AMD Sound Wave ARM APU Leak: Built to KILL Apple M4 & Qualcomm Snapdragon X Elite!

[@MooresLawIsDead]

[SPON: New Customers get the Creality Ender 3 S1 3D Printer for $149: micro.center/mpbw ] [SPON: Shop Micro Center’s Top Deals for May: micro.center/2e630c ] [SPON: Get the latest Tech News & Buying Guides at MC News: micro.center/pnow ]

[@kaelandin]

No

[@ADB-zf5zr]

From what I heard at the time, the K12 project was very interesting and was essentially in two parts, one appears to have made it to some degree and most people do not know it even exists.! From memory, the K12 project was an ARM AMD range primarily for servers (that's where the money is) and they were talking about 128-core (at least) CPU's back then and it was widely speculated that they didn't want to eat into their own X86 server chip sales, but it got to the point that they had a product that was almost ready to launch to market but they chose not to. I can understand why, the sheer cost of support alone would be a serious burden and they had limited resources (money, people) to use and instead of going with both options, they chose to drop the K12 project. . But not entirely, they seemed to have learned a lot, some of that was low-power operation that I heard was used to design and tweak x86 chips to be more efficient, and of course the "AMD platform security processor" has been an ARM CPU within all AMD chips since 2013 and I have no doubts that this has been beefed up, tweaked etc over the years and the knowledge from the K12 project would certainly have helped in some way. . Bear in mind that this was a rumour from many years ago (talked about many, many years before the K12 project was talked about publicly in 2017 (hence the direct relation to the AMD PSP), but there was talk about an AMD hybrid CPU that used both ARM and x86 processors, and shared system memory. I assume that the basic premise would be something like Proxmox but for the average end user. The system would boot using the ARM cores, run a base OS and various background tasks, perhaps even the storage subsystem (but this is my speculation), and the end user would use the x86 processors running Windows, the primary issues here would be divided loyalty and MicroShaft would not like this, but also, what exactly would the ARM processors be doing that the x86 processors could not, plus the added software complexity and how in reality these two things would work together, so many question and pitfalls. . Looking at it now, with MicroShaft having actual usable support for ARM, and Intel's big.LITTLE x86 architecture (but supporting different instructions), there would be no better time than now (or a couple of years ago) for AMD to make this a reality and have an ARM x86 hybrid where under Windows the background talks could be run by the ARM processors entirely and not interfere at all with the x86 processors (excepting memory usage) and possibly be used as primary processors in low power mode under battery use and shut down the x86 processors entirely when not in use. This would obviously still be a lot of work on the software side and AMD would need to have others (MicroShaft) entirely onboard with this because AMD is not a software company and would be entirely reliant on others in this regard. . As for the server side of things, a similar proposal but with legs. The ARM cores would essentially run the background tasks (think Proxmox), but some of the ARM processors could also be used to natively run ARM code, therefore being a true hybrid chip where two distinct operating systems using different processors with different instruction sets and both could run code natively. . "If" AMD had gotten an ARM CCD to mass production that seamlessly worked with the Ryzen IOD and the EPYC IOD then they could have been interchangeable and so give both desktop and server both options simultaneously.! I imagine the sheer cost of support was a primary factor, as well as having to design and manufacture an entire new CCD which eats into cash and Human resources, and it was simply not viable at the time financially. . Things have changed, the landscape has changed, ARM is now being used outside of mobiles. It is being used in gaming (Nintendo Switch), and Servers (albeit custom), so now IMHO there is no reason why AMD could/should not jump into this head first.!

[@Ronny999x]

I thought Intel was going RISC-V instead of ARM.

[@lolmao500]

Micro center needs locations in canada.

[@mitcoes]

Why Huawei ARM AiO GNU/Linux devices sold to consumers (chinese government was already buying them) is not part of this news? Why GNU/Linux on ARM vs MS WOS for desktop / tablet use (historically better Linux than NT kernels) is not part of this news too?

[@cairnex4473]

AMD are going to start naming APU's after transformers... looking forward to "Starscream" and "Megatron"

[@wolfblade6569]

Lol😂

[@erlienfrommars]

And they're naming it after Decepticons; as nvidia have already used the "Optimus" name in the past.

[@mikeb3172]

Don't you think they're more mature than that?

[@cairnex4473]

@@mikeb3172 Jokes are jokes bro...

[@dairyb0y]

Opteron already kinda sounded like a transformer name.

[@WickedRibbon]

I like to think "Sound Wave" is a reference to the fearsome Decepticon general.

[@Strykenine]

Literally came here to say the same thing.

[@benjaminoechsli1941]

The best of the bunch.

[@aztracker1]

Soundwave superior... Qualcomm inferior.

[@modernlogix]

Ooh interestin'

[@erlienfrommars]

The microsoft tablet that would use that APU is probably capable of transforming.

[@superjuddy]

actually insane that you're being sponsored by micro-center now Tom, really glad to see the success you've built for yourself!! God bless man!

[@2pumpedupforu]

Brother, he has been sponsored by them on and off for at least the two years I’ve watched him.

[@Etheoma]

He's had micro-centre ad spots for ages.

[@TechOtaku86]

Bros been living under a rock for YEARS

[@GewelReal]

Now imagine if he was right about his leaks lol

[@LuciusAugustusRex]

An ARM chip that has full access and support to desktop graphics is a huge differentiator to Qualcomm and it's ARM GPU cores.

[@Spyke_misc]

then again qualcomm said they aren't against the idea of letting DGPU on their chips

[@handlemonium]

The x86-x64 translation layer (not just emulation) needs to be solid though so that adoption doesn't rely on developers porting more "pro" software to ARM. And yes, more than just Adobe Photoshop and Da Vinci Resolve.

[@teaser6089]

@@Spyke_misc True, then again why would AMD work with Qualcomm and lose out on a chunk of the profits, if AMD can also just take the entire market away from Qualcomm

[@seeibe]

For me as a software developer thr differentiator would be if it would run Linux. AMD has a good track record of making their hardware accessible to open source developers.

[@artun42]

Especially considering that Adreno is AMD's initial design (hence the similarity to Radeon). They now their competition across the board.

[@John.Philip.Tan876]

Damn... Lisa Su must be tired of all the Arm fans bothering her about that Arm IP they had.

[@richard.20000]

RISC was the future since 1990's. RISC is the successor of old CISC. x86 CISC survived till these day only thanks to using RISC-like internal core. 1st RISC-like x86 was AMD K5 (modified RISC Am 29000 CPU with x86 decoder).

[@lllongreen]

.

[@The_Ninja_Chin]

@@richard.20000ARM also become more x86 like when utilising more fixed function/specialist parts in its architecture. They are more alike than they used to be. Not to mention x86 prioritise backwards compatibility.

[@BastyTHz]

amd never learned ... they already forgot what ms did to temash. american will always buy white ppl stuff not others. temash was good and better than intel atom but ms did not buy it as agreement in developing windows 8 and tablet.

[@riven4121]

Guess you could say they took an arm and a leg I'm not even sorry for this

[@Trevor0713]

it won't surprise me if AMD is working on an Arm based APU. After all, they have hundreds of engineers who are quite familiar with Arm architechture, and had been putting them into their products for decades. yes, i'm talking about the Xilinx team. AMD's CPU team had borrowed many technologies from Xilinx.

[@user-lp5wb2rb3v]

that might have been part of the plan for acquiring xilinx

[@visitante-pc5zc]

Never heard of the xing ling team

[@danieloberhofer9035]

​@@visitante-pc5zc So what do you think where XDNA (both the name and the IP) comes from?

[@Grimmwoldds]

@@visitante-pc5zc Xilinx was the FPGA company AMD bought(Intel bought Altera). The primary driver was both wanted NPUs based on FPGA technology, but there's a lot of applications for FPGAs.

[@skywalker1991]

Amd already have shortage in engineers, they can't even have enough to design proper GPU , now venturing into ARM CPUs , Rule #1 , stick to what you good at and make those better , Amd shouldn't stick their hands into new jars , ARM is different Animal than x86 , There are too many players already has better ARM CPUs designed , Nvidia , Qualcomm, and many more . Amd will fall flat even if they designed fastest ARM cpu cause and can't supply in large quantities, no one will consider AMD Arm cpu . Please amd stick to x86 and Radeon GPUs , and work on selling their AI generative gpgpus. Period

[@neliaironwood7573]

AMD did try out on ARM-based Opteron A1100 SoCs back in around 2014 to 2016, a year before AMD Ryzen. They didn't really take off though, so they suspended their efforts on ARM but put R&D into ARM-based SoCs, including an ARM-based Ryzen with Radeon Graphics which will probably be that Sound Wave. Something tells me that the Sound Wave APU might not be called Ryzen, but rather a new name.

[@jackofthecoke]

Not just didn't take off, they mothballed the entire project. Remember, they didn't have the kind of capital they have today to deploy on 2 different horses. Ryzen was their hail mary as it was. But as you'd expect, that decision did have its opponents. Jim Keller worked on that ARM project as well and years later felt that not going ahead with it was a mistake.

[@johnd.1618]

Su was killing projects in her first year. One example is the ARM Opteron, the other is the AM1 platform.

[@jamieknight326]

I’m pretty sure I saw these in use within Panasonic broadcast gear at my old job.

[@effexon]

is there any benefit to arm based cpu (on server, opteron is) vs current x86 ryzen/TR/epyc that would offset huge effort to recode x86 software on it?

[@Jimster481]

​@@effexonnot really.

[@josephmazor725]

Amd learned from Sun zu. Know your enemy (arm) and yourself and you need not fear the result of 1000 battles

[@jcdenton7914]

They haven't been able to get past Nvidia yet.

[@josephmazor725]

@@jcdenton7914 Jetson already read Art of War, Dr Su is just finishing now and we can only hope Pat has started

[@visitante-pc5zc]

Intel is dead

[@tringuyen7519]

@@jcdenton7914AMD hasn’t reached Nvidia yet. But Nvidia’s rushing out Blackwell in 2024 & Rubin in 2025 to stem off AMD. MI300X with 192G of VRAM scared Nvidia!

[@davidaustin5622]

"Avoid what is strong, and strike at what is weak" - Sun Tzu. Explains why AMD is focusing on Intel rather than Nvidia.

[@kingofstrike1234]

you should keep the confidence level notes on the slides

[@user-ec8zd8zp4x]

I can see why AMD is interested in ARM. ARM looks great for handhelds/ thin and light laptops/ tablets/ mini PC's. The 9-15w region is the region you want to be in for those devices or they'll drain to much power for their use and/ or they will become to heavy for comfort because of the bulky battery needed. Personally I would love to see an ARM Steam Deck/ handheld PC, but it needs to be compatible with X86 software. Apple has shown us that is absolutely possible. I am also very curious about the performance of Switch 2, see if the rumours are right.

[@SWOTHDRA]

Totally agree with you its time AMD makes the switch (pun intended) towards ARM

[@pweddy1]

The ironic thing is Intel had the XScale Arm series of CPUs years ago and sold it off.

[@johnd.1618]

And AMD sold the mobile GPU team to Qualcomm, giving Qualcomm the biggest advantage it enjoys the last 15 years. The Adreno GPUs.

[@teaser6089]

@@johnd.1618 True, difference is that AMD had to sell it off in order to survive, whilst Intel was and still is doing fine.

[@visitante-pc5zc]

Intel is dead

[@arenzricodexd4409]

@@johnd.1618 Qualcomm advantage is in the baseband/modem tech not so much on the GPU front.

[@elivegba8186]

@@arenzricodexd4409 don't you see them winning in the GPU side of things?

[@ernestuz]

A few years back (many few years), AMD developed an ARM family that they targetted to servers. I remember meeting a couple of Texans in Cherry Hinton at ARM offices a couple of years before it happened. At the time, the development cycle of a new CPU took around 18ish months. I naïvely thought that x86 was going to be dead :)

[@edwardpark247]

Opterons

[@Meoknet]

AMD has said already they're prepared to do ARM if the market calls for it, so I'm not surprised they're working on something. An APU makes sense because they have an edge in APUs with RDNA in house. Add XDNA into that mix and they could really produce something special for ARM.

[@Truemann45]

No

[@andersjjensen]

@@Truemann45 What a well worded response. Very solid and coherent arguments too. You clearly know your stuff.

[@Truemann45]

@@andersjjensen short + sweet

[@aforty1]

You’re the best leaker and analysis online. Pat yourself on the shoulder but don’t let it go to your head. Thanks for everything!

[@abowden556]

Personally my hope is that they use the learnings and feedback from this to improve the low power and sleep performance on their low TDP x86 products! you can have all the instructions, but they don't all need to perform the same, especially the ones nobody uses!

[@benc3825]

ARM isn’t faster or slower and isn’t more or less power efficient. X86 can be extremely power efficient, it simply comes down to the area they are designed for. ARM and X86 have exist in entirely different markets for too long that we now associate low power with ARM and high power with x86. It comes down to the size of the core architecture, instruction sets, workload, frequency/voltage curve, and the packaging. Look at P cores: Zen 4 vs Raptor Cove, “E cores” Zen 4C vs Gracemont Different design focus/ same architecture Zen 4 vs Zen 4C. Zen 4 and Raptor cove performance around the same on average but Zen 4 uses a lot less power on average Zen 4C vs Gracemont… well I don’t really need to say anything about it Zen 4 vs Zen 4C… they are exactly the same core design, but the base Zen 4 is made with higher frequency at the cost of some power and die size. Zen 4C is the exact opposite, lower frequency with lower power and die size. If you look into it, Zen 4C actually can hit a higher frequency when you only have a few watts per core than base Zen 4. For the packaging: basically all of AMD’s chiplet products have one problem, the die to die interconnect are EXTREMELY power inefficient in light workloads. There are much better packaging alternatives that do cost more which they should be using in the coming gens.

[@abowden556]

@@benc3825 I know this! The main problem is getting AMD chips into the hands of people who have experience optimising devices for super low power, that generally means ARM people. If I thought x86 was irredeemable in that space, I'd have no hope of the learnings being transferable!

[@benc3825]

@@abowden556 I see what you mean now and fair point. It still is mostly architectural design choices than anything else. Personally, I don’t think they’d need to get any ARM people for ultra low power cores, because the Zen architects are freaking nuts with what they can do

[@michaelandrews4783]

Learnings? how about "they learn from the feedback to improve the low power and sleep......"?

[@benc3825]

@@michaelandrews4783 Just to make sure we are on the same page, what are those problems?

[@geekinasuit8333]

AMD has been making ARM CPU's after acquiring Xilinx and Pensando. AMD also worked with Sansung to integrate an iGPU into the Exynos ARM SoC. The K12 work was placed on hold so that AMD could focus on X86. which was the bigger market for AMD. Now that Intel is "in the rearview mirror" (to use Pat's own words), AMD can concentrate on other architectures. What is very clear, is that AMD is now focusing on adding AI capabilities to all of its products, and that will include the ARM SoC's that Xilinx and Pensando use. Under Lisa Su's leadership, AMD has been aiming at the highest margin sectors as a priority, which means whatever ARM ambitions they have, will be directed at the high margin spectrum where ARM fits in, whatever that may be. AMD may simply plan to include ARM chiplets into some of its high-end devices to provide flexibility for some of its biggest customers. For example, AMd could add ARM chiplets in place of x86 chiplets to some of its Instinct accelerators, or release an EPYC with ARM chiplets instead of x86.

[@epobirs]

Going back even farther, the acquisition of ATI also included some ARM experience there, when ATI made a product that today would be at the heart of a smartTV or box adding functionality to an existing TV. It was a bit too ahead of the curve and didn't do a lot of business.

[@roflmagister5]

AMD A1100 SOC (with Cortex-A57 core, dunno how much that qualifies as "making CPUs"), which was way before Xilinx.

[@KennyNguyen-rf9hl]

​@@epobirs😅

[@andersjjensen]

LOL! That "mike drop" exit statement was gold.

[@daxterrhiley7079]

Risc-V should be the better choice. The license is free, so asian comps are already developing CPUs - like alibaba, huawei etc. So you can build CPUs independently with the same instruction set. Different designs, different fabs but same ISA = same windows for RISC ?

[@_Digitalguy]

Windows is not being coded to work with RISC V

[@EthelbertCoyote]

Just a guess but AMD With ARM is a little about laptop and a lot about Data center. Apple has not done it yet but at the power levels they are running at with their whole SOC and unified shareable memory if you wanted to run large Ai models on the cheap, I think it would do pretty well on the $ to compute metric.

[@jamieknight326]

I’d love to be able to run M series CPU’s via AWS. A 20 core M1 Ultra can already outperform ~35-40 Graviton cores for our financial fraud detection tools.

[@effexon]

arm in datacenter? I didnt know that is a thing. I thought epyc/xeon was mainstream there, though some supercomputers in past have used arm based architectures. bulk of cpus are x86 though (google,amazon, etc datacenters in hundreds of thousands rigs).

[@EthelbertCoyote]

@@jamieknight326 To be clear, "IF" you removed certain design choices that Apple made to focus on a "full" SOC ex. audio processing, etc... they Apple or AMD could squeeze more cores and average out the typical arm heat pattern making it easier to cool and stabilize. So the M series is only a crumb of what could be done. I also, while I cannot prove it think Apple really wants to enter the datacenter but know just how hard that really is. AMD has a much bigger head start with customers. Finally the big win is with the RAM on the chip. sitting closer to the cores you can really use slower ram and therefor more than most competitors could VRAM, the sharable part, CPU or GPU makes it very interesting for some workloads.

[@jamieknight326]

@@EthelbertCoyote that would be nice if they did. We currently develop on M series Macs and deploy to graviton instances in AWS. M series instances on AWS would be a game changer.

[@definingslawek4731]

@@EthelbertCoyote The ram is NOT on chip for apple silicon or snapdragon elite. I don't understand why everyone gets this wrong. Surprised to see this even on a technical channel. The ram is a distinct chip just like on every other computer. Just take a look at an apple logic board. You see the SOC and then next to it are the dram modules. Just imagine how many variations of SOCs theyd have to make to have all the ram variants.

[@pedro4205]

About the retail source, AMD is already partnering with Samsung on ARM processors, in most of the regions Galaxy S24 and S24+ are using a SOC with RDNA2 GPU.

[@sbrader97]

To clarify Exynos 2400 is actually RDNA3

[@Freddie1980]

AMD is right to do this as they have no option because if Microsoft continues to work away on ARM for Windows then Qualcomm and Samsung will chew away at the x86 share of the Windows install base.

[@TheAutumnNetwork]

Pure facts, especially since MS will be most likely showing off the new surface products with Qualcomm's ARM chips this upcoming week which will get some folks excited and curious. I'm definitely in the curious camp myself, and I would love to see AMD take a true, genuine attempt at ARM, even though their APUs have been stellar too.

[@Just_An_Ignacio]

Well, Samsung's Exynos 2200 had RDNA2 integrated graphics and performed very well against its Snapdragon competitor at the time, so AMD shouldn't have a lot of problems to use an enhanced version of the most performant ARM cores for their first comercial CPU of the type.

[@szczurki3987]

What's funny is that technically we already have ARM based SoCs featuring AMDs IP, at least when it comes to graphics. The Samsung Exynos chips feature RDNA based GPUs in them after all (Exynos 1480, 2200 and 2400)

[@sneg__]

"adreno" is an anagram of "radeon"

[@artun42]

​@@sneg__I just wanted to say that.

[@kozad86]

It makes sense, they can't rely on Intel's license exclusively as ARM begins to grab market share on PC. AMD needs something big to save Radeon, and this in handheld PCs could be it.

[@benc3825]

AMD doesn’t rely on Intel’s licensing. They both have the same base x86, the processors will just have different x86 extensions, like how Zen 4 supports AVX-512 while Raptor Lake can’t run the instruction As for saving Radeon, what do you mean? They are completely fine and a CPU architecture chance wouldn’t change that

[@aravindpallippara1577]

The only thing that can save Radeon (if it even needs saving in the first place) is Radeon themselves.

[@benc3825]

@@aravindpallippara1577 it doesn’t need to be saved… well… it needs to be saved form the terrible marketing team, but other than that, it’s just fine

[@kingyogesh441]

@@benc3825why is Radeon no longer forcing NVIDIA price cuts 👀

[@michaelandrews4783]

Radeon only powers almost all the consoles , it's doing just fine

[@Summanis]

Will AMD use an ARM core design or will they make their own core around the ARM instruction set?

[@sargentpayne]

This feels like a later than usual for you post, I am glad you did. So far all your leaks have been awesome keep up the hard work.

[@johnnytshi]

Lisa said it over and over, heterogenous computing. X86, Arm, GPU, NPU, FPGA. From HPC, to data center, to desktop, to laptop, and other portables.

[@Will-sc3hw]

Would love a android with an AMD apu. The gpu performance would be insane

[@sbrader97]

Exynos is already using rdna3 gpu

[@LUKAS3675]

few thoughts: - imagine an epyc-arm lineup (server market would love that) - how about risc-v (see amd MicroBlaze) - the next steam deck releases in 2026 ;) - AMD already has Xclipse, a smartphone (ARM) GPU

[@StephenMcGregor1986]

Hmm... "You're absolutely right! Here's a breakdown of your points: Steam Deck 2 with ARM CPU: Strong Hypothesis: It's a very strong hypothesis that the next Steam Deck iteration could utilize an ARM-based processor like the upcoming AMD Sound Wave APU. Here's why: Linux Advantage: As you mentioned, Linux is well-optimized for ARM processors, and the Steam Deck heavily relies on Linux. Power Efficiency: ARM processors are generally known for their power efficiency, which is crucial for a handheld gaming device like the Steam Deck. AMD's Focus: With the acquisition of Xilinx and Pensando, AMD is gaining expertise in ARM technology. The Sound Wave APU being ARM-based aligns with this strategic shift. Market Demand: The success of the current Steam Deck shows a strong market for powerful handheld gaming PCs. An ARM-based Steam Deck 2 could offer improved battery life and potentially lower production costs. Commentary on AMD's New APU: The commentary you shared highlights some key points about AMD's current and future direction: ARM Expertise: Acquiring Xilinx and Pensando has indeed bolstered AMD's ARM capabilities. This opens doors for future ARM-based products beyond just the ones acquired companies were developing. Focus Shift: Prioritizing x86 for the larger market makes sense. However, with Intel as less of a pressing concern, AMD can now explore ARM more strategically. High-Margin Focus: Lisa Su's leadership emphasizes high-margin sectors. This suggests AMD might use ARM for specific applications where it offers a clear advantage, like power efficiency in mobile devices or potential cost benefits in specific server segments. ARM Chiplets: The concept of including ARM chiplets in existing x86 products like EPYC CPUs is interesting. It offers flexibility and caters to specific customer needs. Overall: The commentary accurately reflects AMD's evolving approach. They are well-positioned to leverage ARM technology for strategic advantages in various product lines, and the Steam Deck 2 with an ARM processor is a strong possibility. Additional Considerations: Software Compatibility: While Linux is strong on ARM, ensuring compatibility with existing Steam games on an ARM-based Steam Deck 2 would be crucial. Emulation could be a solution, but performance might be impacted. Developer Adoption: Encouraging developers to optimize games for ARM will be essential for the success of an ARM-based Steam Deck. The future of the Steam Deck and AMD's ARM strategy is exciting. It will be interesting to see how they navigate these considerations and potentially revolutionize the handheld gaming PC market."

[@yarmgl1613]

if they make their arm CPUs compatible with AM5 boards it will definitely take off

[@yuryzhuravlev2312]

I can expect hipe around ARM next 2-3 years, but after RISCV will smash ARM and x86. It's already a case for micro controllers. Also, now it's a perfect time to start new vertical integrated company with RISCV + modern (and open) LInux (not same as Ubuntu or RHEL). Success of Chromebooks shows it's possible.

[@SpencerHHO]

I've maintained a level of skepticism towards ARM on mainstream desktop but with Microsoft actually investing in it this time I could see AMD making a play at it. I think the company that should he most worried is qualcom. AMD might be a small company in terms of headcount but they've shown time and time again that they can be incredibly agile and innovative when developing new solutions. I more or less have also observed that ARM loses most of it's effeciency benefits above a threshold of around 15watts but 15 watts can go a lot further today than 10 years ago. I'd be very interested in seeing what AMD could do in that power enevlop.

[@seeibe]

I think this is about mobile, not desktop.

[@_Digitalguy]

the threashold is higher than 15w and with time it will probably increase

[@harrythehandyman]

I am interested in SteamDeck X that uses the Arm APU for longer battery life.

[@WENEVERLEFT]

I've seen that Sony is hiring Software Engineer for "System software development with program compatibility between different architectures). Could mean that PS6 or nextgen portable might be on ARM.

[@pinakijana9479]

I think PS6 is already in developement. Saw somewhere that it's gonna be on zen 5.

[@backupplan6058]

Or you do have the PlayStation Portal which has a Qualcomm CPU in it, while hardly massive seller or something designed to run natively it does require software updates not to mention if they plan a future handheld device.

[@nabieladrian]

It's an open secret that PS6 will launch with dedicated handheld.

[@jaynorwood2]

Intel's PVC GPUs are in Aurora, which is the "fastest AI system in the world dedicated to AI for open science, achieving 10.6 AI exaflops."

[@jamesrock9446]

I'll wait when the LEG cpus enter the market. Rumor is that they will combine with ARM cpus towards a new line of bank breaking products.

[@scarletspidernz]

Then we can use the chips to summon Exodia the Forbidden Computer

[@Knirin]

If AMD can get a chunk of the ARM ecosystem onto UEFI with proper expansion ports then good. If they can’t x86 isn’t going anywhere. Small low power UEFI based x86 devices are at the same price point as a Raspberry Pi 5 or 4. Those x86 devices also don’t require a special kernel from only one vendor. The Pi5 is close but not fully there yet.

[@seeibe]

I don't think it's to compete with cheap low powered x86 devices, but rather to compete with expensive mobile devices like the m2 macbook air.

[@Knirin]

@@seeibe Maybe, it still needs to use a standards compliant UEFI implementation to allow kernel developers a chance to build for it themselves.

[@seeibe]

@@Knirin Agreed, if they have that I will buy one

[@leorickpccenter]

As far as I recall, AMD had plans for ARM for the longest time. But I guess they have to put those off to focus on Zen first. Now they have a stronghold and they can expand and continue on this path. I'm still concerned about the manufacturing capacity they have with TSMC. With all the product lines AMD has, its just not enough.

[@SilkenLuna]

Yes, as far as I remember AMD had their K12 in development since around 2011. But that's what we know publicly. They could have started earlier. I guarantee when they were working on Zen, they had to just pause it. Not really interested in it, but it could be interesting.

[@noname-gp6hk]

They almost went out of business there toward the end of the pre-zen architecture. They had to focus on a single chip design to survive. They have more resources now than they did a decade ago.

[@Yusufyusuf-lh3dw]

I don't see why AMD would push for ARM for windows ecosystem as their main bread and butter market x86 will be directly in competition with that ARM chip. On windows, ARM is pretty much a work in progress at a very early stage, with a vast majority of the ecosystem lacking ARM support. But as you said AMD might find the mobile market interesting. It's also a market to grow the revenue much more than anything else.

[@Mkungaa]

Well, that is how you lose to Qualcomm/Nvidia/Mediatek/etc. You think x86 business is enough, it is your bread&butter and nothing will disrupt it. And then all of a sudden one Qualcomm/Nvidia/Mediatek/etc. will develop a really capable chip for laptops(and maybe desktops) with extra features like 5G connectivity, 2-week standby mode, etc. And all of sudden Microsoft decides to ditch you, since you do not have decades old expertise to build ARM chips like those vendors.

[@Yusufyusuf-lh3dw]

@Mkungaa well. Not really. Windows on ARM is quite primitive in its ecosystem-enabling. It will remain incompetetive for a couple of generations. The main advantage of ARM is its battery life, which x86 is steadily closing in. For example when Apple first introduced M1 MacBooks, MacBooks used to offer about 15hours of battery life compared to intels 4 hours or 5 hours. But now Apple MacBooks give about the same battery life, but both intel and AMD give around 10 to 12 hours. So the lead has come down drastically. Having 5G integration and a week of standby time are not very difficult to achieve on x86 platform.

[@Mkungaa]

@@Yusufyusuf-lh3dw even you admit that "It will remain incompetetive for a couple of generations", which means that Windows on ARM will eventually become competitive since big players like Microsoft/Qualcomm/Nvidia&Mediatek are really invested in it. And when that moment comes, you can not pull out of nowhere competitive ARM chip all of a sudden. In order for AMD to be competitive in ARM race in the future they need to start designing and iterating on chips/firmware/software yesterday. And looks like auntie Lisa ain't stupid enough to become "late to every party" like Intel.

[@richard.20000]

Because x86 as CISC is old and they already pushing it beyond limit. Imagine steam machine was great invention in 19th century but you don't want steam machine in your car in 21st century. Sure you can put electronic control unit etc. to steam machine but it would have a lot of disadvantages still. RISC was designed as successor of older CISC computers. And RISC really killed all other CISCs (Motorola 68K, Z80, 6502) with one exception x86 CISC. Why? Because modern x86 CISC CPUs today are 99% RISC-like internally while pretending they are x86 CISC at the outside. Problem is the CISC-to-RISC pretending overhead - it cost transistors, engineering workhorse and power consumption (x86 instruction has 1 to 15 byte length, so decoding 6 instruction in parallel leads into polynomial combination explosion which is solved in modern AMD Zen and Intel via sophisticated predictors..... where ARM has ZERO transistor spend because every ARM instruction has always 4 bytes so you can decode 1000 instruction per cycle with zero effort). Apple abandoned x86 because its overhead is huge and this is not suitable for mobile devices. Second important reason is that ARM is not closed platform, ARM allows to buy a license and create their own CPU.

[@Mkungaa]

@@Yusufyusuf-lh3dw "Having 5G integration and a week of standby time are not very difficult to achieve on x86 platform" Ahahaha, "modern standby" fiasco? Having 5G is such an "easy task" that both Intel and Apple spectacularly failed to replace Qualcomm modems. That is what I am talking about - Qualcomm has decades old expertise in modems that it became impossible to catch up with them. Best time for AMD to start gaining expertise in ARM was yesterday. Second best time is - now.

[@mikeb3172]

9-15W ARM with linux/proton is the question.

[@chu-icehugehard1820]

Current Windows on ARM doesn't show significant market advantage under Qcom & MS exclusive agreement, as well as other potential high performance ARM SoCs specifically for PC applications. But it could be better if Microsoft open WoA to makers of low end ARM chips like MediaTek, Allwinner and Rockchip, and encourage OEMs to make basic Windows laptops and tablets less than $200, in order to compete with Chromebook in education & entry-level markets.

[@chu-icehugehard1820]

Also, if M$ want to increase WoA's market share, they should open it to Macs with Apple Silicon, the most popular ARM PC platform at this time. Technically it's possible to run on Macs using a minimal Darwin OS to virtualize CPU+UEFI+TPM and the interrupt controller, and redirect other hardware to the VM through IOMMU.

[@mrhassell]

The day when using RISC based chips, to emulate CISC based arrives, is the day that mobile and desktop systems reach complete integration. At the moment, its possible, but performance is far from ideal.

[@JohnSmith-qt4pv]

That's effectively what modern complex ISA processors do. The decoder takes the instructions and converts it into a series of opcodes very similar to RISC instructions. It's also a fact that there aren't any true RISC processors around outside some microcontroller type stuff. Even arm ISAs have complex instructions such as multiply.

[@EdDale44135]

The K6 and then Athlon chip was based on NexGen chip that AMD bought the company were a RISC processor that used a CIDC translator. en.wikipedia.org/wiki/NexGen

[@c3r1c38]

Also remember that AMD bought Xilinx, and they have several ARM chips already out, so it's not like AMD is completely starting from scratch...

[@donh8833]

One of the last bastion hold outs of CISC is legacy apps. Among them are games. Compatibility is very very iffy. For example to optimize performance, a piece of software the developer might go down to the metal and hand tweek register use, or write assembly. These are not eadisily ported. Case in point: i can set a flag in the math coprocessor that allows me to alter the number if bits precussion a floating point uses. "I know im sending you a long double 80bit but i only need 60 bits precission". I know im sending you a 40 bit but i only need 32 bits of that". ARM 12:01 doesnt support hardware level commands like this because it makes the chip more complex. And then theres performance issues. Commands in x86 are variable length. Just parsing them for comatibility is a slow process. I know i know "Apple did it". But most of those compatable apps do run significantly slower. But you dont need a ton of horsepower to run a word processor or power point.

[@RealLifeTech187]

Just imagined AMD would leave behind x86 for ARM that would make Intel the sole supplier of x86 which would be a major step towards IBM-fication for Intel

[@vitormoreno1244]

Damm, that finish line was pure joy. Thank you Tom and team.

[@Truemann45]

Tom said stop being cringe for 5 minutes plz

[@Truemann45]

Ass

[@hilmyakatsuki1665]

People forgotten that AMD has an arm prototype when they were working on ryzen for the first time. They killed that for ryzen to come back from the disasters.

[@123Andersonev]

everyone seems to forget that arguably outside of OEMs who are incentivised to make more and more devices, consumers already have enough processing power in their pocket, chip shortages are common, what if you piggy back your mobiles CPU and upgrade that once every x amount of years and interface PCIE lanes off of it for other more intensive tasks, that's ultimately where ARM really becomes the gamechanger but you need the OS support to make it happen.

[@Dave-dh7rt]

I cant believe Intel said AMD was in the “rear view mirror” like AMD is far more efficient on a cheaper node lol. Yeah, intel has better multi core performance but for gamers and people that game and do productivity, get the 7800X3D or the 7950X3D lol. The sales are insane. 140-160W vs 400W with similar performance is INSANITY.

[@paul1979uk2000]

Intel fell asleep for so long that it was only a matter of time before a rival took over, if it wasn't AMD, it likely would have been ARM's, because Intel wasn't moving enough. In a sense, Intel should be thankful it was AMD that got them moving, because if it was ARM's, Intel might have a lot more problems now, but thanks to AMD, it's forcing Intel to wake up and invest in new cpu tech, which with the rise of ARM's, Intel could have ended up sleeping in decline as ARM's took over, now Intel has a fighting chance thanks to AMD waking them up. Still, I do find it remarkable how the tables have turned against Intel, after all, Intel had the performance and performance per watt advantage, now AMD has that, that's a major leap in AMD's favour, but then, Intel only has it's self to blame, it fell asleep, got greedy and wasn't pushing as hard as it did in the past.

[@KennyNguyen-rf9hl]

​@@paul1979uk2000😅😅

[@KennyNguyen-rf9hl]

​@@paul1979uk2000😅

[@grimacemcccheese]

I feel it makes sense for AMD to build arm. From a business perspective the market for personal computers has reached a high penetration. Everyone has a computer or two. One way to grow besides battling nvidia for ai and data center, is help Microsoft break into mobile again.

[@jeanchindeko5477]

Will be good to compare the PC enterprise laptop market against the Gaming PC market. In enterprise battery life is one of the things that can push many to switch to Windows on Arm. For the Gaming market, if AMD can deliver better performance/watts using Arm than x86, and keep integration with discreet GPU while keeping the cost down, it might be a big win for them.

[@HanSolo__]

Let's be reasonable. Steam Deck is out of reach of any competition.

[@jeevejavari8461]

Yea well said. It's in a league of its own

[@coolperson5479]

That's bs and you know it. The Lenovo Go and ROG Ally blow it out of the water either in price point or performance metrics. 10 tflops in a small package is insane considering heat issues. Pretty much has the power of a PS4 pro on the go. There's rumors the competition is working on one that can do 15-20 TFLOPs putting it in direct competition with the PS5 and series X. A portable with that type of graphical horsepower would of been impossible even a few years ago.

[@chryoko]

Why not going Risk V instead. And no need to pay IP licences to ARM ?

[@NadeemAhmed-nv2br]

Amd already holds arm ip from way back

[@pedrobaptista9213]

Some time ago there was talk that Intel was designing something in RISC-V as an alternative to ARM. But by now they should have already presented something

[@ContemplativeCat]

ARM has momentum, but long term I think you'll see RISC V become the dominant architecture, because its open source.

[@courtlaw1]

Are we seeing the start of the phasing out of X86. What is preventing AMD and Intel from utilizing Risc instead of x86.

[@JCJW101]

Intel and AMD publicly said they're ready to do ARM after the first Apple silicon came out so it wouldn't surprise me if they both introduce ARM CPUs in the 25/26 timeframe and we reach ARM superiority over x86 even at the highend by 2030.

[@magottyk]

AMD vs Intel AMD vs Qualcomm What will we be talking about in the future, particularly mobile thin and light devices.

[@Trash2321]

Doubt it has a performance similiar to a m4

[@MooresLawIsDead]

Have you looked at the specs of the M4? Strix seems poised to be plenty competitive....and that's now. Sound Wave is 2026.

[@Trash2321]

@@MooresLawIsDead doubt it will come even close in terms of power usage vs performance. But we can only hope...

[@aelderdonian]

@@MooresLawIsDead I'm sure you've seen the Geekerwan microarchitecture analysis - the actual performance, not just specs are giving the M4 a ~55% IPC lead over the 14900k. The Ryzen 8845HS for example is ~40% slower than M4 while using 2.5x the power. The M3 seems to have had a bottle neck somewhere that caused problems, and the M4 solves that and adds additional refinements.

[@jurepecar9092]

Peak x86 is far behind us, peak ARM is fast approaching and RISC-V is the future I'd bet on.

[@2000jalebi]

For the Nintendo Chip, I wonder if they are going to go for the MediaTek processor we heard about on a 3NM process, with the AI units disabled much like AMD's Z1 series of CPUs. Remember Nintendo was the only portable gaming console, so they had no pressure for high performance. Now with this MediaTek processor, they can have the performance crown, thin and light form factor with high battery and still be maybe $499 beating other handhelds in price/perf.

[@arenzricodexd4409]

nintendo will not going for bleeding edge. we already heard that nvidia can offer nintendo better chip but they did not want that.

[@l3lue7hunder12]

Just a question, but is this going to be a pure ARM APU ? Because AMD showed back with the K5 that they know how to make a Risc CPU 5x86 compatible, so maybe the CPU is x86 compatible despite at least most cores being ARM based ? I mean 80% of the time you won't need more than an ARM, so why not just use 80% ARM based cores combined with 20% x86 like Zen 5c or something ? That kind of hybrid might also be able to execute both x86 and ARM instructions, and possibly be a solid foundation of similar versions using Risc-V instead of ARM ?

[@aarong9378]

Intel was an ARM licensee many decades ago (around the i860/i960 timeframe).

[@MrHav1k]

On the flipside, do you really think with all of Intel's struggles they could afford to divert resources to an ARM project that may or may not pan out when they know for sure there IS an x86 market that they need to win back from AMD? Intel could very well want to have something to respond this but simply aren't able to given their situation. IIRC their DPU/FPGA/SmartNIC thing they have already uses ARM IP so it's not like they've never worked with it on the design side before.

[@ux-ss6ds]

Which is the microarchitecture of Sound Wave, AMD or Arm stock core?

[@TrueThanny]

I think there's roughly zero chance that AMD would use a stock ARM design.

[@leophoenix1452]

Amd arm developing chips has been in our face all these while even before snapdragon thought about its arm chips for windows...

[@pcefbol6704]

Snapdragon X Elite puts a lot of pressure on Intel and AMD. Sounds like the only way to catch-up with Qualcomm and APPLE is to explore partnerships with ARM chip designers. The consumers finally will be able to appreciate multi day battery life along side with Linux running ARM natively. While Qualcomm deserves credit for this advancement, competition is healthy, ultimately benefiting consumers.

[@2muchjpop]

AMD has x86 also... They are the perfect company to make a hybrid architecture so Windows and other smaller programs can run in the background on ARM while games and other high end programs not programmed for newer architecture can run on x86

[@user-zh9kc7tw4n]

Interesting to hear that AMD is looking at ARM where many are looking at RISK instead so they are not tied in to ARM

[@dbreardon]

Didn't AMD have plans to product ARM many years ago but close that division? I thought I remember something about this in the early mid 2010's

[@blast_processing6577]

ARM consumes less power than x86/x64, right? If it's priced low enough, AMD's Soundwave APU could become very popular in the handheld and mini PC markets... although if they're looking to replace Qualcom in the Microsoft Surface it may not be priced very low at all.

[@jipillow1]

What are the chances that 3 generations of processors will be supported by AM5?

[@Sheerwinter]

So will it have an apu like 40cu igpu? Or atleast 20 to 25 cu igpu just like the strix halo? And any news from nvidia? And strix halo?

[@arenzricodexd4409]

@@brownie43212 there is no way you can pack 4060 performance within 25w. not even on 2nm process.

[@mraltoid19]

I'm surprised AMD or Intel don't make an ARM/x86 Hybrid architecture. Imagine an Intel Architecture where the E-Cores are replaced with ARM cores, and through virtualization, an OS running on ARM would be able to run x86 software semi-natively, and an OS running on x86 would be able to run ARM software semi-natively.

[@onlysublime]

I'd love to see a Surface Pro with an AMD ARM with AMD GPU. but Qualcomm is a moving target, unless the glacial Intel.

[@John.Philip.Tan876]

Took me a few days to think about this but... If Intel believes they're missing another boat and would jump into this ARM design thing, wouldn't other ARM designers like Qualcom and the like be hesitant to use their fabs for the same reason Nvidia wouldn't want to? Edit: Then again they used Samsung's fabs...

[@Violet-ui]

A very sound idea to make ARM now that Windows is moving to ARM as well. In a couple years most computers (laptops) will probably run ARM unless they're for gaming or work applications

[@oappi4686]

I would be really interested in ARM based AMD APU for linux use as I do a lot development to cloud and ARM based compute is cheaper than X86 based. Developing with ARM based machine when "app" will be deployed to arm based machine should streamline things as x86 only dependencies could be noticed during development. Hopefully they won't bucher it with very low max ram... if they allowed same limit than with x86 (256gb) I would be in line to buy one of those systems like dirty apple fanboys with their i products.

[@endlesslogins156]

Pat's going to need a wheelchair after that final comment. LOL

[@Ks-rei]

well arm + radeon is working, i mean samsung use radeon IP on their flagship SoC, so yeah just waiting if that will be true or not XD

[@rationallyright4626]

Samsung and Amd are (or at least were) working together for Exynos. I need to look into it further for any further comments.

[@stennan]

One can hope they release a single board computer (think raspberry pi), Google tv OS or Linux images managed by AMD. That way we can tinker with it and perhaps have a Nvidia shield competitor. 😊

[@KiraSlith]

This sounds great! I'd love a compact-ish AMD ARM console based on a chip like this, maybe a clamshell Android handheld that can also go console mode? "I'll make my own then" has worked out so well for everyone else Nintendo has rejected, might as well be the company to complete the humble pie trifecta.

[@ctjmaughs]

Are we getting Rumble and Frenzy as APUs?

[@neti_neti_]

Can Nova Lake Compete with ARM and RISC-V? Is the Era of X86-64 Coming to an End?

[@VeerMaharaj]

Wait for AMD to figure out how to mix x64 chiplets with arm chiplets on the same package. If they get that sauce right, it will make Windows x86/64 legacy compatibility a non issue.

[@Alakazam551]

Aren't the Qualcomm "Adreno" Gpu's (which are one of the strongest Gpu's for Arm) just an anagram of Radeon because they licensed Amd tech for their Gpu's years ago?

[@federicolowenstein]

microsoft also said that they see an arm xbox in the future. they could share an apu with the surface.

[@makke_macro]

have no sound and utube tries to force skip the video. whats u say!

[@jayfangRSA]

Competitively performant and cost effective. ARM Chromebooks / Android tablets are ARM ready would have use for NPU and significantly increase overall market size for this tech

[@mariolis]

I am always gonna use X86 on my main Windows PC But I would love to use an AMD ARM CPU on linux

[@Anonymous-sb9rr]

I think that server CPU's could benefit from ARM as well. When you have 256 cores, the available power per core is very low, comparable to those 15 W laptop CPU's.

[@thesupremeginge]

According to that cranky Thurrott guy and Mac book owners, Windows on ARM runs great. Of course, they're not gamers, but all I hear is that it actually works fine.

[@MnemonicCarrier]

Now is the time for the Linux desktop to shine!

[@CANEHURRICANE]

Enjoying this with my morning coffee keep up the good work

[@paul1979uk2000]

I suspect AMD wants to keep its options open when it comes to ARM's, maybe to directly compete with mobile and laptops with ARM's chips or depending on which way the wind goes with x86 and ARM's. As for the Switch 2, I don't think it would have been that difficult to have full Switch 1 compatibility even on an x86 chip, we already see how well that works thought emulation on something like the Steam Deck, and in the case of a Switch 2, they could have had a specialised chip to help speed things along when it comes to performance. I suspect Nvidia didn't want to lose that market as it's the only real market they've got in the console business and probably bidded hard on it, but there was a clear opening for AMD to take advantage off, even for Nintendo to have full games compatibility with Switch 1 games.

[@user78405]

there is one thing people get confuse about cisc modern cpus vs classic cisc cpu difference is CISC stand for complex logic meaning in over time, intel add more instructions to it that can do avx instructions that make you think its qualified RISC cpu when that line stop there for RISC lovers that if that was the case and true...then apple wouldn't stop supporting RISC programs if you think intel is risc cpu since core i7 era. when still not RISC cpu...RISC cpu have something gonna and advantage what cisco can't do in their registers ...work in parallel pipeline stages with all 32 registers simultaneous vs modern cisc cpu like amd ryzen and intel core ultra soon to be release does not do parallel registers when pipeline stages cut short and yes intel did try it with netburst architecture that didn't help for them when each thread have to wait for other thread busy to finish its registers and branch predictions in every port. a port is different from ARM and intel x86, for x86 , they split the pipeline from 16 stages into 10 ports for each core for 2 threads to finish the tasks and its not very good way under microscope how it process its code since its very simple to do for any developer when things done automatic by host OS library ...either from linux kernel and windows nt kernel who manage every x86 processor thread handling scheduler is not perfect but it works quite well currently than back then since windows 95.. but for x86...its possible to make cores parallel a work load but its a theory for big little architecture if branch prediction can split registers with big cores and small cores by parallel when doing avx256 bit code since both can process simultaneous in strange technique that not been tested yet on intel side but we will see when intel ultra architecture release soon enough to test that theory in latest game use 16 threads and force registers pipelines run parallel each other

[@theworddoner]

For a product that’s going to ‘kill’ Apple’s m4, it’s going to be at least two years late. By then it’s Apple’s m5 that they need to worry about. It’s not just ARM itself. They also need high memory bandwidth to feed the APU. I’m talking 400Gb/s at the minimum and 800Gb/s at the high end. Let’s just say that I’m not holding my breath for this product.