This year’s CES has become a notebook-centric event in the PC space, and look no further than AMD’s CES conference for proof of that. The densely packed presentation began immediately with the announcement of AMD’s 2023 mobile product stack, which will see the CPU supplier mixing and matching silicon across multiple design generations to bring together a new product stack for the new year . This includes the return of some old favorites, including bringing desktop Zen 4 silicon to mobile, as well as the introduction of AMD’s new Phoenix CPU silicon, its first mobile-focused Zen 4 CPU design. There’s a lot to unpack here, so let’s dive in.
New Names, New Categories, Mobile Silicon Smorgasbord
First, this year marks the departure point for AMD’s new mobile CPU naming system. First introduced by the company in September, AMD is moving to a model year system, incrementing the first digit of its processor model numbers each year. In practice, AMD has committed to offering a combination of new silicon and product updates (eg new or rebranded SKUs) each year. Given AMD’s affinity for announcing new mobile CPUs at CES, I suspect this means we’ll be talking about a new top-to-bottom AMD mobile CPU product stack at each and every CES from now on.
AMD Ryzen Mobile Product Number Decoder (2023-2025)
AnandTech
Meaning values first digit Model Year 7 = 20238 = 20249 = 2025 Second digit Main market segment 1 = Athlon Silver2 = Athlon Gold3 and 4 = Ryzen 35 and 6 = Ryzen 57 = Ryzen 78 = Ryzen 7 or Ryzen 99 = Ryzen 9 Third digit Architecture 1 = Zen/Zen+2 = Zen 23 = Zen 34 = Zen 45 = Zen 5 Fourth Digit Feature Isolation/Lower Performance Segment 0 = Lower Segment5 = Upper Segment Suffix TDP/HX Form Factor = 55W+HS = 35W – 54WU = 15 – 28W = 9WC = Chromebook (15 – 28W)
If you take away only two things from AMD’s new product naming system, the most important is that the leading digit is the model year and the third digit is the CPU architecture (eg Zen 4). The rest, while still important, are there to define segments and TDP, rather than the hardware itself.
As part of the product name refactoring, AMD has also revised its TDP classes for mobile CPUs. Starting with the Ryzen Mobile 7000 series, the H-series class has been retired. Instead, the HS series now covers 35W to 45W via configurable TDPs, making the H and HS series redundant in AMD’s eyes. In practice, we’ve seen AMD push their H(S) silicon above 45W in the past, and I wouldn’t consider that off the table in the future, but these are the official ranges for now.
For anyone who still needs a more powerful (and more energy-efficient) CPU, there’s the HX class. Which in this generation has moved away from being a high-TDP variant of AMD’s mobile silicon and instead uses AMD’s desktop silicon, similar to what Intel does with its own HX-series processors. Gavin has a great write-up on the new HX mobile SKUs in another piece, but in short, the 7045 series Dragon Range platform is the mobile variant of AMD’s Ryzen 7000 “Raphael” desktop silicon, chips and all.
For the rest, AMD offers a mix of monolithic mobile-focused processors in the rest of the Ryzen Mobile 7000 family, including the new Phoenix platform in the 7040 series.
The line ends up similar to the previous one, with Dragon Range dedicated to DTR-class laptops. Meanwhile, AMD’s most portable laptops will use Phoenix for the fastest and most expensive laptops, with the refreshed Zen 3-based Rembrandt-R (7035 series) and Barcelo-R (7030 series) platforms offering cheaper options, especially with the 15-28. U-series Watt parts. Finally, AMD’s Zen 2 Mendocino platform, which was introduced late last year and is already part of the Ryzen Mobile 7000 family, will continue to be its budget PC offering thin and light laptops.
And here’s what it looks like in terms of hardware features/specs:
The end result is that, among its various sub-series, the Ryzen Mobile 7000 family is made up of a mix of 3 different Zen CPU architectures, built on 4 different process nodes. This is, for better or for worse, a very dense stack of mobile products, with 19 different consumer SKUs and another 3 PRO SKUs, covering a wide range of features and performance levels, as well as price ranges. So between all these configurations, AMD should have something for everyone in its 2023 mobile CPU product stack.
Ryzen Mobile 7040 series: Zen 4 Phoenix takes flight
The only truly new silicon among today’s announcements, and possibly the main announcement among mobile announcements tonight, is AMD’s Ryzen Mobile 7040 series of chips. Codenamed Phoenix, this is AMD’s first mobile-focused Zen 4 CPU design, incorporating the new CPU cores, new RDNA 3 iGPU and more in a new monolithic silicon die. AMD initially teased Phoenix at its 2022 Financial Analyst Day, so while tonight’s announcement wasn’t entirely unexpected, it’s been long-awaited as the architecture of Zen 4 CPU was designed to shine on mobile as much as it does on the desktop.
Phoenix is AMD’s flagship mobile silicon and the successor to the Zen3+/RDNA2 Rembrandt, which was the basis of the Ryzen Mobile 6000 series. Iterating on top of that design, Phoenix incorporates up to 8 of the new Zen 4 CPU cores from AMD, as well as a high-performance integrated GPU based on AMD’s new RDNA 3 architecture, with up to 12 CUs.
All of this, in turn, is being built on top of TSMC’s 4nm process, making it AMD’s most advanced piece of silicon to date, eclipsing even the Ryzen 7000 and Radeon RX 7000-based families 5nm What’s not to oversell 4nm here: TMSC’s 4nm nodes are all variants of their 5nm process, and we don’t know which specific variant AMD is using here, but it does emphasize that AMD has spared no expense here by using the best (and most ). expensive) process node they could get their hands on for their next generation mobile CPU.
Along with the new CPU and GPU architectures, Phoenix also marks another first for an AMD processor: it’s the first AMD CPU to incorporate an AI engine/neural network processor.
Adopting its XDNA architecture developed by Xilinx, AMD has placed an XDNA processing block, which they have dubbed Ryzen AI, into Phoenix to accelerate the processing of AI workloads. As with other dedicated silicon blocks/domain-specific accelerators, the inclusion of an AI engine is both to provide better performance at a specific task and to execute that task in a much more power-efficient manner. This, AMD reasons, will give them a performance advantage, along with power efficiency in tasks that can take advantage of the Ryzen AI block.
AMD doesn’t quote performance figures for the AI engine at this point, so we don’t have any hard numbers to use for comparison. But AMD has told us that the block can process up to 4 streams of AI workloads and is 20% faster than the neural engine in Apple’s M2 SoC. Obviously, any Apple comparison will raise a few eyebrows, as they’re the player to beat in the laptop space right now, so it’ll be interesting to see how those claims play out once Phoenix starts shipping.
In any case, the Apple comparison is apt, as AMD is clearly taking a page from the mobile (phone) market playbook with the inclusion of a neural network processor. These AI processors have proven to be very effective in mobile devices, and with Moore’s Law slowing down, the tech industry as a whole is moving towards using more DSAs to maintain the performance and efficiency of the increasing performance So for some time now, it has been clear that it would only be a matter of time; and in 2023, AMD is in a great position to do so thanks to its Xilinx acquisition.
That said, these are clearly early days for an AI engine on a PC (Windows) laptop. The software ecosystem to make use of the hardware isn’t quite there yet, especially in the consumer space, so AMD is taking the first step by providing hardware for developers to start coding. AMD sees the Ryzen AI engine as a long-term investment, and we should expect it to appear in more AMD products over the next few years, as well as receiving hardware updates of its own. In the meantime, it will be more of a conversational feature until the consumer software is ready.
Moving on, AMD is initially releasing 3 Phoenix-based SKUs, which according to the new naming system, are part of the Ryzen Mobile 7040 series. These are all HS-series parts, meaning they have TDPs of 35W to 45W and while the 7040 series is also planned for the U series, AMD will push its first Phoenix chips into the most visible (and profitable) notebook market.
AMD Ryzen 7040 “Phoenix” 4nm mobile CPUs
AnandTech
C/T BaseFreq TurboFreq GPU Freq. GPU L3 Cache (MB) TDP HS Series 35W – 54W Ryzen 9 7940HS 8/16 4000 5200 RDNA 312 CUs 3000 32 35W – 54W Ryzen 7 3 802 8004000 3000 3000 – 54 W Ryzen 5 7640HS 6/12 4300 RDNA 380 CUs 2800 32 35W – 54W
For the 7040 launch, there is one SKU for the Ryzen 9, Ryzen 7 and Ryzen 5 segments. The Ryzen 9 7940HS and Ryzen 7 7840HS have all 8 CPU cores enabled and differ only in their clock speeds . Meanwhile, the Ryzen 5 part, the 7640HS, drops to 6 CPU cores, and we’re still waiting to hear from AMD how many RDNA 3 CUs are enabled (R5 parts usually have a half-enabled GPU).
At the high end, the 7940HS will be able to turbo as…