AMD Ryzen AI MAX 300 "Strix Halo" iGPU to Feature Radeon 8000S Branding

[btarunr]

AMD Ryzen AI MAX 300-series processors, codenamed "Strix Halo," have been on in the news for close to a year now. These mobile processors combine "Zen 5" CPU cores with an oversized iGPU that offers performance rivaling discrete GPUs, with the idea behind these chips being to rival the Apple M3 Pro and M3 Max processors powering MacBook Pros. The "Strix Halo" mobile processor is an MCM that combines one or two "Zen 5" CCDs (some ones featured on "Granite Ridge" desktop processors and "Turin" server processors), with a large SoC die. This die is built either on the 5 nm (TSMC N5) or 4 nm (TSMC N4P) node. It packs a large iGPU based on the RDNA 3.5 graphics architecture, with 40 compute units (CU), and a 50 TOPS-class XDNA 2 NPU carried over from "Strix Point." The memory interface is a 256-bit wide LPDDR5X-8000 for sufficient memory bandwidth for the up to 16 "Zen 5" CPU cores, the 50 TOPS NPU, and the large 40 CU iGPU.

Golden Pig Upgrade leaked what looks like a company slide from a notebook OEM, which reveals the iGPU model names for the various Ryzen AI MAX 300-series SKUs. Leading the pack is the Ryzen AI MAX+ 395. This is a maxed out SKU with a 16-core/32-thread "Zen 5" CPU that uses two CCDs. All 16 cores are full-sized "Zen 5." The CPU has 64 MB of L3 cache (32 MB per CCD), each of the 16 cores has 1 MB of dedicated L2 cache. The iGPU is branded Radeon 8060S, it comes with all 40 CU (2,560 stream processors) enabled, besides 80 AI accelerators, and 40 Ray accelerators. The Ryzen AI MAX 390 is the next processor SKU, it comes with a 12-core/24-thread "Zen 5" CPU. Like the 395, the 390 is a dual-CCD processor, all 12 cores are full-sized "Zen 5." There's 64 MB of L3 cache, and 1 MB of L2 cache per core. The Radeon 8060S graphics solution is the same as the one on the Ryzen AI MAX+ 395, it comes with all 40 CU enabled.



We now get to what will probably go down as the most popular processor SKU from the "Strix Halo" series, the Ryzen AI MAX 385. This processor comes with an 8-core/16-thread "Zen 5" CPU using a single CCD, all 8 of these are full-sized "Zen 5," share a 32 MB L3 cache, and have 1 MB of L2 cache per core. The iGPU is branded Radeon 8050S, and is configured with 32 CU (2,048 stream processors), 64 AI accelerators, and 32 Ray accelerators.

Lastly, there's the Ryzen AI MAX 380, the entry-level to this series. It comes with just a 6-core/12-thread CPU using a single CCD, but all six of these are full-sized "Zen 5," and have 32 MB of L3 cache on tap, besides 1 MB of L2 cache per core. AMD hasn't decided on what to call the iGPU solution, but it's configured with just 16 CU—the same number as the iGPU on the "Strix Point" monolithic silicon. Perhaps the advantages the Ryzen AI MAX 380 offers over the Ryzen AI 9 HX 370 is the fact that it has more full-sized "Zen 5" cores. The HX 370 has a 12-core/24-thread CPU, but 8 out of 12 of these are low-power "Zen 5c" cores that run in a lower frequency band than the "Zen 5" cores, and have a lower L3 cache share in their CCX.

We're not sure if all four processor SKUs get the 256-bit LPDDR5X memory interface, it's possible that the entry-level Ryzen AI MAX 380 comes with a narrower memory interface like 128-bit or 192-bit.

View at TechPowerUp Main Site | Source

 

[AleXXX666]

yeah, Ultra and Max namings in cpus, that's marketing we deserved

 

[_roman_]

It'S not like NVIDIA did not use MAX before, right?

 

[Kohl Baas]

yeah, Ultra and Max namings in cpus, that's marketing we deserved

"It just works!"

 

[Hardware1906]

MAX/PRO for CPU
and 80x0 for GPU????

we are going to have the 8060 before the 5060

 

[GoldenX]

Dear god, the naming.
Wish they fix the core latency this time, and don't starve that iGPU with only a 128-bit wide memory bus.

 

[The Quim Reaper]

I can't wait to see these....in all 2 laptops that will ever get them.

AMDs Laptop dGPU efforts have been a complete fail for the last decade, not so much their silicon performance but their ability to convince OEMs that they will ever be able to supply them with enough chips to make it worthwhile them going big on them.

 

[Hakker]

I can't wait to see these....in all 2 laptops that will ever get them.

AMDs Laptop dGPU efforts have been a complete fail for the last decade, not so much their silicon performance but their ability to convince OEMs that they will ever be able to supply them with enough chips to make it worthwhile them going big on them.

Not true it's just that their dGPU isn't geared towards OEMs at all and for a good reason. There is no market for it. At least 95% of all OEMs are work machines they are more than well enough with the iGPU on them. The Discrete GPU market is already a small one the Discrete GPU on an OEM is absolutely tiny. AMD much rather goes for the console market with 100 million guaranteed sales with just 2 chips. Than going for a 8 chips 1M market at best. I went and still go to LAN parties some rather big and over the last decade of people that had a gaming laptop I could count on both my hands compared the couple of thousand who brought their normal PC. And that's just about the biggest group you could target with a gaming laptop.

dGPU's are a niche in a niche market. sure Nvidia sells the 3050 and 4050 with somewhat okayish numbers but that's literally it.

 

[TumbleGeorge]

Radeon 8060S

Wait to be added to GPU database in TPU with it's smaller variants too.

 

[A&P211]

It'S not like NVIDIA did not use MAX before, right?

I love the MAX name, the name is so SUPER. Nvidia will make lots of money with that (TI)tle.

 

[ymdhis]

Dear god, the naming.
Wish they fix the core latency this time, and don't starve that iGPU with only a 128-bit wide memory bus.

Nearly all leaks suggested 256bit memory so far.

I just want this to come to AM5 socket, even if it's 128 bit only.

 

[Daven]

Not true it's just that their dGPU isn't geared towards OEMs at all and for a good reason. There is no market for it. At least 95% of all OEMs are work machines they are more than well enough with the iGPU on them. The Discrete GPU market is already a small one the Discrete GPU on an OEM is absolutely tiny. AMD much rather goes for the console market with 100 million guaranteed sales with just 2 chips. Than going for a 8 chips 1M market at best. I went and still go to LAN parties some rather big and over the last decade of people that had a gaming laptop I could count on both my hands compared the couple of thousand who brought their normal PC. And that's just about the biggest group you could target with a gaming laptop.

dGPU's are a niche in a niche market. sure Nvidia sells the 3050 and 4050 with somewhat okayish numbers but that's literally it.

I would add that Strix Halo and Apple M series will make the mobile dGPU market even smaller to the point of irrelvance.

 

[bitsandboots]

AMD is getting to USB levels of naming nonsense.
The 7xxx CPU mobile naming convention was already deliberately, maliciously misleading but now we have an (AI) 3xx series which has 8xxx GPU. Too much to keep track of.

 

[tpuuser256]

2.5x the CUs of 890M and 8000MHz LPDDR5x. That will hit hard baby

 

[T1beriu]

I just want this to come to AM5 socket, even if it's 128 bit only.

How would that even work?!

 

[Neo_Morpheus]

This is my biggest issue with them, they dont respect their own naming convection.

The only way that I would use the name Radeon 8000 is if its based on RDNA4, otherwise, this should be a 7000.

 

[mate123]

So when Intel makes weird naming scheme AMD feels compelled to double down?

 

[igormp]

Nearly all leaks suggested 256bit memory so far.

I just want this to come to AM5 socket, even if it's 128 bit only.

Don't think it'll be possible due to the IOD size, and it likely only supports LPDDR5X, so no regular DDR5 support.
Going for half of the memory bandwidth would also starve the iGPU pretty bad, making the extra CUs kinda moot.

How would that even work?!

I guess by disabling half of the memory controllers, making them support regular DDR5, and having all of that fit withing the regular AM5 socket package? Not sure if it's doable, but it doesn't sound impossible.

 

[anonuser57]

So when Intel makes weird naming scheme AMD feels compelled to double down?

No, it's that everyone has to copy Apple. Some just joined the bandwagon sooner than others.

 

[TheinsanegamerN]

Nearly all leaks suggested 256bit memory so far.

I just want this to come to AM5 socket, even if it's 128 bit only.

it would be totally pointless on AM5, because a 128 bit DDR5 bus cant come close to 256 bit LPDDR5X bus. Strix halo is going to have 9500 mhz memory. Socket AM5 can barely supply a third of that, the iGPU would be choked beyond belief. This is why the 8000g series exist.

Also, just get a 7600?

 

[mate123]

No, it's that everyone has to copy Apple. Some just joined the bandwagon sooner than others.

At least they are not calling the iPhone 17 as iPhone AI MAX 300 8000S (yet)

 

[T1beriu]

Don't think it'll be possible due to the IOD size, and it likely only supports LPDDR5X, so no regular DDR5 support.
Going for half of the memory bandwidth would also starve the iGPU pretty bad, making the extra CUs kinda moot.


I guess by disabling half of the memory controllers, making them support regular DDR5, and having all of that fit withing the regular AM5 socket package? Not sure if it's doable, but it doesn't sound impossible.

Doesn't make any sense. Cutting the memory bandwidth in half will destroy the performance benefits of having it 256-bit. Cut the bandwidth of a 4090 in half and you will see it perform as a 4070 Super. Any other brilliant ideas you've got?

 

[igormp]

Doesn't make any sense. Cutting the memory bandwidth in half will destroy the performance benefits of having it 256-bit. Cut the bandwidth of a 4090 in half and you will see it perform as a 4070 Super. Any other brilliant ideas you've got?

Let me help you to read my post once again:

Going for half of the memory bandwidth would also starve the iGPU pretty bad, making the extra CUs kinda moot.

I didn't say it was a good idea, just said it wasn't impossible.