AMD Ryzen AI 300 Pro Series Could be Equipped with up to 128 GB of Memory

[AleksandarK]

According to the leaked listing posted on X by user @Orlak29_, reports suggest that Pro versions of the AMD Ryzen 7 AI and Ryzen 9 AI are in the pipeline, with a potential game-changer in the form of the high-end "Strix Halo" model. The standout feature of the Strix Halo is its rumored support for up to 128 GB of RAM, a significant leap from AMD's current offerings. This massive memory capacity could prove valuable for AI workloads and data-intensive applications, potentially positioning AMD better against offerings from Intel and Qualcomm. Leaked diagrams hint at a unique design for the Strix Halo, featuring a chiplet layout reminiscent of a graphics card. The processor is reportedly surrounded by memory on three sides, enabling the massive 128 GB capacity.

While this top-tier model is expected to carry a premium price, it could find a ready market among professionals and enthusiasts demanding both raw processing power and extensive memory resources. On the performance front, rumors suggest the Strix Halo will boast up to 16 Zen 5 cores and a GPU with 40 Compute Units based on RDNA 3.5 architecture. This combination might rival the performance of high-end mobile GPUs like the RTX 4060 or even the RTX 4070 for laptops.
As with previous generations, AMD is expected to release Pro versions of these processors with additional features like ECC memory support.



View at TechPowerUp Main Site | Source

 

[JWNoctis]

Remains to be seen: Would it see even more RAM bandwidth from the CCDs than, say, a non-Pro Threadripper 7000 with octa-channel regular DDR5, what the latency would look like, and would it cost much less, presumably in a premium portable form factor.

 

[hsew]

If I had a nickel for every time a tech publication predicted an upcoming AMD iGPU to match nVidia’s current X060 GPU…

 

[dj-electric]

If I had a nickel for every time a tech publication predicted an upcoming AMD iGPU to match nVidia’s current X060 GPU…

These very much exist, they just come in weirdly shaped boxes and sold to gamers with locked operating systems and minimal IO.

 

[hsew]

These very much exist, they just come in weirdly shaped boxes and sold to gamers with locked operating systems and minimal IO.

Right… today’s PS5 and Xbox SOCs are blessed with 6-8 times the memory bandwidth possible from even today’s fastest LPDDR5 configurations, as well as 10-15 times the power budget allowed to the GPU alone. While *technically* an iGPU, it’s really not in the same class (or market) as the aforementioned laptop SOCs…

 

[Wirko]

If I had a nickel for every time a tech publication predicted an upcoming AMD iGPU to match nVidia’s current X060 GPU…

You should have been buying "AMD" at Nasdaq, not Newegg.

 

[hsew]

You should have been buying "AMD" at Nasdaq, not Newegg.

Just to clarify… My comment was more aimed towards the tech press as opposed to AMD. I just think it’s absurd to suggest an upcoming laptop iGPU, which has to *share* a lower power budget as well as comparatively scarce memory bandwidth with multiple CPU cores, is somehow going to perform on-par with any current midrange dGPU… which in contrast would have its own independent memory subsystem and power budget. 4060-class performance is an extremely optimistic forecast to put it mildly.

 

[oxrufiioxo]

Just to clarify… My comment was more aimed towards the tech press as opposed to AMD. I just think it’s absurd to suggest an upcoming laptop iGPU, which has to *share* a lower power budget as well as comparatively scarce memory bandwidth with multiple CPU cores, is somehow going to perform on-par with any current midrange dGPU… which in contrast would have its own independent memory subsystem and power budget. 4060-class performance is an extremely optimistic forecast to put it mildly.

People really want an APU like a series X or a PS5/PS5 pro that's untill they see they'd need to be on proprietary boards with ram stuck onto the board to get anywhere near equal performance. Also likely a $$$$ price tag to match.

It would still be cool to see somthing like a devils canyon nuc that actually has decent support though with Strix Halo but I'd never buy it.

 

[usiname]

Just to clarify… My comment was more aimed towards the tech press as opposed to AMD. I just think it’s absurd to suggest an upcoming laptop iGPU, which has to *share* a lower power budget as well as comparatively scarce memory bandwidth with multiple CPU cores, is somehow going to perform on-par with any current midrange dGPU… which in contrast would have its own independent memory subsystem and power budget. 4060-class performance is an extremely optimistic forecast to put it mildly.

The comparison is with mobile x70 card, not desktop. About 4070 mobile, this card is insult for the x70 tier so its very achievable, 25% slower then 7600xt which has 32 CUs, 20% less than Strix Halo. Also 7600xt has 288GB/s bandwidth vs ~200GB/s shared between the CPU and GPU.
Plus this is RDNA 3.5 vs RDNA 3, 780m 12CU is already as fast as the desktop rx 6400 with 12CU, so I expect it to be at least as fast as 4070 mobile

 

[SL2]

The comparison is with mobile x70 card, not descktop.

Another problem is that you have to go for a 4080 mobile to get the current generation with more than 8 GB VRAM, that's a lot of money for a laptop.

Unfortunately, some rumors says it won't get better with the next generation. I hope they're wrong.

@AleksandarK: Is it Strix Point Halo or Strix Halo?

I've only heard the latter.

 

[AnarchoPrimitiv]

Well, despite what some people may say, I think Strix Halo will be a revolutionary product and will fundamentally shake up the product stack and offer the most competition to nvidia in years, and most interestingly, not from the above (halo products), but below (entry and mid range).

 

[Denver]

If I had a nickel for every time a tech publication predicted an upcoming AMD iGPU to match nVidia’s current X060 GPU…

Technically this is a regular GPU chip integrated into an MCM design, so if you consider this an iGPU, also consider that the mi300A is an APU that fulfills your nonsensical sarcastic dream.

 

[Bruno Vieira]

The comparison is with mobile x70 card, not desktop. About 4070 mobile, this card is insult for the x70 tier so its very achievable, 25% slower then 7600xt which has 32 CUs, 20% less than Strix Halo. Also 7600xt has 288GB/s bandwidth vs ~200GB/s shared between the CPU and GPU.
Plus this is RDNA 3.5 vs RDNA 3, 780m 12CU is already as fast as the desktop rx 6400 with 12CU, so I expect it to be at least as fast as 4070 mobile
View attachment 352646

Is has ~200GBps, but it is a much lower latency than GDDR6, and much larger pool (from 4/8gb to 32/64/128gb). GPU seeks system mem if the internal memory is full. I hope to buy a mini PC with this thing.

 

[Daven]

There is no doubt that the rumored specs make this a powerful SoC / RAM combo product. The importance of such configuration will not be found in the rumored specs but rather the ramifications on final laptop/desktop design, battery life and price. If there isn't much to be gained between Strix Halo and a regular APU + GPU combination then the product will be a failure. Make something just for the sake of making something is not success in and of itself.

 

[Sound_Card]

AMD is putting an IGP in their APU that will be as fast as a 6800m. (Both 40cu). That blows my mind.
I was already impressed my RDNA 2.0 12cu in my mini PC. A 40cu RDNA 3 probably has a few tech giants absolutely worried.

 

[LabRat 891]

AMD APU with On-package ultra low latency RAM?
Been waiting for this for *years*.

Now, do HBM

 

[Lew Zealand]

Plus this is RDNA 3.5 vs RDNA 3, 780m 12CU is already as fast as the desktop rx 6400 with 12CU, so I expect it to be at least as fast as 4070 mobile
View attachment 352646

No. The 780m iGPU is about 20% slower than the RX 6400 and it's likely due to lower memory bandwidth.

 

[Max(IT)]

Proper iGPU performances has been announced since a while, but so far everyone failed to deliver

 

[TumbleGeorge]

AMD is putting an IGP in their APU that will be as fast as a 6800m

No, impossible. Maybe between 6650m and 6700m.

 

[Sound_Card]

No. The 780m iGPU is about 20% slower than the RX 6400 and it's likely due to lower memory bandwidth.

Proper iGPU performances has been announced since a while, but so far everyone failed to deliver

Not sure what you define as proper. The 680m and 780m have been amazing IGP performers. 1080p medium settings, and FSR takes them even further. We are getting an IGP that has 4x the CU's as 780m. This will be the first IGP that can genuinely game on 2k, especially with FSR.

 

[Lew Zealand]

View attachment 352693

Those TPU charts are guesses and have huge variations when the GPUs aren't tested directly against each other. Which these weren't for some reason.

I have the 780m (with 7840HS at 65W) and had the 6400 for a while and even used the 6400 at PCIe 3.0 the whole time. It's still about 20% faster, all due to 128 GB/s vs. ~90 GB/s memory speed. The big advantage of the 780m is you can assign it more than 4GB as I use mine at 8 and it can even allocate 16 if you want, but with only 32GB DDR5 installed in my UM780, I keep it at 8GB.

Edit: found my 3DM comparos

Having some RDNA2 and 3 GPUs, this trend of RDNA3 being kinda crap at DX11 (or RDNA2 being really good?) seems consistent though RDNA3 closes the gap in heavier DX11 games like Ark: SE with settings turned up.

 

[Yashyyyk]

Right… today’s PS5 and Xbox SOCs are blessed with 6-8 times the memory bandwidth possible from even today’s fastest LPDDR5 configurations, as well as 10-15 times the power budget allowed to the GPU alone. While *technically* an iGPU, it’s really not in the same class (or market) as the aforementioned laptop SOCs…

I like to think that you are paying for the efficiency increases over the past 4 years

Another problem is that you have to go for a 4080 mobile to get the current generation with more than 8 GB VRAM, that's a lot of money for a laptop.

Unfortunately, some rumors says it won't get better with the next generation. I hope they're wrong.

@AleksandarK: Is it Strix Point Halo or Strix Halo?

I've only heard the latter.

Yeah, pretty sad, but hopefully this is the answer/competition needed

and more efficient, if only at lower power limits; Nvidia is already more efficient than AMD, perhaps that only widens

 

[R0H1T]

I think Strix Halo will be a revolutionary product and will fundamentally shake up the product stack and offer the most competition to nvidia in years, and most interestingly, not from the above (halo products), but below (entry and mid range).

If AMD is smart enough they'll release the full fat version in a NUC or similar form factor as well against Apple M4 or higher variants. Heck a zen5c based 8c(16c?) chips could be just the kick Apple needs

AMD isn't really competing with Intel or QC at the moment their real target should be Apple!

 

[Tek-Check]

Also 7600xt has 288GB/s bandwidth vs ~200GB/s shared between the CPU and GPU.

Top Strix Halo SKU should have LPDDR5X at 8000 or 8500 MT/s, so 256-270 GB/s.

 

[SL2]

Besides the 40 CU's, the doubled bus width and soldered RAM will also make a difference.