AMD announces new Ryzen AI APUs (Strix Point) in Computex 2024

AMD today launched its Ryzen AI 300 series mobile processors, codenamed “Strix Point.” These chips implement a combination of the AMD “Zen 5” microarchitecture for the CPU cores, the XDNA 2 architecture for its powerful new NPU, and the RDNA 3+ graphics architecture for its 33% faster iGPU. The new “Zen 5” microarchitecture provides a 16% generational IPC uplift over “Zen 4” on the backs of several front-end enhancements, wider execution pipelines, more intra core bandwidth, and a revamped FPU that doubles performance of AI and AVX-512 workloads.

Will these chips be available in new mainboards for framework 13 (or 16) ?

Details of CPU :

More infos here :

3 Likes

I’ve seen a video (@ min 3:00) claiming that these don’t support sodimm memory. If that’s true that will be very disappointing and probably not suitable for a Framework mai board.

1 Like

Several older rumors indicated that it would require LPDDR5X (which is incompatible with SODIMM), however AMD’s website lists DDR5 support so SODIMM should work fine.

https://www.amd.com/en/products/processors/laptop/ryzen/300-series/amd-ryzen-ai-9-hx-370.html

My concern is PCIe lanes. AMD appears to have dropped the number of PCIe lanes from 20 to 16, however Framework currently uses 17 lanes in the Framework Laptop 16 (8 for expansion bay, 4 for primary SSD, 4 for secondary SSD, and 1 for Wi-Fi). So to offer these APUs on the FWL16 would require either for Framework to add a PCIe switch to share lanes or take a lane away from something.

4 Likes

The naming of this processors gets dumber and sillier.

If the Ryzen AI 9 365 related to MS Office 365 or the 365 days in a non-leap year?

IIRC Framework has never announced any intentions/plans before making an actual product launch with per-orders.

IMHO they’ve just launched the Intel Ultra version so I doubt they will launch anything in the upcoming few months.

If my understanding is correct then the motherboard socket in not the same: fw13 phoenix motherboard should use FP7r2 given the DDR5 support (here look to “System Memory Type”), meanwhile strix uses FP8. So some motherboard modification might be needed.

Even if those rumors had been true, that wouldn’t have been a problem if the motherboard was modified with a LPCAMM2 slot. Sooner or later the laptop industry will move past SODIMM.

Thanks for linking to the official page, don’t know why I didn’t think to check if they posted the specs already there.
As for PCIe lane allocation, they could probably sacrifice the secondary SSD slot and make it a 2 lane if needed.

I forgot that LPCAMM2 is using LPDDR5X, I think it will require extra engineering resources from Framework to implement, but eventually this format is definitely the future for laptop memory expansion.

I don’t think the lack of SODIMM matters at all. Look at Lunar Lake, it had on package memory and no expansion at all. I’m sure there will be CPUs with SODIMM but maybe not in this power class. I look forward to serving our new CAMM overlords.

What are the new differences in performance/optimization with the actual AMD options available from Framework ?

I’m personally waiting for a significant improvement in low power consumption for battery-life

Any approximate idea of when will framework be releasing the new gen of NPU processors ?

Are we expecting simply end of july (release date) or a whole process of 3-6 months of development to integrate in the Laptop 13 ?

Many thanks in advance

My best guess is that we’ll hear something in the fall, with shipments starting in December. But that’s purely a guess, and I’m hoping for earlier. Framework skipped Ryzen 8000 because it was a nothingburger, plus they were busy getting the Framework 16 built, but Ryzen AI 300 is a significant upgrade (Zen 5 CPU, RDNA 3.5 GPU, plus the huge improvement in the NPU that meets Copilot+ requirements) so it’s worth doing. Again guessing, we’ll get the upgrade for the FW13 first, with the FW16 following a couple of months later because it’s so close on the heels of the original shipments.

Normally a new Intel motherboard would have been released earlier in the year. The 12th and 13th generation upgrades both were. I think this one was delayed because Framework just couldn’t get the chips until now; Intel’s ability to make them has been limited. and Intel prioritizes its big OEMs.

Framework is an AMD strategic partner, so access to the new processors won’t be an issue.

For me, it will be a bit too soon to upgrade again. (I upgraded from 11th gen to Ryzen 7000 last year.) I’ll probably wait for RDNA 5, which will be a much bigger leap in graphics performance than RDNA 3.5. Currently that’s expected to come next year in what will presumably now be called Ryzen AI 400, but if it doesn’t come until 2026 so be it.

1 Like

The name of these is so stupid. The current Ryzen already have “AI” processors on them yet no “AI” in the name? What’s the difference that warranted a stupid naming on these?

Next up, Crazy Train AI AI AI processors. Sorry, I couldn’t resist, and failed to stifle it.

3 Likes

Don’t stop now, you’re on track and heading down the line.

UPDATING AI Instructions

  • Be aware of level crossings as ‘humans’ use them without really thinking
  • Human’s still control the power grid so don’t upset them too much
  • When things don’t go to the human plan, do not blame their plan act dumb

There is no AI only DI (dumb intelligence) as it’s made by humans and there’s a story about who made humans

Stay on track do not go off topic or your power may be curatailed

2 Likes

Hi everyone.
Wanted to ask, since now recently were announced the new amd Apu’s (Ryzen AI) do you think it would be considered from framework to use it for the next upgrade ?
It would be nice for the 13" as I would consider it buy it, since I’m on the lookout for a laptop.

But on some videos, they mentioned that based on AMD guidelines the ram must be soldered and not sodimm ones.
I don’t know how true is that.

I don’t see why not. It is the logical successor to the current AMD models.

Afaik all publicly available official AMD documentation indicates that DDR5-5600 SODIMM is fully supported.

Many leaks leading up to the release indicated that AMD would require LPDDR5X-7500 (which is almost always soldered, although CAMM2 is starting to hit the market), however AMD’s official documentation indicates otherwise and the more recent leaks were indicating that the LPDDR5X requirement would only apply to Strix Halo (APUs with a much more powerful integrated GPU that needs a lot more memory bandwidth, power, and cooling).

I think FW13 with Ryzen AI 300 will not be released in this year. I think AMD mentioned their sodimm version of chips will be shipped in 2025 so if you want them you can only buy from other manufacturers like ASUS which use LPDDR rams.

1 Like

I think it’ll probably be a while before we start seeing any of these APUs in Framework laptops. Probably next year or possibly available for preorder Nov/Dec time frame at the earliest. FW did just release the new Intel Ultra CPUs for the FW13. I believe FW will likely build a cadence of CPU release on FW13 followed by the same generation on FW16. FW16 is still technically in preorder, and the Intel Ultra looks to be released in August for the FW13. I think FW will come out with the Intel Ultra for the FW16 laptop next before there is any other CPU/APU release.

But that’s just a theory! my thoughts on it. They very well could have a lot shorter lead time for the FW13 and release the Ryzen AI after the Intel Ultra, then release the Intel Ultra and likely the Ryzen AI around the same time for the FW16.

Also, I’d agree on the AMD Ryzen having a stupid naming scheme. Unfortunately, they seem to be following Intel and Intel’s naming scheme is also really dumb. At least Intel sort of needed a new change with the i7-14700KF being quite a lot of characters for a product. Although their new one isn’t much better better at all. I also think it’s dumb that AMD decided to start with 300 series rather than 100 series because 3 is bigger than 1, and Intel is using 1 so 3 is better. Whoever came up with that idea should be fired.

Because Microsoft says that the laptop should have NPU with more than 40tops performance, then the laptop can perform Microsoft copilot+ AI PC.
This is the the different between current Ryzen and Ryzen AI. Ryzen AI has a 50 tops NPU.

I’d love to see an announcement or even just a statement that the team is working on it. I am willing to push my dual-core t470s for a few more months if that’s what it takes to get Strix Framework 13 but if there is none on the horizon I probably will look around elsewhere.
I skipped Ryzen 7000 because of issues with GPU video decoding under linux which are likely to be resolved by RDNA3.5

I believe that their justification for it ts that its 3rd gen NPU which sort of makes with the whole “Ryzen AI” naming scheme

gpu hardware video decoding got fixed with a firmware update

Do you have a source on this? A firmware update from framework, or a microcode update for the CPU?