yes, but you’ll be missing out on ALL games, none of which are compiled for arm64.
also no, box64 is not hardware accelerated. both snapdragon and apple m series cpus have hardware magic that makes translation viable at all (as demonstrated by box64’s (and even fex) piss poor performance)
One of the biggest problems I had with windows on ARM was drivers. Most of my devices that needed drivers didn’t have an arm compatible version available. This needs to change more urgently than simply being able to run software, for me, at least.
They control the ecosystem in the way that they provide what hardware is new on MacOS and what capabilities it has. So if any developer wants to support modern devices they have to port to that new hardware. They don’t have any choice, if they want to stay relevant.
I wouldn’t necessarily present that as a good thing. If operating systems become incompatible with old software, that means that such software cannot be effectively preserved and may be lost to time eventually without a committed maintainer.
This is and always will be true about software. Progress sometimes means abandoning bad ideas, even ideas that were good at the time but are bad now that something else has changed.
Old Windows games generally don’t work on modern Windows without a virtual machine.
It has more to do with manufacturers than Microsoft. Nobody was making high performance ARM chips, so there was never a market for windows on ARM until now
Windows on arm was a thing, I had a surface 2 rt about a decade ago, too bad it never felt like microsoft ever really fully committed to the idea imo, and yeah x86 apps wouldn’t run on it (though there was an emulation tool apparently, was community developed). Market was definitely there (though I’m not sure how big it was, probably a cross over with netbook users), they just fumbled it like they did windows phone in my view.
Although it may look like Windows is a platform for any-hardware, reality is MS can and does push manufacturers to shape hardware as it’s desire, like forcing all mainboards to have TPM.
both apple m cpus and snapdragon have hardware magic to make translation viable when paired with usermode tools like Rozetta or Prism
otherwise it’s painfully slow, as demonstrated by both box64 and fex.
well apple cpus have a hardware flag which slightly changes side effects of some existing arm instructions AND completely changes the memory model to one that imitates Intel x86 for the current thread. (TSO)
with this flag enabled, the code can be almost fully recompiled ahead-of-time (from x86_64 to arm64) with minimal overhead (no need to remap memory addresses etc), with no reason to resort to e.g. jit recompilation on the fly (which is exactly what box64 does all the time)
i assume snapdragon and microsoft are doing a similar thing (it definitely involves some hw magic too, maybe not exactly the same as in apple) since it’s seemingly the only way to achieve parity with apple in terms of performance
What Apple did for Macs when switching architectures, though, was to port their own software to the new architecture. Microsoft doesn’t even port fucking Minesweeper to ARM.
Emulation is always slower and eats more battery. Microsoft’s laziness is proof they don’t care about that hardware, so may just as well buy an iPad Pro instead.
Its not laziness, they have lost developers support over time and lets be honest here, Windows 8 arm was roundly laughed out the door. Expectations are now marketing hallucinated by copilot.
This is typical Microsoft “agile”: minimum effort and delivery.
Yuzu can exhibit superior performance because the Switch is rocking the Tegra X1 from 2015. Yuzu absolutely cannot beat the Switch with contemporary hardware and/or comparable power consumption.
Oh yeah, clearly I did not read the article well. Still, it doesn’t mean what you think it does.
First, Yuzu is more of an alternative API implementation than an emulator in this setup. The stock Switch OS and API implementation have been entirely replaced with Linux and the Yuzu implementation of the API. Given recent performance uplifts in the Linux kernel, I’m not surprised that Linux+Yuzu beats the first-party implementation.
Second, the use of the word “emulation” in the above thread is really a misnomer: Rosetta 2, Prism and the like all perform what is called dynamic ISA translation. Yuzu need not perform ISA translation when running on ARM hardware.
It is always quite amusing to see a billion dollar corporation beaten in its own game :)
More information/context, if you’re curious:
Rosetta 2 in particular isn’t full emulation because the API is the same for both architectures - it is only dynamic ISA translation. I expect that Prism will be slightly closer to full emulation; there is simply no way Microsoft will reimplement all of the legacy Windows APIs on ARM.
WINE is a great example of something that is also not a full emulator, but for the opposite reason: it does not perform any ISA translation or hardware emulation, but rather only syscall (API) translation.
Emulation is almost always slower and eats more battery.
FTFY. There have been some cases where emulation actually outperforms native execution, though these might be, “the exceptions that prove the rule.” For example, in the early days of World of Warcraft, it actually ran better on WINE on Linux than natively on Windows.
To be fair this is also a translation layer and not an emulator.
Prism is an x86 emulator for ARM. If you think that Prism is “a translation layer and not an emulator”, I refer you to the very first word of the second to last paragraph of the submitted article.
This article seems to conflate “emulation” and “translation layer”. I don’t think there is anything that confirms “Prism emulates an x86 CPU”, only that it allows for running x86 code on ARM. This does not inherently require emulation as demonstrated by Rosetta 2, which is a translation layer.
only that it allows for running x86 code on ARM. This does not inherently require emulation as demonstrated by Rosetta 2, which is a translation layer.
WINE doesn’t “translate” one CPU architecture to another CPU architecture either, so the WINE comparison is still wrong, no mater if CPU translation is called emulation by you or not. WINE is a wrapper for API calls within the same CPU architecture. That’s it.
“Windows apps are mostly compiled for x86 and they won’t run on ARM with bare Wine”
What you linked is an effort to combine WINE with the QEMU x86 emulator which is an emulator because it emulates CPU calls. Hint that it’s an emulator is in the name “QEMU” and an actual quote from the wiki page you linked and clearly didn’t care to read: “Running Windows/x86 Applications: See Emulation”
EDIT: Let me also quote from the readme file of the Hangover project:
Hangover uses various emulators as DLLs (pick one that suits your needs, e.g. works for you) to only emulate the application you want to run instead of emulating a complete Wine installation.
The 68k to PPC transition was rough though. It wasn’t until system 8 that Mac OS on a PPC mac was fully PPC code. But that was also a much different Apple that’s nothing like the Apple of today.
I firmly maintain that if Microsoft gave a shit about ARM, they would be defaulting every one of their compilers to produce fat x86/aarch64 binaries. The reality is, however, that they don’t care about the hardware so long as it is good enough.
if Microsoft gave a shit about ARM, they would be defaulting every one of their compilers to produce fat x86/aarch64 binaries
Wasn’t the point of .NET once that native binary code isn’t needed? I’d say if Microsoft gave a shit about ARM, everything would have been ported to .NET.
Another thing they did is add hardware support for the x86 strong memory model to their ARM chips, allowing for efficient emulation. Without this, translated code takes a big performance hit.
Did Qualcomm add something similar to their ARM CPUs ?
this is for the transition. no point in porting your software if nobody has the hardware. This will get people to get the hardware, as they can just keep using the existing software, and wait until it’s properly ported
Edit: you people really think windows is the only software that needs a translation? Do you only ever use your OS on your computer, and not a single software more?
No, this won’t get people to get hardware that looks horribly slow because everything needs to run through a translation layer. They do have the sources. They could just recompile them for the new hardware. If their sources are not total crap.
I’d expect there’s quite a lot of assembly and endianness-dependent stuff here and there. It’s Microsoft. Their culture is about pride of things being arcane-complex inside, cause if you can untangle that, you are a good programmer. They think that. I think they think that. Maybe they are just vile.
Windows NT historically ran on lots of CPU architectures, PowerPC, MIPS, Alpha, Itanium, etc, and that included the bundled software like 3D Pinball. I would have expected it to still be quite portable.
You may be spot on with that. Though the assembly part can be fixed by code translation at compile time. Endianess and shitty programming habits are another thing that the cross-executor must deal with, too, so maybe this has been covered already. Or it will blow up in their faces, anyway.
Sure, but the vast majority of Mac software at the time, including loads of first applications from Apple, couldn’t run on Tiger. You had to run it in the “Classic” environment - and they never ported that to Intel.
Tiger shipped just 4 years after the MacOS 9.2 and plenty of people hadn’t switched to MacOS X yet.
The reality is Apple only brings things forward when they can do it easily.
Apple has done eight major CPU transitions in the last 40 years (mix of architecture and bit length changes) and a single team worked on every single transition. Also, Apple co-founded the ARM processor before they did the first transition. It’s safe to assume the team that did all those transitions was also well aware of and involved in ARM for as long as the architecture has existed.
Nobody will buy the hardware if they can’t commit to supporting the software. In a previous role, I was responsible for advising purchasing decisions for my company’s laptop fleet. The Surface X (Arm edition) looked cool, but we weren’t willing to take the risk, because at the time Microsoft had far worse transitional support than they do now. It’s gotten better, but no one in their right mind is going to make the kind of volume purchases that actually drive adoption until they demonstrate they are in it for the long haul. It’s a chicken and egg problem, and Microsoft doesn’t care what hardware you are using, so long as it is running Windows or using (expensive) Windows services.
I don’t really know if ARM adds benefits I’d really notice as an end user, but it’ll be interesting to see if this really goes through and upends the dominant architecture we’ve seen for really 40+ years.
The benefits, basically, are that it can provide an architecture that is designed for modern computing needs that can scale well into the future. That means high performance with low power consumption and heat.
The x86/64 model has been up against a wall for a while now, pumping out red-hot power hogs that don’t suit modern needs and don’t have much of a path forward wrt development compared to ARM.
If nothing else it breaks the stranglehold the 2.1 x86 licensees (Intel and AMD) have on the Windows market. Its just that that market is much MUCH smaller than it was 20 or 30 years ago.
Right? I’m much more excited to see RISC-V start to become more powerful and have more commercial offers of hardware to compete against the global tech brokers. We need the FOSS version of hardware or else our future privacy and ownership rights will forever be in jeopardy with info tech.
RISC-V is just an ISA, the same for ARM and other RISCs and CISCs. There’s no guarantee that RISC-V will be any freer than current CPUs, because the actual implementation and manufacturing are the job of the OEMs.
RISC-V is an open standard under an open and free license. Which means that it doesn’t require an expensive proprietary licensing fee. It is the necessary development bed upon which open source hardware can be created. Effectively it means that it has the potential of creating cheaper hardware that manufacturers can create with lower cost overhead and whatever improvements they make upon the designs can be used for free by other manufacturers.
The RISC-V ISA is free and open with a permissive license for use by anyone in all types of implementations. Designers are free to develop proprietary or open source implementations for commercial or other exploitations as they see fit. RISC-V International encourages all implementations that are compliant to the specifications. […] There is no fee to use the RISC-V ISA. FAQ
While all other ISAs are proprietary standards that charge chip designers up the nose to even look at the specifications. Hence why there’s so few chip manufacturers in the world.
ARM is the licensor, not the licensee. At the very least, they are willing to license the ARM architecture to more companies (the licensees) than Intel is with x86. More RISC-V support would be ideal though for sure...
Second this. Not to mention INSTANT resume from hibernation! It’s fucking crazy. I can use this thing ALL DAY doing webGL CAD work and Orca Slicer and barely scratch 50%.
With a modern system, I honestly don’t think there’s a noticeable difference between suspend to ram and suspend to disk. They’ve gotten the boot times down so much that it’s lightning-fast. My work laptop’s default is suspend to disk, and I don’t notice a difference except when it prompts for the bitlocker password.
S0 standby is borderline unusable on many PCs. On Apple silicon macs it’s damn near flawless.
My current laptop is probably the last machine to support S3 standby and I do not look forward to replacing it and being forced back into a laptop that overheats and crashes in my backpack in less than 15 minutes. On my basic T14 it works ok for the most part, but my full fat Thinkpad P1 with an i9 is in S0 standby for longer than a few minutes, and sometimes uses more power than when it was fully on. Maybe Meteor lake with it’s LP E cores will fix this but I doubt it.
There’s nothing stopping x86-64 processors from being power efficient. This article is pretty technical but does a really good explanation of why that’s the case: chipsandcheese.com/…/why-x86-doesnt-need-to-die/
It’s just that traditionally Intel and AMD earn most of their money from the server and enterprise sectors where high performance is more important than super low power usage. And even with that, AMD’s Z1 Extreme also gets within striking distance of the M3 at a similar power draw. It also helps that Apple is generally one node ahead.
If there’s ‘nothing stopping’ it then why has nobody done it? Apple moved from x86 to ARM. Mobile is all ARM. All the big cloud providers are doing their own ARM chips. Intel killed off much of the architectural competition with Itanic in the early 2000’s. Why stop?
Their primary money makers are what’s stopping them I reckon. Apple’s move to ARM is because they already had a ton of experience with building their own in house processors for their mobile devices and ARM licenses stock chip designs, making it easier for other companies to come up with their own custom chips whereas there really isn’t any equivalent for x86-64. There were some disagreements between Intel and AMD over patents on the x86 instruction set too.
This article fails to mention the single biggest differentiator between x86 and ARM: their memory models. Considering the sheer amount of everyday software that is going multithreaded, this is a huge issue, and the reason why ARM drastically outperforms x86 running software like modern web browsers.
Here is a great article on the topic. Basically, x86 spends a comparatively enormous amount of energy ensuring that its strong memory guarantees are not violated, even in cases where such violations would not affect program behavior. As it turns out, the majority of modern multithreaded programs only occasionally rely on these guarantees, and including special (expensive) instructions to provide these guarantees when necessary is still beneficial for performance/efficiency in the long run.
Thanks for the links, they’re really informative. That said, it doesn’t seem to be entirely certain that the extra work done by the x86 arch would incur a comparatively huge difference in energy consumption. Granted, that isn’t really the point of the article. I would love to hear from someone who’s more well versed in CPU design on the impact of it’s memory model. The paper is more interesting with regards to performance but I don’t find it very conclusive since it’s comparing ARM vs TSO on an ARM processor. It does link this paper which seems more relevant to our discussion but a shame that it’s paywalled.
On the x86 architecture, RAM is used by the CPU and the GPU has a huge penalty when accessing main RAM. It therefore has onboard graphics memory.
On ARM this is unified so GPU and CPU can both access the same memory, at the same penalty. This means a huge class of embarrassingly parallel problems can be solved quicker on this architecture.
Do x86 CPUs with iGPUs not already use unified memory? I’m not exactly sure what you mean but are you referring to the overhead of having to do data copying over from CPU to GPU memory on discrete graphics cards when performing GPU calculations?
Do you have any sources for this? Can’t seem to find anything specific describing the behaviour. It’s quite surprising to me since the Xbox and PS5 uses unified memory on x86-64 and would be strange if it is extremely slow for such a use case.
It’s been a while since I’ve coded on the Xbox, but at least in the 360, the memory wasn’t really unified as such. You had 10 MB of EDRAM that formed your render target and then there was specialised functions to copy the EDRAM output to DRAM. So it was still separated and while you could create buffers in main memory that you access in the shaders, at some penalty.
It’s not that unified memory can’t be created, but it’s not the architecture of a PC, where peripheral cards communicate over the PCI bus, with great penalties to touch RAM.
Well for the current generation consoles they’re both x86-64 CPUs with only a single set of GDDR6 memory shared across the CPU and GPU so I’m not sure if you have such a penalty anymore
It’s not that unified memory can’t be created, but it’s not the architecture of a PC, where peripheral cards communicate over the PCI bus, with great penalties to touch RAM.
Are there any tests showing the difference in memory access of x86-64 CPUs with iGPUs compared to ARM chips?
That’s actually not what I was referring to, although the unified memory architecture is certainly more power efficient for mixed-intensive workloads. The cost of transferring to/from dedicated GPU memory is (unsurprisingly) quite large.
I’m not expert, but I can tell you that Apple Silicon gave the new Macbooks insane battery life, and they run a lot cooler with less overheating. Intel really fucked up the processors in the 2015-2019 Macbooks, especially the higher-spec i7 and i9 variants. Those things overheat constantly. All Intel did was take existing architectures and raise the clock speeds. Apple really exposed Intel’s laziness by releasing processors that were just as performant in quick tasks, they REALLY kicked Intel’s ass in sustained workloads, not because they were faster on paper, but simply because they didn’t have to thermal throttle after 2 minutes of work. Hell, the Macbook Air doesn’t even have any active cooling!
I’m not saying these Snapdragon chips will do exactly the same thing for Windows PC’s, obviously we can’t say that for sure yet. But if they do, it will be fucking awesome for end users.
Add comment