In a blog post today, amid the usual PR blabber, Microsoft announced a number of improvements that will be coming to Windows 10 on ARM later this year. The most important ones:
We are excited about the momentum we are seeing from app partners embracing Windows 10 on ARM, taking advantage of the power and performance benefits of Qualcomm Snapdragon processors. We heard your feedback and are making Microsoft Edge faster while using less battery, and announced that we will soon release a native Microsoft Teams client optimized for Windows 10 on ARM. We will also expand support for running x64 apps, with x64 emulation starting to roll out to the Windows Insider Program in November. Because developers asked, Visual Studio code has also been updated and optimized for Windows 10 on ARM.
Adding support for 64bit x86 applications to Windows 10 on ARM’s emulation technology will surely help with application availability, but so far, OEMs and Qualcomm haven’t really managed to set the world of PCs on fire, so performance will remain questionable.
Was waiting for this article to be posted here.
We are going through really interesting times. For decades Intel tried to scale down, while ARM tried to scale up.
Intel has some really low power processors, and even i7 ones that could work without a fan:
https://ark.intel.com/content/www/us/en/ark/compare.html?productIds=196452
For example this one can be configured to 4.5W, which is actually on par with a Raspberry PI.
https://raspi.tv/2019/how-much-power-does-the-pi4b-use-power-measurements
However that thing costs $400. That is a lot of money, in fact you might get a whole ARM system.
That is why it seems to be more interesting to watch ARM climb up. There is still long ways to go, and their “workstation” chips are nowhere near Intel or AMD territory, but they could easily conquer portable computers.
As you may know already, but here for general knowledge, the fastest supercomputer now runs on ARM. Ampere Max and Graviton 2 ARM CPUs are in Intel ballpark, the Neoverse N2 ARM design is specced to even go beyond them.
For super-computers, “fast” is the wrong word – a better word would be “wide” (and “high throughput” would be even better).
A single slow CPU will be slow with bad throughput, and ~160000 slow CPUs working in parallel will be slow and have high throughput.
For something that is embarrassingly parallel, “slow and wide” is extremely good. That’s why Fujitsu A64FX is more like a GPU (lots of weak cores with very wide SIMD) and isn’t comparable to any normal CPU.
For anything that isn’t embarrassingly parallel, “slow and wide” is worthless due to Amdahl’s law. E.g. for something like a browser trying to handle a JavaScript infested web page, a cheap phone will probably give better performance than every super computer (because super computers are slow).
Yes,
With ARM you can get cheap and okay performance, and scale up for massively parallel super computer tasks.
Before this, there was a push for GPUs (and same with BitCoin/crypto currencies).
Before that, there was some with PS3 (with the Cell processors).
However on the desktop, and especially for workstation tasks, there is still a long way to go. This has came up before, and might warrant resharing:
https://www.anandtech.com/show/15733/ampere-emag-system-a-32core-arm64-workstation/4
Brendan,
+1, great points.
Sometimes people overlook this, but many supercomputers achieve high TFLOP using fairly average CPUs. For highly parallel workloads that can scale by going extremely wide, it can be more economical and energy efficient to scale using more slower cores rather than faster cores. Graphics cards are a good example of this too, they’re actually pretty slow per core, it’s just that you’ve got so many of them.
I am more optimistic about the future of extremely wide architectures over extremely fast ones.
Isn’t this a chicken and egg comment?
We need the return of netbooks. When you compare $1000+ laptops, ARM is currently a hard sell. I want a cheap sub-notebook for typing duties in a VAIO P-like form-factor but all current options (e.g. GPD Pocket 2 / P2 MAX) are too expensive, given that it wouldn’t be something to replace my current main laptop / desktop.
Adobe Flash is dead. Make £200 ARM netbooks and _then_ the platform will take off.
Tablets became a thing and killed off the netbook market.
We just need devices of all form factors that can run ARM Windows. Doesn’t matter if they’re netbooks or desktops or tablets….
VAIO P-like form-factor
Actually VAIO-P was a relatively expensive device ($800 – $1899).
But on the other hand Microsoft Surface Pro X price is ridiculous. They charge $200 for extra 8GB of RAM and $300 for extra 128GB storage. For this difference you can buy a phone with both of these options included 🙂
Make £200 ARM netbooks
There is an $199 ARM laptop
https://www.pine64.org/pinebook-pro/
Get a chromebook.
Why don’t M$ create a bytecode for Windows, so when you install a program this bytecode is compiled/translated/converted to the binary code of the target system’s architecture?
They actually did this, twice.
The first one is the “Hardware Abstraction Layer”. Believe it or not, the device driver architecture in the NT kernel is (was) architecture agnostic.
The second attempt was .Net. But the infighting between Dev and Windows teams during the Longhorn / Vista fiasco killed this for the foreseeable future.
Currently they actually have universal like app support in the Store, but that will take a while to catch up.
It’s just a testament to the poor software architecture culture in Microsoft.
Meanwhile Apple/NeXT solved this issue over 3 decades ago…
It is a bit odd for some to claim the concept is dead, due to poor performance, when the platform has never really existed before.
We can’t expect performance issues to be resolved before the platform exists. There is no way MS or any other provider will be able to offer such a system without some performance issues or bottlenecks, they are good but they are not that good.
Making the platform available is the first part of making it usable, so Devs should blame a new platform’s performance for not adopting that platform, that is just code for “I can’t be bothered doing the stuff I need to do to make it work, let somebody else do the work”!