New 2024 Microsoft Announcements - CoPilot+ PCs w/ ARM

[BONES]

Anyone seen these announcements? It looks like Microsoft re getting very serious about AI and ARM. They are saying their new CoPilot+ PCs are 58% faster than an M3 powered MB Air, twice as powerful as previous gen ARM PCs and that their x86 emulator is 20% more efficient than it used to be. It also has an NPU that means a lot of AI tasks can run locally, rather than relying on the cloud.

They also rattle off a list of software that I could use in my job - Adobe CC, Blender and DaVinci Resolve, that runs natively on ARM now. No mention of DAWs, though. Still, interesting enough. You can check it all out at Engadget or The Verge or wherever.

https://www.youtube.com/watch?v=5JmkWJNng2I

[machinesworking]

Here’s the MS blog announcement.
https://blogs.microsoft.com/blog/2024/0 ... pilot-pcs/

Looks pretty cool, exactly what I expected. Hopefully they generate enough interest to have porting to MS Arm happen at at least the pace of Apples transition. I have no idea why MS never bought Sonar? Right now that would be a huge advantage the way that Logic helped force the transitions to x86 and Apple Silicon.

One cool thing that a trillion dollar company will do is force a massive pace on developing a competitive SOC, and that seems to have happened.

Hopefully some of this comes to the Surface, no fans and a touch screen with a full copy of Bitwig DP or Studio One is IMO a Logic for iPad killer.

[BONES]

I think you vastly over-estimate the size of the home studio market. I doubt it is anywhere on Microsoft's radar. The only reason Apple bought Logic was to try and emulate the success of Final Cut Pro, which it hasn't come close to doing. That is why it's so cheap these days.

Personally, I don't see the point in bothering with ARM. Once Intel gets their manufacturing process down to the same size as everyone else, they'll be a lot more efficient without the headache of having to port everything.

[machinesworking]

I think you vastly over-estimate the size of the home studio market. I doubt it is anywhere on Microsoft's radar. The only reason Apple bought Logic was to try and emulate the success of Final Cut Pro, which it hasn't come close to doing. That is why it's so cheap these days.

Yes it's so out of their radar they flew Bitwig developers to Redmond to work with them on touch support for Windows. Something not being the major part of the business in no way makes it not important to a company, otherwise Apple and Microsoft wouldn't sell computers.

Personally, I don't see the point in bothering with ARM. Once Intel gets their manufacturing process down to the same size as everyone else, they'll be a lot more efficient without the headache of having to port everything.

Yeah you don't understand the tech, we get it.
RISC uses less energy than CISC at the instruction level, it's not just the size of the chip. The move over literally will help slow down our destruction of the planet, not that it matters to these companies, but authoritarians stupid enough to believe we have no impact on the planet are a loud minority, even though they don't think they are. So it makes sense to market to the 70% who aren't science deniers.

[sprnva]

One thing I can't find in all the articles about these ARM laptops if if the RAM and storage are user replaceable/upgradeable. That's what is keeping me from going near Apple Silicon.

I'm not really in the market for a laptop, but an ARM-powered mini-pc (MacMini style) would be great if it was an actual computer that could be upgraded and repaired.

[machinesworking]

One thing I can't find in all the articles about these ARM laptops if if the RAM and storage are user replaceable/upgradeable. That's what is keeping me from going near Apple Silicon.

I'm not really in the market for a laptop, but an ARM-powered mini-pc (MacMini style) would be great if it was an actual computer that could be upgraded and repaired.

So one of the ways they're getting good performance out of RISC is through the SOC design, it's something that would be much harder with the CISC chips. So the reduced instruction set takes less power which means everything can sit right on top of each other. Here's where MS and PCs could come ahead though, creating a market for external RAM and GPU to be used along side the SOC. It's not something Apple would do, and it's not likely that external RAM and GPU would be as fast as the internal, but it would be close enough.
Apple is selling overpriced tower computers that are only as fast as their Mac Studios, with only certain PCI cards available, but not extra RAM or GPU, if MS and the third party developers added that to tower PCs that would be it. I think the days of user upgradable laptops are numbered though.

[BONES]

Yes it's so out of their radar they flew Bitwig developers to Redmond to work with them on touch support for Windows.

WTF do the Bitwig devs know about touch support?

Something not being the major part of the business in no way makes it not important to a company, otherwise Apple and Microsoft wouldn't sell computers.

I imagine they learned their lesson when they made a specialist TypeCover for music producers, for the first Surface Pro, and absolutely nobody was interested.

RISC uses less energy than CISC at the instruction level, it's not just the size of the chip.

That difference is less significant than the extra power needed for larger die sizes. As Intel moves to smaller and smaller die sizes, their power requirements and TDP will shrink, too. And Apple silicon has shown us that if you want the processing power, you aren't making huge power savings by switching to ARM. It was the one area where I thought the new Macs would be unassailable but they don't seem to have much, if any advantage in that area.

authoritarians stupid enough to believe we have no impact on the planet are a loud minority, even though they don't think they are. So it makes sense to market to the 70% who aren't science deniers.

The science of which you speak is really f**king bad at predicting next week's weather and you want to trust it to predict what's going to happen in 50 years? That's definitely an act of faith too far for me, I'm afraid. Any idiot can see the climate is changing but the climate has always been changing. You know, Ice Ages and stuff like that. It feels like incredible arrogance to me to think that humans can stop the climate from changing.

[SLiC]

I imagine they learned their lesson when they made a specialist TypeCover for music producers, for the first Surface Pro, and absolutely nobody was interested.

That's because it was a bit crap (same as the ones people have made for the iPad). The concept is cool though if anyone ever gets it right with expression.

https://www.reddit.com/r/Surface/commen ... y_surface/

[ghettosynth]

If these laptops, especially a surface, can run local LLMs at a decent rate, then I'll buy one on release day. What I'd really like is to replace my Go with something that is actually powerful and can do LLM inference, but I'm not holding my breath because you really need quite a bit of memory and I doubt that they will go that route. We'll see.

[machinesworking]

Yes it's so out of their radar they flew Bitwig developers to Redmond to work with them on touch support for Windows.

WTF do the Bitwig devs know about touch support?

Christ, it's not that difficult to understand, touch support for the Windows version of Bitwig, MS was interested enough to fly them to Redmond to help, and used them in upcoming announcement for Windows etc. You forget sometimes where I live, I know people on the development and marketing teams devoted to audio and video.

die sizes. As Intel moves to smaller and smaller die sizes, their power requirements and TDP will shrink, too. And Apple silicon has shown us that if you want the processing power, you aren't making huge power savings by switching to ARM. It was the one area where I thought the new Macs would be unassailable but they don't seem to have much, if any advantage in that area.



The science of which you speak is really f**king bad at predicting next week's weather and you want to trust it to predict what's going to happen in 50 years? That's definitely an act of faith too far for me, I'm afraid. Any idiot can see the climate is changing but the climate has always been changing. You know, Ice Ages and stuff like that. It feels like incredible arrogance to me to think that humans can stop the climate from changing.

All of that is just painful to read. We live in a time when science denying is so prevalent that people like you come up with your own facts, with no degree, no actual science to back it up. Unfortunately one of the downsides to the internet is the prevalence of misinformation, and people that back up the 3% (of likely mostly christian/religious scientists) who claim we have no effect on the environment, are commonly very confident in their lack of real knowledge. None of this is about your hunches, which as a non expert in the field is all you can do. What I believe is overwhelming consensus among experts, 97% is a hard number to debate with anything like logic, you're literally going on your "feelings". Dunning Kruger is a real phenomena, don't be that person.

In terms of x86 VS Arm, the numbers have always been there, Arm uses less energy at the same core speeds, and has traditionally had a hard time scaling up, but that's changing. In the mean time x86 is of course getting better, but here's the rub, none of this isn't "political" in a sense. Apple has rarely cared about scoring the best results on tests, so their Arm chips IMO have never been a great example, they love thermals, and tiny designs, the Mac Studio is a great example, they couldn't give a rats ass about the tower, it's not any faster than the Studio, and this is inherently because they don't really care about the tower market. They love abstractions like USB4/Thunderbolt, because it's less close to the core = less issues. I wish I was wrong, but the chips produced say otherwise. We can see from 64 core designs of Arm chips that this isn't because it's not possible, but because it's not profitable, or logical for them to market to the 1% of the user base that could use more power. In terms of internet debates, there are plenty of fudged numbers, because that's how it goes in both directions when Apple is involved (the politics of choice is an insanely prevalent phenomena) but it's impossible to deny logic, a reduced instruction set resulting in the same outcome would obviously use less power, and results back that up.

Microsoft and the hardware companies that build specifically for MS OTOH are something like 10 times the size or more. That 1% becomes a more significant market share. As they mentioned, 85% of the things most people use laptops and computers for are more than accommodated for with applications already ported to Arm Windows. The thing is power is everything on the PC side, so someone like Asus is going to instal a 48 core Arm chip in a laptop in the next 5 years, and/or someone will toss out the first 128 core Arm tower. Because if they corner the 1% that use that power it's profitable and worth designing a chip for.

Basically Arm is thermally superior, and that equals better scaling, Apple, (as well as the Arm supercomputers etc.) proved 5 years ago that you can scale. but it takes the PC side to really stretch it to it's limits in terms of consumer product.

[pekbro]

Parallel processing is the way to go, but these arm chips have a long way to go before they
can really be overly good for it. you need core counts to be in the thousands, not 10s.

*To acheive this, they need to simplify them,
so they are like the stream processors in your
gpu.

*btw, i did used to work with cray and ibm
supercomputers.

[machinesworking]

Parallel processing is the way to go, but these arm chips have a long way to go before they
can really be overly good for it. you need core counts to be in the thousands, not 10s.

*To acheive this, they need to simplify them,
so they are like the stream processors in your
gpu.

*btw, i did used to work with cray and ibm
supercomputers.

In terms of home computers though, the things we use, what are your thoughts?
It seems to me that Arm is the way forward, and it's likely that we're now going to see massive jumps in improvements now that MS is teaming up with PC manufacturers to push it to it's limits.

[pekbro]

No doubt arm is a good way to go for home computing, there are already some very powerful 'relatively' high core count chips that are causing some problems for linux apparently. I doubt these chips will make it into consumer devices in the near future though. What is it, like 192 cores vs 10

Personally, I think the CPU should be probably dropped in favor of the GPU. CPU manufacturers will fight tooth and nail against that though, as it would likely spell the end for them. The concept of the CPU is mostly just a bottle neck these days and an
antiquated technology.

[machinesworking]

No doubt arm is a good way to go for home computing, there are already some very powerful 'relatively' high core count chips that are causing some problems for linux apparently. I doubt these chips will make it into consumer devices in the near future though. What is it, like 192 cores vs 10

Personally, I think the CPU should be probably dropped in favor of the GPU. CPU manufacturers will fight tooth and nail against that though, as it would likely spell the end for them. The concept of the CPU is mostly just a bottle neck these days and an
antiquated technology.

Interesting take, recently I've been messing around with Stem splitting software that uses the GPU if you want it to, to work it's algorithms on the stereo file, it's easily 3-5 times faster.

I thought the same thing about RAM, at some point what's the reason to have a separate memory space for random access?

[pekbro]

Well you only need to store something if you have no room to carry it. I think a good way to think about the cpu paradigm, would be to consider the neural networks that have been touted so much recently, and realize that something like that can exist infinitely across any scale. You don't need a cpu or memory when everything is everywhere all at once.

*Anyway, definitely beyond the scope of the present, not to mention this topic.

*So this may be hard to see how this might
work, but science is beginning to understand
that there are laws governing information,
just like the laws of physics, these laws
will be leveraged towards computing in the
years to come to account for systems like
i have described. At least, that would be
my guess.