Tuesday, January 6th 2026

AMD Expands Ryzen AI MAX "Strix Halo" Processor Lineup

AMD today expanded its Ryzen AI MAX "Strix Halo" line of super-APUs with two new processor models, the Ryzen AI Max+ 392 and Ryzen AI Max+ 388. Both are based on the existing "Strix Halo" package, which combines a 16-core/32-thread "Zen 5" CPU based on standard "Zen 5" CCDs, with a large I/O die featuring a powerful iGPU based on RDNA 3.5 graphics architecture that has 40 compute units (2,560 stream processors), a Copilot+ ready NPU with 50 TOPS on tap, and a unified 256-bit LPDDR5X memory interface.

The Ryzen AI Max+ 392 is configured with a 12-core/24-thread CPU using two CCDs, clock speeds of 3.20 GHz base and 5.00 GHz maximum boost, a full-featured iGPU with all 40 CU enabled, a 50 TOPS NPU, and a configurable TDP range of 45 W to 120 W. Then there's the Ryzen AI Max+ 388, featuring an 8-core/16-thread CPU using a single CCD, which ticks at 3.60 GHz base with 5.00 GHz boost, again, the fully unlocked iGPU with all 40 CU enabled, the 50 TOPS NPU, and 45 W to 120 W cTDP. The key difference with these two SKUs over the others in the lineup within their brand extension, is the focus on increased graphics power at the expense of CPU core count, which should make these formidable gaming chips at lower price-points.
Add your own comment

12 Comments on AMD Expands Ryzen AI MAX "Strix Halo" Processor Lineup

#1
kondamin
and these should really have an x3d version
Posted on Reply
#2
T1beriu
kondaminand these should really have an x3d version
No it should. X3D only makes sense with powerful GPUs, not to bottleneck them. How many 4060 GPUs did you see paired with top of the line X3D CPU? None. And when you'll see that, it will be because the builder had no clue what he was doing.
Posted on Reply
#3
Darmok N Jalad
The 8C model looks the most interesting to me. It's the closest thing to the current generation of consoles with the fully-enabled iGPU. Seems like the GPU would get more resources versus the 16C and 12C models, and be cheaper to boot.
Posted on Reply
#4
kondamin
T1beriuNo it should. X3D only makes sense with powerful GPUs, not to bottleneck them. How many 4060 GPUs did you see paired with top of the line X3D CPU? None. And when you'll see that, it will be because the builder had no clue what he was doing.
plenty of things will work better with the extra cache, say you love playing factorio or other like large rts games those run fine on a 4060
Posted on Reply
#5
RinSenpai
T1beriuNo it should. X3D only makes sense with powerful GPUs, not to bottleneck them. How many 4060 GPUs did you see paired with top of the line X3D CPU? None. And when you'll see that, it will be because the builder had no clue what he was doing.
I have a friend who has a 7800X3D + RX 7600 build, but he plays mostly MMORPGs.

Ryzen AI MAX is already a big chip, which is necessary for its iGPU, so I doubt they could find more space for 3D V-Cache without increasing size further anyways.

In any case, the only variable for success here is the price. Current Ryzen AI MAX prices almost everyone out in the laptop market, regardless of the use you have of it.
Just for comparison, in France you could get a top-end RTX 5070 Ti laptop or a 5080 laptop for about the same price as these 14" ASUS ROG Flow with 32GB RAM, and these are more powerful (in literally every way, including higher power consumption :laugh: ) and easier to keep cool, and despite being slightly less portable, they're coming with bigger monitors than these hybrids of laptops-tablets.
And if you're into "more RAM", you are paying at least €4k in order to get these 128GB of RAM for the Ryzen AI MAX, that's 5090 laptop pricing and adding 300-500 bucks get you 64GB of RAM on these gaming laptops.


(+ the fact LDLC is damn expensive, that also doesn't help Ryzen AI since their prices never moved an inch, while some other websites/retailers can generally offer cheaper prices for gaming laptops but not for Ryzen AI)


If your goal is to play on a bigger monitor and not move at all, just buy these Nvidia-GPU laptops.
If you're into AI, chances are that it'll run better on these Nvidia-GPU laptops, unless you use a model that needs more RAM+VRAM than these 64GB RAM + 5090 laptops (and that's a total 80GB if we exclude the base Windows RAM usage which sits between 4-8GB).
If you're in productivity, depending on the software suite you're using/wanting to use, even a Macbook actually makes sense.
Battery life is the new battleground and pretty much everyone has a product to claim the highest battery life today; performance on battery being another matter, but they generally all suck equally there if we consider 30-150 minutes battery life to be bad.
AMD is behind everyone else on software support (except on Linux where they're just barely behind instead of way behind) and it doesn't look like they're in a hurry to release their productivity API suite.

Only price will determine if these new Ryzen AI MAX are going to be a success or not.
Posted on Reply
#6
ToTTenTranz
RinSenpaiRyzen AI MAX is already a big chip, which is necessary for its iGPU, so I doubt they could find more space for 3D V-Cache without increasing size further anyways.
The 3D V-Cache sits below the CCD. It wouldn't occupy more space, other than like a couple cubic millimeters. The "V" stands for "Vertical".

Whether or not Strix Halo's IOD and CCDs support VCache is another discussion (especially as the CCDs themselves are different from the regular Fire Range / Granite Ridge ones), but volume wouldn't be an issue.
RinSenpaiIn any case, the only variable for success here is the price. Current Ryzen AI MAX prices almost everyone out in the laptop market, regardless of the use you have of it.
It does look like these 388 and 392 SKUs are made to justify the entrance of a big iGPU SoC in the PC market for a lower price.
Posted on Reply
#7
TheinsanegamerN
A 388+ with x3d cache would make such a sweet chip for mini PCs or laptops. I'm really disappointed that laptops are non existent except for the overpriced HP machine.
T1beriuNo it should. X3D only makes sense with powerful GPUs, not to bottleneck them. How many 4060 GPUs did you see paired with top of the line X3D CPU? None. And when you'll see that, it will be because the builder had no clue what he was doing.
Even the low end 2CU iGPU in the x3d desktop chips sees a performance improvement from 3d cache. rDNA 2/3/4 "infinity cache" serves the same purpose. AMD iGPUs can access the shared L3 cache so just like any other processor, having a much larger pool of cache can help tremendously with performance, especially in memory limited scenarios which iGPUs always struggle with.
Posted on Reply
#8
TheGeekn°72
kondaminand these should really have an x3d version
you wished, too bad it gets mobile level cache...
T1beriuNo it should. X3D only makes sense with powerful GPUs, not to bottleneck them.
I mean, my 9600X has nearly twice the cache these things have...
Darmok N JaladThe 8C model looks the most interesting to me. It's the closest thing to the current generation of consoles with the fully-enabled iGPU. Seems like the GPU would get more resources versus the 16C and 12C models, and be cheaper to boot.
makes me wonder if they made that with the intent to put it into the XSX2/PS6? 24GB of unified memory and they could call it a day pretty much, that would be solid imo
RinSenpaiRyzen AI MAX is already a big chip, which is necessary for its iGPU, so I doubt they could find more space for 3D V-Cache without increasing size further anyways.
Good thing VERTICAL-cache doesn't need more space on the package, only an extra layer of silicon.
TheinsanegamerNA 388+ with x3d cache would make such a sweet chip for mini PCs or laptops. I'm really disappointed that laptops are non existent except for the overpriced HP machine.

Even the low end 2CU iGPU in the x3d desktop chips sees a performance improvement from 3d cache. rDNA 2/3/4 "infinity cache" serves the same purpose. AMD iGPUs can access the shared L3 cache so just like any other processor, having a much larger pool of cache can help tremendously with performance, especially in memory limited scenarios which iGPUs always struggle with.
man, a 388X3D would go so damn hard...
Posted on Reply
#9
T1beriu
TheGeekn°72I mean, my 9600X has nearly twice the cache these things have...
That's false. Strix Halo uses a slightly modified Zen 5 die. Both have exactly the same cache.
kondaminplenty of things will work better with the extra cache, say you love playing factorio or other like large rts games those run fine on a 4060
When you said plenty you actually meant like 5 things, right?
A 388+ with x3d cache would make such a sweet chip for mini PCs or laptops. I'm really disappointed that laptops are non existent except for the overpriced HP machine.
Even the low end 2CU iGPU in the x3d desktop chips sees a performance improvement from 3d cache. rDNA 2/3/4 "infinity cache" serves the same purpose. AMD iGPUs can access the shared L3 cache so just like any other processor, having a much larger pool of cache can help tremendously with performance, especially in memory limited scenarios which iGPUs always struggle with.
TheinsanegamerNA 388+ with x3d cache would make such a sweet chip for mini PCs or laptops. I'm really disappointed that laptops are non existent except for the overpriced HP machine.

Even the low end 2CU iGPU in the x3d desktop chips sees a performance improvement from 3d cache. rDNA 2/3/4 "infinity cache" serves the same purpose. AMD iGPUs can access the shared L3 cache so just like any other processor, having a much larger pool of cache can help tremendously with performance, especially in memory limited scenarios which iGPUs always struggle with.
You confusing infinity cache to X3D cache and you made some pretty wild and inaccurate claims. Please prove everything you said.
Posted on Reply
#10
kondamin
T1beriuWhen you said plenty you actually meant like 5 things, right?
A 388+ with x3d cache would make such a sweet chip for mini PCs or laptops. I'm really disappointed that laptops are non existent except for the overpriced HP machine.
Even the low end 2CU iGPU in the x3d desktop chips sees a performance improvement from 3d cache. rDNA 2/3/4 "infinity cache" serves the same purpose. AMD iGPUs can access the shared L3 cache so just like any other processor, having a much larger pool of cache can help tremendously with performance, especially in memory limited scenarios which iGPUs always struggle with.
when i say plenty of things i mean things that have large datasets or just benefit from large victim caches.
That is a lot more than just 5 things.

it could even improve battery life
Posted on Reply
#11
TheGeekn°72
T1beriuThat's false. Strix Halo uses a slightly modified Zen 5 die. Both have exactly the same cache.
Hmmmm, I just checked AMD's product DB and you're right, I think the third party article I've read then had a serious error, I'll go and report it if I find it again

Well shit, I was under the assumption all this time that Halo had regular laptop cache...
Now that I think about it, how come laptop chips have gimped cache capacity? At first I assumed it was cost savings but since that cache is integrated in the CCD, that doesn't make that much sense...
Posted on Reply
#12
kondamin
TheGeekn°72Hmmmm, I just checked AMD's product DB and you're right, I think the third party article I've read then had a serious error, I'll go and report it if I find it again

Well shit, I was under the assumption all this time that Halo had regular laptop cache...
Now that I think about it, how come laptop chips have gimped cache capacity? At first I assumed it was cost savings but since that cache is integrated in the CCD, that doesn't make that much sense...
isn't that because ryzen ai has mixed cores zen and zen c
Posted on Reply
Add your own comment
Nov 17th, 2026 13:14 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts