Windows ARM Dev Kit 2023
microsoft.com
microsoft.com
If you don’t care about Windows software, Chromebooks are popular and cheap. They don’t have much in the way of security issues, don’t need much management, and store your setup in the cloud.
If you care about the software you probably don’t want ARM. The translation layer is supposed to be quite noticeable unlike Rosetta 1 & 2 on Macs.
And PCs no longer start at $750 so ARM can’t undercut by much. I’ve seen sub-$200 PC laptops. Sure they’re garbage but PCs are at basically every price point now.
Apple had it easy. “Our processors are way better than our old ones, we’re doing this, enjoy or leave.”
MS doesn’t have the kind of position to force this. So why would the market move?
Well enough that it doesn't matter that much you don't have a native Google Chrome, for instance. AMD64/x86-64 version runs just fine.
If Apple released a cheap M1 laptop for like $500 it would be game over for laptop PCs and probably would make a huge dent in the Chromebook market.
We know there's a Windows for RISC-V in the works, as the foundation's talks at the Summit referenced it multiple times.
ARM-based, hardware-wise, there's only Qualcomm-based laptops. Qualcomm is very obviously not going to be making SoCs based on ARM CPUs for long.
They have been RISC-V members for a while, they are being sued by ARM, and they've clearly expressed their intent re: RISC-V in their talks at the RISC-V Summit.
As for other SoC vendors, I doubt there is interest. Not now that RISC-V is already taking off.
I have never cared about Apple hardware to begin with.
Till then, it does not exist for Windows purposes.
https://www.zdnet.com/google-amp/article/microsoft-finally-a...
or an article from Microsoft themselves: https://support.microsoft.com/en-us/windows/options-for-usin...
Feel free to move the goalposts to "non-virtualized, even though you didn't specify that originally :)
I doubt it won't be bigger than ARM ever was or ever will be; within just 3 years from now I expect it to be taking on x86.
For the "up the value chain" bit, things like the very-early-stage PineTab-V:
https://www.pine64.org/2023/04/01/march-update-tablet-bonanz...
Plus the Tenstorrent (Canadian based) server oriented stuff:
https://www.tomshardware.com/news/tenstorrent-shares-roadmap...
For the supercomputer bit:
https://www.tomshardware.com/news/intel-to-explore-risc-v-is...
and:
https://www.nextplatform.com/2022/06/09/strong-showing-for-f...
Whether any of that will get traction is clearly a "who knows" thing, but the signs seem to be pointing in the positive direction.
Assuming hardware was developed alongside the specs, which didn't change much near the time of ratification, then 2-3 years need to be added for actual products. We are about to see the first wave.
We know Tenstorrent (CEO: Jim Keller) has a design they expect to release in 2024 with performance competitive with projected Zen5 (also 2024), but using less power. That's Ascalon, a design led by the same person who was M1's chief architect at Apple.
We also know SiFive, MIPS, Rivos, Ventana are each working on their own very high performance designs, and have teams full of extremely capable people who have a history of success in high performance chips. That's a lot, and that's just what we know of. There's necessarily more efforts beyond the surface, which we don't yet know about.
Ventana's Veyron might be the first to market, with server chips expected 2023H2.
Microsoft and Google, their RISC-V efforts we already know about.
Android RISC-V port, already merged upstream, was an effort funded by by Chinese SoC vendors. Capable chips and smartphones will follow, this is an inevitability.
As for AWS, GCP and Azure (sorted them in actual order of relevance), they will definitely start offering RISC-V support at some point and then we can see whether it picks up.
But high performance server hardware is a prerequisite for this to be possible, thus my parent post.
Citation needed.
Apple is the most successful ARM platform as of now. Everything else is minor compare to them. I would really like to see that we have RISC-V as an alternative platform but right now it is a nice toy to play with and even for the toy usecases the selection is very limited.
I honestly do not see it happening.
Until then, whatever hardware Apple has is irrelevant to the Windows discussion.
They already care about RISC-V as I explained in other posts in this discussion, and there's not even hardware yet.
Hardware is finally about to show up, also explained, to you specifically, in another post in this discussion.
RISC-V is inevitable.
While it does seem pretty likely, "inevitable" is perhaps overstating things.
If RISC-V does make solid inroads into servers and supercomputers, then (perhaps ironically?) the US gov may start doing funky things to suppress it. They don't want China successfully countering the recent chip-making suppression efforts, which RISC-V could enable.
Probably not something we'll see in 2023 though, so that's totally wild guessing. ;)
The momentum has grown far beyond the point of no return.
As we can see it already happening, it is far more likely for me to be struck by lightning than RISC-V not take over the world.
>If RISC-V does make solid inroads into servers and supercomputers
See recent talk from BSC's Mateo Valero[0]. EU's supercomputer roadmap already is RISC-V centric, and their researchers already have a few generations of finished RISC-V designs.
>then (perhaps ironically?) the US gov may start doing funky things to suppress it
They cannot. RISC-V Foundation is based in Switzerland.
0. https://www.youtube.com/@tenstorrentinc (it's somewhere in the recent "Nerds Talking to Nerds About RISC V", which has a load of good talks)
You're not taking into account the difference between theory and reality.
"In theory" the US government "cannot" due to technical aspects like the RISC-V Foundation being based in Switzerland.
"In reality" though, the US government could (for example) introduce targetted sanctions on companies having anything further to do with RISC-V.
The vast majority of those companies would terminate their RISC-V involvement on the spot, as they have substantial US involvement with their businesses.
The remaining ones could indeed keep on pursing RISC-V, but the whole "take over the world" thing would be either dead or have its timeframe pushed out by many years.
Note - I'm trying to illustrate that "inevitable" isn't correct when powerful players are involved. So the above example is just the first one to spring to mind.
RISC-V is inevitable.
There are also ARM-based servers.
Interestingly, none use microarchitectures from ARM themselves, but are instead the work of ISA licensees.
If this could be a solid docker host, it would be a really interesting home server. It still cannot six months later and instead I am building a home K8s cluster with Dell Micro 3060s with i5-8500t CPUs. My per unit cost with 64GB RAM, 1TB NVME, and 2.5Gbe is under $310.
If you're looking for a free Unix-y environment to play with, with >11000 binaries packages available.
You need to use the mini-DisplayPort for video output, not Type-C. This is a UEFI limitation on the machine.
The hardware is simply not competitive at the price, even if you ignore the hassles of hardware support.
99% of their uses are for appliances/phones, so it makes sense that they're still totally immature at being a computer. But weighed against how much time arm has had to try to improve their world, looking at how plodding progress has been, how broadly chaotic & grossly inconsistent the ecosystem is, it's still a stunning testament.
Computing in general could have been this bad. It's such a miracle we got the PC. The Gang of Nine just wanted to sell hardware but they launched a major Democratizing phase change, one we still have yet to see reproduced.
That hasn’t existed in ARM land. Seems to be the wild west in comparison. So every board requires custom bring-up.
Just looking at the 1980's there were tons of operating systems being created & getting used. That there could be so many OSes was part & parcel to it suddenly being possible to create so much hardware.
I see this as a time of diversification & enablement, a great opportunity, that yes, latter Microsoft did emerge top of the heap on. But they were more incidental a player than anything else, in my mind; they rode the wave that was happening. And they contributed very little. DOS with TSR driver stuff wasn't exactly a place for cooperation; it too was wild west.
They didn’t cause it initially, but they sure took advantage of being the thing IBM chose.
The N100 is interesting, but it's capped at 16GB RAM by Intel.