Monday, July 31st 2023

AMD's Upcoming Strix Halo Mobile SoC Said To Feature 16 Cores, Improved IO Die and GPU

by
TheLostSwede
 Discuss (45 Comments)
Based on details posted on Twitter/X by a pair of well known leakers, AMD appears to be working on a pair of different Ryzen 8000-series mobile processors. The previously known Strix Point is said to get up to four Zen 5 cores and eight Zen 5c cores, whereas the Strix Halo is said to get 16 Zen 5 cores, according to @Olrak29_. This is something that was posted by Moore's Law is Dead back in April as well, who claimed the chip will launch sometime at the end of 2024. MLID also suggested that the Strix Halo will feature a 40 CU GPU and a 256-bit LPDDR5X memory interface, making it a very different proposition from your average APU from AMD.

@kopite7kimi chimes in on Twitter to point out that "Strix Halo looks like a desktop Zen 5 with a different IOD." This is definitely something that would be possible for AMD to do and if we look at the MLID information, the Strix Halo processor appears to have something called a Mall Cache, which seems to be something of a catch all cache for the various components inside the chip, such as the AI Engine and the GPU. Time will tell if AMD delivers on Strix Halo or not, but this might be the first notebook processor that can handle gaming at a decent resolution without needing a discrete GPU. Then again, with a rumoured peak TDP of 120 W, this chip is also going to run hotter and draw more power than most mobile processors to date.
Sources: @Olrak29_ on Twitter, @kopite7kimi on Twitter, Moore's Law Is Dead on YouTube, via VideoCardz

Related News

Add your own comment

45 Comments on AMD's Upcoming Strix Halo Mobile SoC Said To Feature 16 Cores, Improved IO Die and GPU

#1
ymdhis
They have something like half a dozen mobile chips now but we are hearing nothing about desktop APUs. This sucks.
Posted on Reply
#2
Post Nut Clairvoyance
256-wide LPDDR5X sounds incredibly interesting. Hope to see these leaks materialise on the market, and compared to a traditional G6 memory subsystem.
Posted on Reply
#3
Raiju
Quad channel on laptop :rockout:
Posted on Reply
#4
doc7000
The 40 cu RDNA GPU for laptops is interesting, this will compete with laptops that have separate GPUs in them. 40 CU that is more then what you get in a 7600 though likely not have as much power to play with so maybe comparable performance to that card.
Posted on Reply
#5
dj-electric
Ive always stuck to the opinion none-console users should have a console-like SoC if they want to get one. The meaning is mostly a chip that contains both powerful CPU cores and a powerful GPU, all paired up with fast memory. Now, we don't get GDDR6, but we can at least access LPDDR5X
Posted on Reply
#6
TumbleGeorge
RaijuQuad channel on laptop :rockout:
Lpddrx 32 bit bus per dimm(2X16 internal) 8 channels(dimm's*) needs for 256 bit wide.
* If will be used dimms.
May math is bad. If I'm is wrong, sorry.
Posted on Reply
#7
AusWolf
Source: MLID. Sure, folks, keep on dreaming about a 40 CU APU! :slap:
Posted on Reply
#8
TheLostSwede
News Editor
TumbleGeorgeLpddrx 32 bit bus per dimm(2X16 internal) 8 channels(dimm's*) needs for 256 bit wide.
* If will be used dimms.
May math is bad. If I'm is wrong, sorry.
LPDDR5X is solder down only, although we might get LPDDR CAMMs at some point in the future if the standard ever takes off.
Posted on Reply
#9
Tek-Check
AusWolfSource: MLID. Sure, folks, keep on dreaming about a 40 CU APU! :slap:
He was the first source several months ago already. And now we are getting similar information from more people. I hope you read the article.
ymdhisThey have something like half a dozen mobile chips now but we are hearing nothing about desktop APUs. This sucks.
Those are being prepared. We had a few leaks some time ago.
Posted on Reply
#10
Post Nut Clairvoyance
TheLostSwedeLPDDR5X is solder down only, although we might get LPDDR CAMMs at some point in the future if the standard ever takes off.
I see little chance for [consumer non-desktop] socket dimm unfortunately.
by itself is not a sufficient selling point, nor will companies advertise such uses.
especially true for GPUs as NVidia\AMD remains antagonistic\indifferent towards everyone.
Posted on Reply
#11
TheLostSwede
News Editor
Post Nut ClairvoyanceI see little chance for [consumer non-desktop] socket dimm unfortunately.
by itself is not a sufficient selling point, nor will companies advertise such uses.
especially true for GPUs as NVidia\AMD remains antagonistic\indifferent towards everyone.
JEDEC has made CAMM a standard for laptop memory modules, so it's not impossible, but so far, Dell is the only company I know of that has adopted the standard, since they came up with it.
www.jedec.org/news/pressreleases/jedec-expands-camm-standardization-include-two-key-memory-technologies
Posted on Reply
#12
Dan.G
Wow, that 40 CU APU looks like a real gaming beast (take notes Intel)! I'm a little skeptical about that 256-bit LPDDR5X, though. Even the 128-bit bus, running at 7500 MT/s, should provide a decent 120 GB/s.
This is the future, boys... No more dedicated GPUs... Hope they make it whisper quiet and cold as ice! :cool:

P.S. Any news on pricing? :rolleyes:
Posted on Reply
#13
Sake
"Then again, with a rumoured peak TDP of 120 W, this chip is also going to run hotter and draw more power than most mobile processors to date."
My Intel Core 12800 HX will dare to contradict you, with 157W TDP, loud as hell and hot. Also much slower than my Ryzen 7900X desktop limited on 105W TDP.
I await for a long time a laptop CPU that doesn't need a dedicated GPU.
Posted on Reply
#14
AnarchoPrimitiv
AusWolfSource: MLID. Sure, folks, keep on dreaming about a 40 CU APU! :slap:
Why is there a small, yet extremely vocal group of people who hate MLID? They constantly rag on him for wrong information, but that's LITERALLY the nature of leaks, a lot of them are going to be incorrect or companies change their plans over time, so I don't get it, it's like they have a specific bone of contention with him despite the fact that most leaks and therefore most leakers are incorrect. I can understand just ignoring him or not having an opinion, but I don't understand the very specific and occasionally aggressive hatred toward him that is not displayed toward other leakers.
Posted on Reply
#15
R0H1T
Except he basically covers an entire spectrum from plausible but unlikely to complete batsh!t & crazy BS leaks. He probably has gotten better in terms of selectively "leaking" this stuff but even then he's just WTFtech on YT atm.
AnarchoPrimitivbut I don't understand the very specific and occasionally aggressive hatred toward him that is not displayed toward other leakers.
When someone constantly say/claims I know stuff from inside sources he better be effin ready to reveal them, because he can just as well make that up!
Posted on Reply
#16
chstamos
R0H1TExcept he basically covers an entire spectrum from plausible but unlikely to complete batsh!t & crazy BS leaks. He probably has gotten better in terms of selectively "leaking" this stuff but even then he's just WTFtech on YT atm.


When someone constantly say/claims I know stuff from inside sources he better be effin ready to reveal them, because he can just as well make that up!
As the previous poster mentioned, doesn't that equally apply to all leakers. If he's going to burn down their employment they're not going to remain his sources for much longer, are they? Unless he has an egregious record of incorrect predictions, far worse than other leakers, I don't really get the particular animosity he's singled out for either. Not taking his "side" or anything, it just seems par for the course for sources specializing in leaks that many of them will be wildly off.
Posted on Reply
#18
AusWolf
AnarchoPrimitivWhy is there a small, yet extremely vocal group of people who hate MLID? They constantly rag on him for wrong information, but that's LITERALLY the nature of leaks, a lot of them are going to be incorrect or companies change their plans over time, so I don't get it, it's like they have a specific bone of contention with him despite the fact that most leaks and therefore most leakers are incorrect. I can understand just ignoring him or not having an opinion, but I don't understand the very specific and occasionally aggressive hatred toward him that is not displayed toward other leakers.
I don't hate the guy. I only hate the fact that such a reputable and high-quality news and review site as TPU uses his "leaks" (whatever made-up BS he means by that word) as a source for articles.

If some information may very well turn out to be false in the near future, then there's nothing to distinguish it from made-up BS. Or if there is, then I'll "leak" that I might travel to Mars tomorrow.
Posted on Reply
#19
R0H1T
AusWolfthen I'll "leak" that I might travel to Mars tomorrow.
Are you planning to take this guy along with you?
Posted on Reply
#20
Vader
AusWolfI don't hate the guy. I only hate the fact that such a reputable and high-quality news and review site as TPU uses his "leaks" (whatever made-up BS he means by that word) as a source for articles.

If some information may very well turn out to be false in the near future, then there's nothing to distinguish it from made-up BS. Or if there is, then I'll "leak" that I might travel to Mars tomorrow.
Rumours, especially of interesting chips like this one, attract views and clicks, is that simple. If the article is based on a leak, and the title correctly implies that ("...said to..."), then it should be enough. This is true for all leakers, and again you're attacking MLID specifically, and not the other leakers that also appear in this page, even in this article itself.
Posted on Reply
#21
TheLostSwede
News Editor
AusWolfI don't hate the guy. I only hate the fact that such a reputable and high-quality news and review site as TPU uses his "leaks" (whatever made-up BS he means by that word) as a source for articles.

If some information may very well turn out to be false in the near future, then there's nothing to distinguish it from made-up BS. Or if there is, then I'll "leak" that I might travel to Mars tomorrow.
Hence why we didn't post in April when he leaked it, only now when others have confirmed similar specs. It's also hard to glance over the fact that he had what appears to be the same info as these other leakers, several months ago.
We still don't know if any of these leakers are correct.
Also, I would say most of what he posts isn't made up, he has some real sources, but I think he sometimes get fed bad info by some of his sources.
Posted on Reply
#22
xrli
256-bit wide bus paired with 8533MT/s LPDDR5X will bring around 270GB/s of total bandwidth. This is almost on par with some mid low range discrete GDDR6 GPU's bandwidth, finally a worth APU!
Posted on Reply
#23
TheinsanegamerN
Dan.GWow, that 40 CU APU looks like a real gaming beast (take notes Intel)! I'm a little skeptical about that 256-bit LPDDR5X, though. Even the 128-bit bus, running at 7500 MT/s, should provide a decent 120 GB/s.
And the RX 7600, of similar spec, has a 288GB/s bus just for itself. If they are going to put 40 CUs into an APU, they have to feed them.
Why do people always go with this "muh APU dont need no bandwidth" ?
Dan.GThis is the future, boys... No more dedicated GPUs... Hope they make it whisper quiet and cold as ice! :cool:
And there's the other argument: "guys APUs are more powerful now and none of those improvements ever go into dGPUs, they'r efinished!"

Just a reminder, the llano APUs were praised for being able to play some low demand games at 1080p. The current APUs are being praised for being able to play lower demand games.....at 1080p. This APU will also be used....for 1080p. Short of sticking a 690mm2 iGPU on there you will never eliminate the need for dGPUs.
Posted on Reply
#24
Evildead666
AusWolfI don't hate the guy. I only hate the fact that such a reputable and high-quality news and review site as TPU uses his "leaks" (whatever made-up BS he means by that word) as a source for articles.

If some information may very well turn out to be false in the near future, then there's nothing to distinguish it from made-up BS. Or if there is, then I'll "leak" that I might travel to Mars tomorrow.
The leak has to be plausible though.
;)
Posted on Reply
#25
AnotherReader
TheinsanegamerNAnd the RX 7600, of similar spec, has a 288GB/s bus just for itself. If they are going to put 40 CUs into an APU, they have to feed them.
Why do people always go with this "muh APU dont need no bandwidth" ?

And there's the other argument: "guys APUs are more powerful now and none of those improvements ever go into dGPUs, they'r efinished!"

Just a reminder, the llano APUs were praised for being able to play some low demand games at 1080p. The current APUs are being praised for being able to play lower demand games.....at 1080p. This APU will also be used....for 1080p. Short of sticking a 690mm2 iGPU on there you will never eliminate the need for dGPUs.
The IGPs did successfully kill off the lower tier discrete GPUs like the GT 1030 and RX 550. We used to have lower tier GPUs catering to the sub $100 market. Integrated GPUs have almost completely taken over that market. They would be doing even better if AMD and Intel were willing to go the Apple route and give wider memory interfaces to their SOCs. There's no technical reason that a properly designed IGP can't beat a mobile 4060. Most laptops have soldered memory so a bigger APU wouldn't be a problem on that front. Dare to go big enough and you could conquer a mobile 4070 as well.
Posted on Reply
Add your own comment
Nov 17th, 2024 15:14 EST change timezone

Latest GPU Drivers

New Forum Posts

Popular Reviews

Controversial News Posts