Hi
Will Fusion ever operate on macs running on Apple Silicon chips as it’s important that I can continue to run my x64 Linux vm images if I upgrade to a new Apple silicon based Mac
I couldn’t see any official comment on this from VMware
https://twitter.com/VMwareFusion/status/1326229094648832000
Admittedly, although "we're happy to say that we are committed to delivering VMware virtual machines on #AppleSilicon!" can mean VMware simply intends Fusion to work on Apple Silicon Macs, so that NEW virtual machines will be created for ARM-based guest operating systems, it MIGHT also mean the perhaps not-too-distant ARM-flavoured Fusion 12 will be able to deliver at least some PREEXISTING virtual machines (i.e., x86/x64 based) to the new Apple Silicon Macs. Naturally, that needn't mean it will emulate Intel Windows, but perhaps something similar to Rosetta 2 might be leveraged so that virtual machines of Intel 32/64-bit OS X/macOS operating systems will run on the new Macs.
That'd be my guess for the latter - apple enabling rosetta 2 for their hypervisor framework. Man if that happens it'd be sweet - I'd jump on a new 16 with an M2 chip next year at release.
Michael,
Ok it is official - Big Sur has been released and Apple Silicon is now in the wild. So please stop the cloak and dagger and tell us if we are going to be able to run x86 VMs on Apple Silicon and what the limitations are if any (e.g no virtualization inside the vm). The messaging on this topic has been murky at best. I will not be renewing my Fusion Pro license until I get some clarification.
Dan
As a public company, VMWare is restricted about how they can share pre-release information. I wouldn't expect any announcement until it's essentially in public beta.
Best guess is that we'll have ARM virtualization for Linux fairly quickly. Emulation of x86 is a whole different beast than virtualizing a new CPU though.
Michael Roy,
I understand that VMware is not ready to announce if it will support Intel based Windows running on Arm based Macs yet. But can you give us an idea of when this decision will be made and communicated? This information greatly impacts buying decisions for Mac hardware for many people who read this forum I suspect. Intel based Macs are now disappearing from the market. If this is the only solution to run Intel based Windows on the Mac in the near/medium term then it would be good to know that sooner rather than later.
Do you seriously expect an official response to your request? Let's imagine that they said something like: "An official announcement regarding Intel Windows emulation on Apple Silicon virtual machines will be made early in February 2021". Would you conclude that, after such a statement they would say "Nope, we won't be emulating Intel Windows"? It would make no sense. By giving such a deadline on their own work, they would be saying they are working on it because it's feasible. Ergo, if it is feasible, they are already working on it in one way or another. Therefore, the announcement will be made, if it is made, when the hypothetical product is ready, not now. In my view, the most likely scenario is this:
Well put. They can't make statements because it impacts revenue recognition accounting rules.
No existing mac magically got worse in the last two weeks. Folks who might consider low-end macs do have a decision to make, but nothing changed about their current situation as a result of the new releases. Most VM users probably want the higher end models yet to be released anyway (16GB is a real limit). We have years, literally, before there's a forcing function.
I think we'll see your first two bullets in the next few weeks or very few months. The third bullet is really up to Microsoft, as they've not announced any sort of emulation capability for Windows ARM, so not holding my breath. The last one is the one I'm hopeful for, since that emulation already exists - my only question there is will it be big sur, or the following OS that enables it.
Of course it is a reasonable response. So now we are down to what would be a reasonable expectation.
The VMware team has ESXi running on ARM and had been working on that for a long time to get so far. This means that critical monitor code - at least for type 1 hypervisors- is code complete.
In theory, the Fusion Type 2 hypervisor (depends on Mac OS to interface with much of the hardware) is mostly a recompile, but clearly VMware is working hard on this.
Now the most important question: Given that Apple has full support for running Intel instructions via Rosetta 2 (which most likely CANT be reused by VMware) AND Geekbench numbers being posted on M1 machines are proving that the M1 chip is emulating x86 instructions FASTER than Apple's own most powerful Macbooks... Will VMware support x86 VMs? Which in theory unlocks the CPU and GPU performance of the M1 for gamers (now that there is no bootcamp)?
Well. Methinks they will at least try.
You surely know much more about this than I do, kjdfhaueiase, but I'm not sure we can truly say that "the M1 chip is [capable of] emulating x86 instructions FASTER than Apple's own most powerful Macbooks". I know that, after one given Intel macOS application has been initially run on Apple Silicon Macs, it runs very well indeed, but that's because the first few seconds (minutes?) of that initial run were devoted to translating the Intel code into ARM code. So, if I'm not wrong, that means that all Intel-only applications run on Apple Silicon effectively become Universal applications after they are run for the first time, which means there's no code translation after that first run; so, in effect, there's no emulation as such. Perhaps the resulting ARM code isn't truly optimized (as it would be if the source code had actually been compiled for the ARM architecture) for the M1 chip, but it's code the new chip can understand natively, without the translation overhead.
The problem, as I see it, is that it is unlikely a virtualization platform, such as Fusion, will benefit from that initial translation of code. I doubt all the intricacies of a full-fledged Intel operating system (like macOS Mojave, or Windows 10) can be successfully translated into Apple Silicon code by simply running said operating system one time. Not only that, how on earth would Fusion manage to ARM-translate the native code of an application run within said operating systems? Take Microsoft Word, for instance. How would the forthcoming Universal Fusion manage to translate the x64 Word into an ARM-palatable version for an equally ARM-translated Windows 10? I don't think it can be done, much less done on the fly every single time you want to run that version of Word within said virtual machine, because it would be difficult to keep a ready made translation for future, nearly native, use. The situation for an Intel macOS Word running in a virtual Mojave or Catalina might be slightly better, but would probably require that the Universal version of Fusion translate each macOS application the first time it is run and for it to save such a translated version within the application itself, making it "universal".
EMR. Rosetta 2 is not a translation engine but an emulation engine. Much closer to what kjdfhaueiase described than what you said. Those first few seconds for the first launch of the first x86/64 application is for the Rosetta 2 engine to load. Once it is loaded for the first application (per benchmark sites that know a lot more than me) it stays running as long as the machine has a need and the memory to allow such. And, yes the benchmarks show that the M1 Macs are MUCH faster than the current fastest Intel Macbook PROs and that even emulating an Intel Mac they are still faster than the current fastest Intel Macbook PROs (though much slower than the pure M1 Mac running code compiled for it).
Regular benchmark:
https://www.macrumors.com/2020/11/11/m1-macbook-air-first-benchmark/
Emulated benchmark included:
So, for the normal standard Macbook PRO user who needs access to one or more windows only applications this proves that this can be done (at least once the M1 chip is available in even fast configurations with much more RAM). But, that is not to say that either VMWare or Parallels will want to spend the time and money it takes to write their own emulator and Apple has already stated that they will not allow (unless they change their minds) VM engines to use the Rosetta 2 engine.
Thanks for clarifying that, rokair. I'm probably mistaken, but I find it surprising that the emulation is carried out every single time an Intel-only app is run. Wouldn't "my" explanation make more sense? Be that as it may, if it's false, so it is. Thanks again.
I don't think it's so much apple won't allow, as apple hasn't enabled the big sur hypervisor framework to use rosetta. That's a non-trivial challenge, but maybe doable. Craig did say in an interview that Apple's aware of the desire/need/issue, and they sell a ton of machines into corporate accounts, so hopefully we'll get some good news.
VMWare isn't in the emulator business, and creating one would take a tremendous amount of work and investment. That's a complex business case, and one that I think would be tough to build, especially since the primary x86 use case will be windows, and microsoft has already announced that windows for arm will have x86 emulation out in november. Now we're down into a very small set of users, and it's probably not cost justified.
In the end, this is one of those things we're just going to have to wait on. VMWare will make the announcements when their accounting and legal folks let them.
If you were only talking about interpreted code or a single known intermediate format then maybe. But, for general x86/64 applications written by thousands or even millions of different developers in hundreds or even thousands of different languages each with their own compiler.... Hmmm. No. What Rosetta 2 is doing is a machine emulation for the base Intel and AMD x86/64 Processor Instructions. It is at a MUCH lower level than any code conversion.
I will give you my frank assessment.
Will we see a Fusion on M1? Yes... for Arm VMs, that is.
Is a Fusion for M1 running x86 VMs possible, a la Rosetta 2, but for system/kernel/hypervisor-level x86 code, via full x86 emulation? Yes.
Will VMware deliver such a Fusion product? Most likely not.
As numerous and vocal as the Fusion user community may be, it simply doesn't strike me as financially sufficient for VMware to warrant the effort.
If lots of Arm-based Windows computers ship, then... well... maybe... but still... I think it would be a remote shot.
So, long story short, I think you are looking at a fork in the road, and it seems wise to me to be prepared -- what matters more to you... the ability to run your VMs... or a specific host OS and/or specific host hardware?
Unix-like host OSs are a dime a dozen, and they all run a modern browser and a modern shell. By contrast, VMs can live a very long time... mine predate the past decade of MacOS, actually. 🙂
But yeah... long term... it is time to think about x86 as a legacy system, i.e. figure out how to run software (VMs) for it, once the hardware is long gone. Not a new concept in history, after all...
Cheers!
> it MIGHT also mean the perhaps not-too-distant ARM-flavoured Fusion 12 will be able to deliver at least some PREEXISTING
What do you mean ARM-flavoured Fusion 12? They're going to call it Fusion 13 and make you pay for the upgrade.
I don't think you are right. I would think that a Universal version of VMware Fusion (i.e., one capable of running natively on Apple Silicon) should drop soon. Naturally, that doesn't mean it will be able to run Intel Windows, but it would be able to run ARM Windows, if Microsoft allows that.
Nah, they've used excuses like this in the past to fleece customers out of more money. New version of OS X drops? Fusion needs to be a new version too when it could have been a simple patch!
No need to get out the pitchforks and torches.
I looked at my past 6 years of Fusion invoices.
$49.99 a year, like clockwork, every Sep/Oct.
With 11->12 being an exception: a 2 year gap
instead of just 1 year, and $79.99 not $49.99.
$50 a year is totally reasonable in my books.
If you disagree, then fine, just stick with your
existing HW+SW, and don't ever upgrade. 🙂