5 ms·
They have a huge leap to make. The scale of difference between SQ3 and M2 is massive...but the bigger problem is Microsoft themselves. Windows ARM support and t
by agloe_dreams 3y ago
They have a huge leap to make. The scale of difference between SQ3 and M2 is massive...but the bigger problem is Microsoft themselves. Windows ARM support and translation is incredibly bad. Apple's magic in M1 was a fast and efficient CPU, yes...but also Apple using their ownership of the entire stack to make it:
A: Easy to port apps
B: Transparent to translate them.
Selling an M1 Mac was easy: It has some new tech, gets better battery life, all while running your old apps faster than your old Mac.
Windows on the other hand fails aggressively here. To sell a Windows on ARM device implies the need to explain a whole lot of mumbo jumbo to the customer about 32Bit and 64Bit apps, about stuff that won't emulate at all, about the massive performance hit.
I honestly think that a Snapdragon X Chromebook has much better odds. Google is in a far better space to have a singular, cohesive, understandable product. Chrome OS doesn't care at all what CPU is in there. It all "just works" for the average user.
- kees99 3y agoI don't get it. On Linux-arm, all I have to do to run x86 apps is this: apt install binfmt-support qemu-user update-binfmts --enable qemu-x86_64 That's it. Neither "ownership of the stack", nor "whole lot of mumbo jumbo" is required.
- user_7832 3y agoHow's the graphical performance in qemu? I tried using ZorinOS via Windows' hyper-v - which should run at native cpu speed - but the graphical performance was certainly not "daily driver-able".
- kees99 3y agoComment above is about enabling [qemu-user]. Hyperv is more akin to [qemu-system]. With latter, like with any full-system emulation, for top graphics performance, you probably want a separate video card and [pci pass-through] it to the guest VM. [qemu-user] https://www.qemu.org/docs/master/user/main.html https://www.qemu.org/docs/master/user/main.html [qemu-system] https://www.qemu.org/docs/master/system/index.html https://www.qemu.org/docs/master/system/index.html [pci pass-through] https://askubuntu.com/questions/1406888/ https://askubuntu.com/questions/1406888/
- bravetraveler 3y agoIt's worth noting for graphical performance with QEMU... guests can be pretty nice with 'virtio' and 'Spice' :) For simpler needs (ie: mostly 2D, very minimal 3D) it's great. I use this mostly for isolated browsing environments and am very pleased. I believe this works with both Windows and Linux guests, though drivers are more annoying where you'd expect.
- user_7832 3y agoThanks!
- agloe_dreams 3y agoI think you just clearly outlined exactly my point. 90% of all PC buyers would look at your qemu install and say "WHAT?!". Hell 80% probably don't know what an x86 app is...and they shouldn't! If a person buys a new PC, it should be better, at time of purchase, than their old one without gotchas. They shouldn't need to learn about why some apps are slower or why a particular app doesn't run. That is why Apple's move was good. They not only made it just work..but the M1 ran x86 apps faster than the outgoing model.
- jandrese 3y agoI have to assume that someone selling an ARM laptop with Linux would do that for you so the compatibility is transparent.
- bravetraveler 3y agoIt's a very small change to the installation medium, though. Your distribution already pre-selects hundreds of packages and performs configuration -- this can be another one. I'm a little surprised the maintainers haven't done so - but I haven't actually used the workflow. The implementation appears transparent; it registers a format with the kernel. Maybe there are downsides. You hint at a very likely one - performance. QEMU can do remarkably well with virtualization... but this, I'm not sure. I'm hopeful! Anyway, with all that in mind... I think the point against requiring such extreme 'vertical integration' remains If the argument against these two commands is it, then it's extremely trivial to resolve.
- leetharris 3y agoI haven't found Windows ARM to be that bad. It is not as good as Rosetta 2 for sure, but it handled most "normal" stuff fine and even some advanced .NET legacy work. I think with more effort they will get there. I think the bigger problem is the one you outlined: M2 is so far ahead. Hard to see Qualcomm catching up and being price competitive
- Shawnj2 3y agoThe real failure of Windows on ARM is that the performance of the hardware is shit Like getting an M1 laptop and virtualizing Windows 11 has better performance than most actual Windows 11 ARM laptops
- seec 3y agoBecause Apple upsell the shit out of those chips with outrageous pricing on other hardware parts. In the PC world that cannot work since nobody would buy a machine with such an uncompetitive price on configuration and that is if we pretend that Apple Silicon is actually competitive with the current x86 offerings. In some ways it is but in many other it is absolutely not. Qualcomm stuff won't sell well, not because it will be terrible but because unless it is priced extremely competitively nobody will care; because "battery life" is not a good enough argument that makes worth the hassle of dealing with a new architecture. There are so many cheap ways to extend the battery life of existing notebooks that it is a non-issue and complete Apple reality distortion field.
- huijzer 3y ago> Windows ARM support and translation is incredibly bad. Also poor support on GitHub. So poor that I’m suspecting they do it on purpose.
- Leherenn 3y agoAlso C: migrate or be left behind. With Windows you know the x64 version will be there "forever", so waiting to see how it turns out is a valid strategy; except it kills any momentum for the platform.