I don't see why Sound Wave would have any advantage, even efficiency, over a similar Zen 5/6 design. Microsoft must really want ARM if they're having this chip made.
personally i totally understood why AMD gave up on its last attempt - the A1100 opterons - about 10 years ago in favor of the back then new ryzen architecture:
but what i would really like to see: an ARM soc/apu on an "open"*) (!) hardware-platform similar to the existing amd64 pc hardware.
*) "open" as in: i'm able to boot whatever (vanilla) arm64 linux-distribution or other OS i want ...
i have to add: i'm personally offended by the amount of tinkering of the firmware/boot-process which is necessary to get for example the raspberry pi 5 (or 4) to boot vanilla debian/arm64 ... ;)
br,
a..z
ps. even if its a bit o.T. in this context, as a reminder a link to a slightly older article about an interview with jim keller about how ISA no longer matters that much ...
It is easy to talk after the fact. The giant market opener for ARMs was Apple, many times in business it is better to he a follower once a big market arises.
I don't think AMD had the money to properly execute on both Zen and this K12 ARM chip. So they chose the more safer bet of Zen which seems to have worked out really well for them.
AMD was being pummeled by Intel during the time Jim was there and was only the giant we know today when Jim left. AMD did this, mostly, by x64 server market. So AMD did what it could to get around intel's and apple's monopolies. Intel on the server end came in at a higher price, and often worse performance or more power usage, or both. I'm not sure who AMD would have sold ARM to in 2010's. Apple didnt want their product and made their own, cell phone companies were cozy with the established ARM vendors, MS wouldnt launch an actual ARM laptop for years, data centers didn't want it, etc.
Look at intel's various arm or embedded offerings it keep canceling. It can't find buyers. Qualcomm and Samsung other vendors just keep eating up sales in ARM.
Now I imagine AMD sees ARM servers as the future and wants to make sure not to be left behind, on top of ARM desktop/laptop and further embedded.
I think this mostly a sign the world is now moving away from the old x86/64 system that ruled technology for so long. AMD is needs to stay competitive here.
Consider that Amd was not far from bankruptcy. They couldn't even execute on their gpu chips consider that they were the duopoly with nvidia and mostly missed the ai wave. Do they even have the capacity to work on arm on top of that?
Well, I'm eager to use it. For my home server I use an old power-hungry Epyc 7B13. It's overkill but it can run a lot of things (my blog, other software I use, my family's various pre-configured MCPs we use in Custom GPTs, rudimentary bioinformatics). The truth though is that I hate having to cross-compile from my M1 Mac to the x86_64 server. I would much rather just do an ARM to ARM platform cross-compile (way easier to do and much faster on the Orbstack container platform).
So I went out looking for an ARM-based server of equivalent strength to a Mac Mini that I could find and there's really not that much out there. There's the Qualcomm Snapdragon X Elite which is in only really one actual buyable thing (The Lenovo Ideacentre) and some vaporware Geekom or something product. But this thing doesn't have very good Linux support (it's built for ARM Windows apparently) and it's much costlier than some Apple Silicon running Asahi Linux.
So I'm eventually going to end up with some M1 Ultra Studio or an M4 Mini running Asahi Linux, which seems like such a complete inversion of the days when people would make Hackintoshes.
Better (or simply more) ARM processors, no matter who makes them, are a win. They tend to be far more power-efficient, and with performance-per-watt improving each generation, pushing for wider ARM adoption is a practical step toward lowering overall energy consumption.
Meh, performance-per-watt is not what everybody wants. I only want it in that it affords more raw performance by allowing more watts to be pumped through it without thermal overload. But if that can't actually happen then I'm still more interested in x86. Sure the lights dim when I turn my PC on, but I want the performance.
I don't think I'm using x86 for anything anymore. All the PC's in my home are ARM, the phones are ARM, the TV's are ARM and even the webservers I'm running are ARM nowadays.
Wow. This could really be a big deal, especially if it’s more of an openly available product than what Qualcomm has on offer.
For me personally I’d love it if this made it to a framework mainboard. I wouldn’t even mind the soldered memory, I understand the technical tradeoff there.
I want a hybrid APU, perhaps an x86 host with ARM co-processors that can be used to run arm64 code natively/do some clever virtualization. Or maybe the other way around, with ARM hosts and x86 co-processors. Or they can do some weird HMP stuff instead of co-processors.
A common trend in audio systems is that the market cap is too small for economies of scale when it comes to commodity parts like processors. There are a handful of audio-specific chips that are common but processors are not one of them (any more).
Could be an interesting chip for a future Raspberry Pi model? With Radeon having nice open source drivers, it would be easy to run a vanilla Linux OS on it. The TDP looks compatible as well.
Oh I hope the price is low enough that this be a real media box chip competitor fir streaming devices. Nvidia Shield Tegra chip from 2015 is still one of the best in this space. And with nvidia making all the AI money is not interested in making a new device. Apple TV the only real alternative does not support audio passthrough so is not as open as android or Linux media boxes.
I think Amlogic, Mediatek and Qualcomm all have SoC which are significantly better than the Tegra for this use case. It’s just that the market barely exists as most consumers use their tv directly so no one really wants to make a media box anymore.
Yeah, whether AMD is willing to go after some market low end has long been TBD. Intel's N100/N97/N150 is everywhere & very affordable, seemingly, based off some of the system prices. AMD doesn't have anything remotely like it.
The chip here is an interesting mix. Fast ddr5-9600! But less GPU CU's than most APUs: 4 down from 6. But if it comes with the other fixings like a good video engine & AMD's very good drivers it could be a real win.
Also a little hopeful that AMD rebadging it's Zen 1 and Zen 2 chips again might possible open up some decent low end space, but Sound Wave with more modern solutions would be a very nice to have power efficient low end.
> Memory support is another highlight: the chip integrates a 128-bit LPDDR5X-9600 controller and will reportedly include 16 GB of onboard RAM, aligning with current trends in unified memory designs used in ARM SoCs. Additionally, the APU carries AMD’s fourth-generation AI engine, enabling on-device inference tasks
128-bit LPDDR5X-9600 is about 150 GB/s, that's 50% better than an Orin NX. If they can sell these things for less than like $500 then it would be a pretty decent deal for edge inference. 16 GB is ridiculously tiny for the use case though when it's actually more like 15 in practice and the OS and other stuff then takes another two or three, leaving you with like 12 maybe. Hopefully there's a 32 GB model eventually...
49 comments
[ 2.2 ms ] story [ 63.1 ms ] threadARM isn't nearly as interesting given the strides both Intel and AMD have made with low power cores.
Any scenario where SoundWave makes sense, using Zen-LP cores align better for AMD.
imho. (!)
i think this would be great!!
personally i totally understood why AMD gave up on its last attempt - the A1100 opterons - about 10 years ago in favor of the back then new ryzen architecture:
* https://en.wikipedia.org/wiki/List_of_AMD_Opteron_processors...
but what i would really like to see: an ARM soc/apu on an "open"*) (!) hardware-platform similar to the existing amd64 pc hardware.
*) "open" as in: i'm able to boot whatever (vanilla) arm64 linux-distribution or other OS i want ...
i have to add: i'm personally offended by the amount of tinkering of the firmware/boot-process which is necessary to get for example the raspberry pi 5 (or 4) to boot vanilla debian/arm64 ... ;)
br, a..z
ps. even if its a bit o.T. in this context, as a reminder a link to a slightly older article about an interview with jim keller about how ISA no longer matters that much ...
"ARM or x86? ISA Doesn’t Matter"
* https://chipsandcheese.com/p/arm-or-x86-isa-doesnt-matter
Could be a revival but for different purposes
Look at intel's various arm or embedded offerings it keep canceling. It can't find buyers. Qualcomm and Samsung other vendors just keep eating up sales in ARM.
Now I imagine AMD sees ARM servers as the future and wants to make sure not to be left behind, on top of ARM desktop/laptop and further embedded.
I think this mostly a sign the world is now moving away from the old x86/64 system that ruled technology for so long. AMD is needs to stay competitive here.
I believe Jim Keller is now working on RISC-V which could take the server market by storm in the next 5 years or so.
There are already RISC-V server offerings:
https://labs.scaleway.com/en/em-rv1/
https://chipsandcheese.com/p/evaluating-the-infinity-cache-i...
So I went out looking for an ARM-based server of equivalent strength to a Mac Mini that I could find and there's really not that much out there. There's the Qualcomm Snapdragon X Elite which is in only really one actual buyable thing (The Lenovo Ideacentre) and some vaporware Geekom or something product. But this thing doesn't have very good Linux support (it's built for ARM Windows apparently) and it's much costlier than some Apple Silicon running Asahi Linux.
So I'm eventually going to end up with some M1 Ultra Studio or an M4 Mini running Asahi Linux, which seems like such a complete inversion of the days when people would make Hackintoshes.
- Low power when only idling through events from the radio networks
- Low power and reasonable performance when classifying objects in a few video feeds.
- Higher power and performance when occasionally doing STT/TTS and inference on a small local LLM
It looks like it is intended to run Windows Arm.
But, wouldn't it make more sense for amd to go into risc-v at this point of time?
For me personally I’d love it if this made it to a framework mainboard. I wouldn’t even mind the soldered memory, I understand the technical tradeoff there.
https://en.wikipedia.org/wiki/Project_Denver
The chip here is an interesting mix. Fast ddr5-9600! But less GPU CU's than most APUs: 4 down from 6. But if it comes with the other fixings like a good video engine & AMD's very good drivers it could be a real win.
Also a little hopeful that AMD rebadging it's Zen 1 and Zen 2 chips again might possible open up some decent low end space, but Sound Wave with more modern solutions would be a very nice to have power efficient low end.
They will be very happy.
128-bit LPDDR5X-9600 is about 150 GB/s, that's 50% better than an Orin NX. If they can sell these things for less than like $500 then it would be a pretty decent deal for edge inference. 16 GB is ridiculously tiny for the use case though when it's actually more like 15 in practice and the OS and other stuff then takes another two or three, leaving you with like 12 maybe. Hopefully there's a 32 GB model eventually...