Monday, December 22nd 2025

Lenovo Legion Gaming CES Line-Up Tips AMD Ryzen AI 400 Refresh

In our reviews of previous Lenovo Legion laptops, we've noted that they deliver a lot of bang for the buck, with devices like the Legion 5i being a great example. Now, it looks like Lenovo is planning to launch an update to its Legion gaming laptop series at CES in January 2026. Recent leaks from Windows Latest indicate that Lenovo is planning on launching AMD-powered Legion 5a and Legion 7a laptops, all of which will feature updated AMD Ryzen AI 400-series APUs and NVIDIA GeForce RTX 5060 Laptop GPUs. Prior leaks about the AMD Ryzen 400 APUs revealed that the next-gen APUs will pack both Zen 5 and RDNA 3.5 tech, with the Ryzen 9 APUs featuring up to 12 cores, while the leaked Ryzen 7 variants both packed 8 cores.

The leak confirms Legion 7a laptops equipped with the AMD Ryzen AI 9 HX 470 and Ryzen AI 9 465 APUs, while the 5a laptops will come with either the Ryzen AI 9 465 or Ryzen 7 250. The Legion 7a and the Ryzen 400-series Legion 5a will lead the line-up with 16-inch, 2,560 × 1,600, 16:10 OLED displays topping out at 240 Hz (165 Hz for the Legion 5a) and 500 nits brightness. It seems as though only the Lenovo Legion 7a will feature NVIDIA G-SYNC, while the entire line-up will deliver 100% DCI-P3 coverage. All except for the Ryzen 7-powered Legion 5a will feature X-Rite calibration from the factory. The top-end Legion 7a will come with up to 64 GB of LPDDR5x RAM at 8533 MT/s, while the lower-end Legion laptops will have slower DDR5-5600 RAM and top out at 32 GB. All three laptops will seemingly also feature keyboards with 24-zone RGB and swappable keycaps, which is interesting for those interested in longevity and repairability.
Source: Windows Latest
Add your own comment

5 Comments on Lenovo Legion Gaming CES Line-Up Tips AMD Ryzen AI 400 Refresh

#1
Chaitanya
AMD has the stupidest naming scheme and the most convoluted product stack when it comes to mobile which makes Intel's offering look quite sane.
Posted on Reply
#2
Squared
Years ago, AMD kept using Vega in iGPUs years after RDNA was released. I assumed then AMD was too poor to afford updating the iGPU and perhaps also wanted the faster GPGPU abilities of Vega. But none of that is true today; AMD has money and RDNA4 beats RDNA3 in every category except transistor count. Moreover during the 7040 series to 8040 series refresh, AMD actually did make new silicon with a bigger NPU. So why is Ryzen 400 launching with RDNA3.5 when Core 300 is launching at the same time with shiny new Xe3?
Posted on Reply
#3
ToTTenTranz
SquaredMoreover during the 7040 series to 8040 series refresh, AMD actually did make new silicon with a bigger NPU.
The NPU didn't get bigger from Phoenix to Hawk Point, it was simply clocked higher. It's pretty much the same footprint with a new stepping, just like this transition from Ryzen 300 to 400 series (Strix Point to Gorgon Point).
SquaredSo why is Ryzen 400 launching with RDNA3.5 when Core 300 is launching at the same time with shiny new Xe3?
One could say AMD cancelled RDNA4 iGPUs because the Zen6 Medusa line is just "around the corner" (probably not here until very late 26 / early 27) and that one will have APU variants with RDNA3.5 and RDNA5 iGPUs.


Though the actual reason is just that AMD's execution on APUs became pretty poor because they're laser-focused on riding the AI wave, which is where their investors want them to put their energy on as that's where all the infinite growth is.
Had they been under the same pressure as Intel to catch up to reach profitability and a foot on the ULP APU market, we'd already have RDNA4 iGPUs with Infinity Cache on the market at this point.
Posted on Reply
#4
Squared
ToTTenTranzThe NPU didn't get bigger from Phoenix to Hawk Point, it was simply clocked higher. It's pretty much the same footprint with a new stepping, just like this transition from Ryzen 300 to 400 series (Strix Point to Gorgon Point).




One could say AMD cancelled RDNA4 iGPUs because the Zen6 Medusa line is just "around the corner" (probably not here until very late 26 / early 27) and that one will have APU variants with RDNA3.5 and RDNA5 iGPUs.


Though the actual reason is just that AMD's execution on APUs became pretty poor because they're laser-focused on riding the AI wave, which is where their investors want them to put their energy on as that's where all the infinite growth is.
Had they been under the same pressure as Intel to catch up to reach profitability and a foot on the ULP APU market, we'd already have RDNA4 iGPUs with Infinity Cache on the market at this point.
TPU said the NPU was the only part not refreshed. 10 TOPS to 16 sounds like a lot for just a clock increase. But it's also kind if rediculously small next to Microsoft's 40 TOPS requirement so maybe it is just a much faster clock speed.
www.techpowerup.com/316436/amd-ryzen-8040-series-hawk-point-mobile-processors-announced-with-a-faster-npu
Posted on Reply
#5
ToTTenTranz
SquaredTPU said the NPU was the only part not refreshed. 10 TOPS to 16 sounds like a lot for just a clock increase. But it's also kind if rediculously small next to Microsoft's 40 TOPS requirement so maybe it is just a much faster clock speed.
www.techpowerup.com/316436/amd-ryzen-8040-series-hawk-point-mobile-processors-announced-with-a-faster-npu
It's probably more like the original NPU implementation was never meant for running local LLMs and 10TOPs was deemed more than enough.
Microsoft's NPU demands were originally for more trivial stuff like face detection for auto login on windows or beauty filters on Teams, but on a lower power budget. The Steam Deck's APU does have a functional NPU (originally planned for a 2021 Microsoft device) and at the time AMD used to call it a "CVML" block, which stands for Computer Vision Machine Learning (there's even a device out there that makes heavy use of it, Magic Leap 2). LLMs were not on Microsoft's radar at the time.


That ship has sailed and Microsoft has now shifted to running big models locally that send all your activity to their servers, and that's why AMD's APUs with RDNA5 are apparently getting rid of the NPU and getting bigger tensor units within the GPU's shaders.
Posted on Reply
Nov 20th, 2026 16:23 PST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts