total cache of 140 MB, compared to 76 MB for 9900X. Same thing for the Ryzen 9950X3D, which has 16 physical cores and 32 logical cores, a maximum boost frequency of 5.7 GHz for a thermal envelope of 170 W, and a total cache of 144 MB, compared to 80 MB for the 9950X.
Their hardware characteristics are therefore almost identical to those of “non-3D” models, with the notable exception of the total cache size. In terms of performance gains, nothing spectacular: AMD announces a meager 8% increase compared to the X3D models of its previous generation, and an advance oscillating between 10 and 20% compared to the Core Ultra 9 285K of its direct competitor Intel.
However, these figures are not very significant in the current state and must be taken for what they are, elements of marketing communication. The exact conditions for carrying out the tests and measurements are not known, and the final performance of these processors will strongly depend on the other components with which they are associated in users’ machines, as well as the real use cases, which give results sometimes very far from those of synthetic benchmarks.
Graphics cards: revised nomenclature, RDN4 architecture and arrival of FSR4
The company was particularly expected on the subject of its graphics cards, a sector in which it is struggling against its main competitor and undisputed market leader, Nvidia. AMD has therefore just unveiled the first two references of its new generation of graphics cards, the Radeon RX 9070 and RX 9070 XT. In passing, the manufacturer took the opportunity to modify and align its nomenclature with that of its rival, by adopting a name where the first two digits (90) represent the generation, the last two (70) the range and the suffix (XT) the variant.
We therefore see that the RX 9070 series should succeed the RX 7800 XT, RX 7900 GRE and RX 7900XT, while positioning itself against Nvidia’s RTX 4070 models. Beyond the change in nomenclature, the RX 9070 introduces the new RDNA4 architecture, which brings several significant improvements according to AMD: optimized graphics computing units, improved AI calculation and Ray Tracing accelerators, better media encoding engine and 4nm manufacturing.
All these changes aim to allow the company to make up part of its technological gap with Nvidia, particularly in the area of Ray Tracing performance, in which its GPUs have until now been less fast than those of its competitor. . The arrival of the RDNA4 architecture is also an opportunity for AMD to deploy FSR4, the latest version of its scaling engine (upscaling).
Like Nvidia, AMD will now rely on deep learning to increase the rendering definition of games, with 4K UHD in sight, and to generate additional images thanks to the frame generation. If this profound change of approach should allow a leap in terms of performance and visual quality, it will however exclude owners of older generation AMD graphics cards.
The FSR4 will in fact rely on hardware components specific to the RDNA4 architecture, and for the moment only the RX 9070 series cards should benefit from it, according to the company. A decision which could prove divisive among players, and whose success will certainly depend on the price positioning of new cards, on which AMD has unfortunately communicated no information to date. As for the RX 9070 launch window, it is scheduled for the first quarter of 2025, without further details.
Mobile processors: the Ryzen AI 300 family expands with four new chips
This is a segment in which AMD has clearly been able to make its mark over the past few years, that of processors for laptops. The company has in fact created a small revolution in this area in 2022, with its chips improperly named APUs, for Accelerated Processing Unitprocessors which bring together a CPU, a GPU and sometimes an NPU within the same component. Although the principle is not new, AMD has managed to stand out thanks to its remarkable iGPUs, the famous Radeon xxxM, which now make it possible to play games, even recent ones and in 3D, without the need for a dedicated graphics card.
The company has therefore continued its momentum and has continued to improve its formula with each generation since. The latest, that of the Ryzen AI 300 which appeared last summer, will therefore expand and accommodate four new models available in seven references (four for professionals and three for the general public). Soberly titled Ryzen AI 300 Max, this new range will therefore offer processors integrating up to 16 cores in Zen5 architecture, 40 RDNA3.5 Computing Units, an XDNA2 NPU offering up to 50 TOPS, clock frequencies up to 5.1 GHz boost and a cache of up to 80 MB.
Hardware characteristics which augur a significant leap in terms of power, and which AMD does not hesitate to highlight. The company, for example, praises the merits of its most muscular APU, the Ryzen AI Max 395, compared to Apple’s very efficient M4 Pro, in 12 and 14 core versions. AMD’s new flagship would indeed be up to 86% more efficient than Apple’s in certain 3D rendering tasks, which would indeed be no small feat. But as before, these measures must be taken with caution, while awaiting tests in a real situation.
On paper, these new chips appear particularly attractive, and suggest a very significant jump in power for laptops and mini-PCs. However, two elements should raise some questions, even some concerns. First of all, the thermal envelope (the TDP in the image above): while it was between 15 and 54 W on the Ryzen AI 300 range, it more than doubles with the Ryzen AI 300 Max, ranging from 45 to 120 W. For chips intended for mobile computers, such as ultraportable computers, or compact ones, such as mini-PCs, such heat release is worrying. To see how manufacturers will manage heat dissipation.
The other area of concern is the price of these new APUs, on which AMD has, again, given no information. When we know that the prices of the previous range, the Ryzen AI 300, were already crazy, almost twice as much compared to the Ryzen 8040 (for performances which did not double at all), there is reason to be worried about the bill for the new Ryzen Ai 300 Max. As efficient as they are, these chips could even see their quality-price ratio erode significantly if they came too close to or exceeded the cost of a CPU + dedicated graphics card pair. Their launch on the market is announced between the first and second quarters of 2025, so we will certainly know more by then.
Portable console PCs: the Ryzen Z2 takes over from the excellent Ryzen Z1
Last general public area in which AMD has made announcements: portable video games. Since the advent of the portable console PC market, the manufacturer has equipped most of the sector’s machines with its chips, such as Valve’s Steam Deck, Asus’ ROG Ally or Lenovo’s Legion Go. The latter two included a Ryzen Z1 processor, which has proven itself and been particularly efficient, even in demanding games. AMD therefore continues its momentum and announces the arrival of Ryzen Z2 chips.
This new generation of processors dedicated to portable console PCs will be available in three versions: Z2 “classic” (mid-range), Z2 Extreme (high-end) and Z2 Go (entry-level, intended only for the future Legion Go for the moment). All variants will benefit from Zen5 architecture cores and RDNA3.5 graphics computing units… which will therefore deprive them of FSR4 as explained above. A regrettable absence for processors intended for portable gaming machines, which would therefore take full advantage of the most recent scaling techniques to increase their performance.
It’s all the more unfortunate since the power gap between the Ryzen Z1 and Z2, while it will undoubtedly be appreciable, will probably only be marginal. However, the most recent games, and in particular AAA, are becoming more and more resource-hungry at an ever-increasing pace. It is therefore to be feared that portable console PCs will gradually be left behind by the graphics standards of 2025, in the absence of support for scaling or image generation technology.
Finally, as with all the other products announced, AMD has not communicated prices for the Ryzen Z2. Not surprising this time, given that these chips are not intended for direct purchase by users. We will therefore have to wait for the main manufacturers to announce their future portable console PCs equipped with these processors, probably starting in the first quarter of 2025. Note however that, contrary to what the first image above suggested, no Steam Deck embedding a Ryzen Z2 is planned, Valve having confirmed after the conference that no update of its machine was planned before a or two years.