Windows 10 for ARM - Details and Limitations Revealed

[Trixanity]

59 minutes ago, Zodiark1593 said:

A slow transition over to ARM and UWP? As we're already emulating at Atom performance, native ARM performance is bound to be far higher, perhaps even approaching Intel's Core-Y chips. Cortex A-75 is coming soon as well. While I don't know the details, it's Meltdown vulnerability points to a more aggressive branch prediction capability, and possibly a wider architecture than seen in previous ARM designs.

I don't think ARM's designs can get anywhere near close to the Core architecture but it'll be good enough for the use case. I mean this isn't exactly meant to be high performance computing. It's web browsing, office work and content consumption with long battery life.

 

A75 is wider than A73 (despite belonging to the same family) but is the same width as A72. That means 3-wide instead of 2 and the pipeline length is the same. The branch predictor is the same too. However it moved from shared queues to individual queues making it more speculative which is probably why it's vulnerable to Meltdown.

[Apepa]

Since it breaks compatibility with older versions of Windows, not to mention games and software designed for x86 Windows, I have no idea why you would choose this over a Linux distro. 

[Apepa]

1 hour ago, Trixanity said:

I don't think ARM's designs can get anywhere near close to the Core architecture but it'll be good enough for the use case. I mean this isn't exactly meant to be high performance computing.

That's something I can't seem to find any info about: What sort of performance could you get out of an ARM/RISC chip if you allowed it to have the same TDP as a high-end x86 desktop chip?

[Trixanity]

46 minutes ago, Apepa said:

That's something I can't seem to find any info about: What sort of performance could you get out of an ARM/RISC chip if you allowed it to have the same TDP as a high-end x86 desktop chip?

I think it depends on the benchmark/workload. I think Apple could get close. Might even win some but current chips from Apple (or any other ARM chip designer) aren't meant to compete in that arena so it's difficult to tell. I am certain Apple intends to ditch Intel within a few years so they're working their way up already although the change might be gradual (eg MacBooks first).

 

I'd say to begin with that risc/cisc is a somewhat outdated terminology as far as I know. I mean x86 processors are at least a hybrid at this point. They're cisc on the surface for legacy reasons but translate it to risc-like instructions internally to do things faster and more efficiently. I don't know enough about the ARM ISA to say for certain. It seems to still be pure risc but can't say for sure.

 

There are ARM chips in servers so there are some comparisons to look at but I'm too lazy to look up the numbers. There are actually whole bunch of them and even more in the works. So it's certainly doable in some sense but I suspect ARM based servers are specific in use case and depends on instructions required. Some of the server guys on this forum can probably tell you a lot more about how it stacks up. I don't read up on ARM for servers too often. The only thing I can recall reading was AMD's brief flirt with ARM where they launched a server processor (think it was 2 or 3 years ago) that was dead on arrival. It was late and outdated when it launched and the performance wasn't there. In fact it was useless if I recall correctly. I think even if it had launched in a timely fashion it would have been a futile effort. AMD also cancelled their K12 ARM processor which was rumored to be even faster than Zen but that remains to be seen. I think if it had launched you'd have had a more clear answer. I'm very disappointed it didn't but I suspect the cost of keeping two separate processor branches going at the same time while having to compete in an uncertain market of ARM servers gave it the death blow.

 

I can try to look up some numbers tomorrow if you can't find any and are still interested. But now I'm going to bed

[mr moose]

1 hour ago, Apepa said:

Since it breaks compatibility with older versions of Windows, not to mention games and software designed for x86 Windows, I have no idea why you would choose this over a Linux distro. 

Windows store, MS account and familiarity would be the main reasons anyone would.  Why learn a new OS if you already use an MS account and email and everything you use is available.

[mr moose]

7 minutes ago, Trixanity said:

I think it depends on the benchmark/workload. I think Apple could get close. Might even win some but current chips from Apple (or any other ARM chip designer) aren't meant to compete in that arena so it's difficult to tell. I am certain Apple intends to ditch Intel within a few years so they're working their way up already although the change might be gradual (eg MacBooks first).

 

I'd say to begin with that risc/cisc is a somewhat outdated terminology as far as I know. I mean x86 processors are at least a hybrid at this point. They're cisc on the surface for legacy reasons but translate it to risc-like instructions internally to do things faster and more efficiently. I don't know enough about the ARM ISA to say for certain. It seems to still be pure risc but can't say for sure.

 

There are ARM chips in servers so there are some comparisons to look at but I'm too lazy to look up the numbers. There are actually whole bunch of them and even more in the works. So it's certainly doable in some sense but I suspect ARM based servers are specific in use case and depends on instructions required. Some of the server guys on this forum can probably tell you a lot more about how it stacks up. I don't read up on ARM for servers too often. The only thing I can recall reading was AMD's brief flirt with ARM where they launched a server processor (think it was 2 or 3 years ago) that was dead on arrival. It was late and outdated when it launched and the performance wasn't there. In fact it was useless if I recall correctly. I think even if it had launched in a timely fashion it would have been a futile effort. AMD also cancelled their K12 ARM processor which was rumored to be even faster than Zen but that remains to be seen. I think if it had launched you'd have had a more clear answer. I'm very disappointed it didn't but I suspect the cost of keeping two separate processor branches going at the same time while having to compete in an uncertain market of ARM servers gave it the death blow.

 

I can try to look up some numbers tomorrow if you can't find any and are still interested. But now I'm going to bed

I was just trying to find a way to say this.  But the the whole thing is just beyond my words.

 

The best I could do was: they are essentially the same with differences in compilers more than anything.   I believe one of the biggest reasons we see variance across ARM devices has as much to do with which compiler was used as it does with the particular hardware,  In other words two identical ARM processors can perform differently using the same software but each compiled with a different compiler.    If the traditional RISC side of processors could be streamlined with more uniform and predictable compilers and the processors where scaled up then there might be no difference or even a slight advantage to more conventional RISC chips.

[Apepa]

13 minutes ago, mr moose said:

Windows store, MS account and familiarity would be the main reasons anyone would.  Why learn a new OS if you already use an MS account and email and everything you use is available.

I don't know anyone who uses anything from the Windows store besides the sort of cross-platform games that are popular on non-Windows mobile devices. Besides, for the sort of basic "Cromebook" functionality this is aimed at, there isn't a lot to learn with something like Ubuntu, and people are used to using different OSes now with the advent of iOS and Android devices. Not to mention it's free.

[Apepa]

24 minutes ago, Trixanity said:

 

I can try to look up some numbers tomorrow if you can't find any and are still interested. But now I'm going to bed

I've found a few sever benchmarks where the ARM processors trail the x86 ones by quite a bit, but the TDP is also quite a bit lower. Not quite a definitive answer! 

[Guest]

2 hours ago, Apepa said:

Since it breaks compatibility with older versions of Windows, not to mention games and software designed for x86 Windows, I have no idea why you would choose this over a Linux distro. 

It isn't like Win32 disappeared on this version. You can still run x86 applications provided they aren't 64 bit or use an OpenGL version newer than 1.1 or requires hardware-accelerated OpenGL. DirectX 9-12 is still supported and 32 bit applications are still common place.

[Apepa]

1 hour ago, tjcater said:

It isn't like Win32 disappeared on this version. You can still run x86 applications provided they aren't 64 bit or use an OpenGL version newer than 1.1 or requires hardware-accelerated OpenGL. DirectX 9-12 is still supported and 32 bit applications are still common place.

"You can still run x86 applications provided they"...are compatible with WINE. It's the same sacrifice. Since you have to migrate to a new ecosystem anyway, you might as well move to one that's free.

[mr moose]

1 hour ago, Apepa said:

"You can still run x86 applications provided they"...are compatible with WINE. It's the same sacrifice. Since you have to migrate to a new ecosystem anyway, you might as well move to one that's free.

Except that majority (upwards of 80%) of users don't know Linux or anything about it. 

[Trixanity]

8 hours ago, mr moose said:

I was just trying to find a way to say this.  But the the whole thing is just beyond my words.

 

The best I could do was: they are essentially the same with differences in compilers more than anything.   I believe one of the biggest reasons we see variance across ARM devices has as much to do with which compiler was used as it does with the particular hardware,  In other words two identical ARM processors can perform differently using the same software but each compiled with a different compiler.    If the traditional RISC side of processors could be streamlined with more uniform and predictable compilers and the processors where scaled up then there might be no difference or even a slight advantage to more conventional RISC chips.

Oh certainly. I don't see a reason why one would be slower than the other simply because of terminology. As you say compilers can affect the results significantly (we even see that with Zen). Thing is though that there isn't really any competition in ARM mobile space so numbers are all over the place and so is developer support. They're all so architecturally different.

 

The server space is still catching up. Intel still overshadow many of their competitors in that area (at least in market share). Would be interesting to see a high performance 50 or 100W ARM chip in the desktop market though. As long as it supports the right instructions and contribute on compilers we could see some interesting things.

[Trixanity]

8 hours ago, Apepa said:

I've found a few sever benchmarks where the ARM processors trail the x86 ones by quite a bit, but the TDP is also quite a bit lower. Not quite a definitive answer! 

This could be interesting

 

No independent data but Qualcomm claims their chip is providing higher performance per watt and per mm^2 than Intel and general every metric, perhaps except absolute performance which I assume based on the fact they didn't show numbers in benchmarks but as percentages on various metrics. They deliberately avoided hard numbers. But if Qualcomm can pull off their claims it really speaks in favor of ARM.

[Apepa]

6 hours ago, mr moose said:

Except that majority (upwards of 80%) of users don't know Linux or anything about it. 

That's not really a barrier anymore, the main issue is what OEMs include on laptops. 

[Guest]

9 hours ago, Apepa said:

"You can still run x86 applications provided they"...are compatible with WINE. It's the same sacrifice. Since you have to migrate to a new ecosystem anyway, you might as well move to one that's free.

While yes its a more limited scope, it isn't like its moving you to a new ecosystem. If its a more casual user, its likely that most if not all of the software they use will fit in this scope. (Just imagine the frustration of them not knowing why things are different with a linux system) I do agree that Linux should become more mainstream, but I don't think this would be much of the reason to do so nor will help OEMs ship it.

[Misanthrope]

Actually not as limiting as I was imagining, might be able to run ok for those tablet/ultralight combos (the cheap ones of course)

[Apepa]

24 minutes ago, tjcater said:

While yes its a more limited scope, it isn't like its moving you to a new ecosystem. If its a more casual user, its likely that most if not all of the software they use will fit in this scope. (Just imagine the frustration of them not knowing why things are different with a linux system) .

I would have agreed with this 10 years ago, but not so much anymore.  MacBooks became hugely more popular off the back of the iPod craze of the last decade, and if anything I have a harder time with Mac OS than most non-techies do, because they don't interact with their computer beyond a basic selection of apps. And then there's the proliferation of iOS and Android devices, not to mention the way Windows has changed over the years (I remember people complaining about the Windows XP start menu).

 

And I'd also argue that the Windows Store Apps are a different ecosystem, since even extremely popular apps like Chrome aren't available, and I'd bet x86 compatibility is a lot more limited than they're trying to make out. Windows compatibility mode doesn't even work half the time even on the same platform.

[Guest]

1 minute ago, Apepa said:

MacBooks became hugely more popular off the back of the iPod craze of the last decade, and if anything I have a harder time with Mac OS than most non-techies do, because they don't interact with their computer beyond a basic selection of apps.

I do agree here, had to use macOS all through highschool and found it far less intuitive than Windows and some *nix DEs.

3 minutes ago, Apepa said:

And I'd also argue that the Windows Store Apps are a different ecosystem, since even extremely popular apps like Chrome aren't available, and I'd bet x86 compatibility is a lot more limited than they're trying to make out. Windows compatibility mode doesn't even work half the time even on the same platform.

I think you got Windows 10 S and Windows 10 for ARM mixed up. Win10A does support proper programs found on the web (Chrome/Firefox 32 bit should be included) unless it was configured in S mode (Then I'd whole heartily agree with you). They could have downplayed compatibility, but we'll need these devices first before we can put it under the gauntlet

[Trixanity]

11 minutes ago, tjcater said:

I do agree here, had to use macOS all through highschool and found it far less intuitive than Windows and some *nix DEs.

I think it depends on your perspective and experience. If you're technically illiterate then I think it might make perfect sense but it can be downright infuriating as a power user.

For example I loathe how everything is drag'n'drop; I find it a very inefficient way of doing things. I also dislike how it handles windows. So many annoying little things but I think it's good for the average user. It's like permanent training wheels in that's it's difficult to do something wrong which is a double edged sword.

[Apepa]

1 hour ago, Trixanity said:

So many annoying little things but I think it's good for the average user. It's like permanent training wheels in that's it's difficult to do something wrong which is a double edged sword.

I borrowed a MacBook last month while I was waiting for a replacement motherboard and I spent most of my time swearing at it. Oh, and it refused to create me a bootable Windows Installer USB drive, so I had to burn one to a DVD, and due to that Apple "Build Quality" I couldn't use the laptop while a DVD was burning because the chassis would flex and disturb the spinning disc.

 

I hate them.

[mr moose]

6 hours ago, Apepa said:

That's not really a barrier anymore, the main issue is what OEMs include on laptops. 

says you who is quite versed with technology.  HP sell Laptops with Linux, they are not very popular for the same reason.  People don't want to learn new stuff even when it's cheaper. We are talking about the general population here.  

[Zodiark1593]

5 hours ago, Trixanity said:

I think it depends on your perspective and experience. If you're technically illiterate then I think it might make perfect sense but it can be downright infuriating as a power user.

For example I loathe how everything is drag'n'drop; I find it a very inefficient way of doing things. I also dislike how it handles windows. So many annoying little things but I think it's good for the average user. It's like permanent training wheels in that's it's difficult to do something wrong which is a double edged sword.

At least in my few instances of troubleshooting and fixing certain programs, I actually liked how Mac OS apps prioritizes use of the Terminal to make under-the-hood changes. If one is just wandering through the gui alone, I could see how it can be frustrating to do much beyond basic tasks. The Terminal is really a necessity (from my very limited Mac experience) to delve much deeper.

[ScratchCat]

23 hours ago, Zodiark1593 said:

On my own phone (SD821), with a specially optimized browser, Kraken scores in the low 2k range. With Chrome, it is in the middle 3k range. Chrome can utilize 8 cores very well btw, so octa core chips probably have the edge. So optimization magic?

How does one produce a browser around 75% faster than that produced by Google, currently the leader in it's field?

If this were possible I would have assumed that Google (plus every web browser developer) would have produced an automated system to produce and distribute this sort of improvement. Furthermore according to this source the Pixel XL scored 2600 with Chrome 53 , therefore would be in the 2400 with improvements (The Pixel being the probably the most optimized device) , if this were beaten it certainly would be impressive.

 

21 hours ago, Kierax said:

Their is already a wealth of tablets running Atom chips with decent battery life, but full Windows 10, where does a ARM cutdown Win10 system fit in? 

ARM could provide further performance/efficiency benefits, better standby performance ( weeks instead of days) and integrated LTE. Additionally Intel seems to have given up on the Windows tablet market, leaving a vacuum  for ARM to fill.

[Tieox]

1 minute ago, ScratchCat said:

ARM could provide further performance/efficiency benefits, better standby performance ( weeks instead of days) and integrated LTE. Additionally Intel seems to have given up on the Windows tablet market, leaving a vacuum  for ARM to fill.

I see, could make for an interesting option then when people need another new device.

 

 

[Zodiark1593]

23 minutes ago, ScratchCat said:

How does one produce a browser around 75% faster than that produced by Google, currently the leader in it's field?

If this were possible I would have assumed that Google (plus every web browser developer) would have produced an automated system to produce and distribute this sort of improvement. Furthermore according to this source the Pixel XL scored 2600 with Chrome 53 , therefore would be in the 2400 with improvements (The Pixel being the probably the most optimized device) , if this were beaten it certainly would be impressive.

 

Using the YuBrowser, this was the best score I've achieved. I ditched the browser for other reasons, most notably, recieving no updates as of late. 

 

 

 

More recent Chrome updates actually seem to degrade the score in this benchmark some with subsequent releases, I've noticed. Something to note. I don't put that much faith into these scores though, as Chrome's smoothness and fluidity is still very good, despite benching far worse.