silicon macbook news 2021 news stories

It’s been a few months since our informal announcement via Twitter back in November where we committed to delivering VMware VMs on Apple silicon devices, so we wanted to take this opportunity to share a bit about how our progress with our little project to bring Fusion to life on Apple silicon Macs this year.

The quick read

Before we get right into it, I just want to summarize our position way up front with a quick tl;dr:

A new generation of Macs

With the introduction of Apple silicon, it was revealed that the new CPU line would be based on the same Arm CPU architecture found in an iPhone or on an iPad as opposed to the x86 or x86_64 Intel (or AMD) architectures found on desktops and notebooks. With the new architecture comes incredible performance gains, thermal improvements, and dramatically improved battery life, but poses some unique challenges for virtualization apps like Fusion Pro and Player.

With first generation of Apple silicon chips, namely the M1, Apple has made significant performance and efficacy improvements, with claims of “Up to 2.8x CPU performance; Up to 5x the graphics speed; Up to 11x faster machine learning; And up to 20 hours of battery life” on a new 13” MacBook Pro.

Seeing improvements like that, it comes as no surprise to us that when users got their hands on M1 devices they naturally wanted to run virtual machines on them! Why not take advantage of that extra CPU power and carry around a single notebook instead of 2 laptops, right? We agree.

In much the same way they did when moving from PowerPC to Intel CPUs back in 2006, Apple introduced a new version of Rosetta to support running Intel apps on Apple silicon. For the most part, apps ‘just work’, even if they’re a bit slower.

However, for those that need to run another operating system like Linux or Windows, Rosetta 2 doesn’t support Virtualization, and Apple silicon Macs don’t support Boot Camp. That means it’s time for us to innovate and rebuild our beloved desktop hypervisor for Macs, VMware Fusion, to support the next generation of Apple hardware.

Fusion’s roots

With the 2006 transition, a tiny (but incredible!) team of engineers at VMware saw an opportunity. As a side project, this small group were able to essentially rebuild Workstation to run on the Mac using Apple’s UI, thus creating the foundation of what we now know as VMware Fusion

One of the benefits our users appreciate of having older “enterprise-grade” siblings with Workstation on the desktop and ESXi in the data center is that it gives organizations a consistent operating model. A VMware VM behaves pretty much the same regardless of what product it’s running on. Developers and Operations teams can move VMs and templates between data centers, desktops, and clouds with ease. This is super important to us and to our customers, particularly as more and more operational workflows become automated.

Some context

Now, we’re no stranger to Arm CPUs, having shipped what is currently a something we call a Fling with ESXi Arm Edition. Delivering ESXi for Arm has been a multi-year effort, and yet it’s still not quite a Product like ESXi on x86 currently is.

So when we learned about the M1 devices, we knew we had the in-house expertise on both the Arm team, and also on the Fusion bench, to set in motion a plan to re-invent our Mac desktop hypervisor to support this incredible new platform. Being able to build on top of what we’ve learned with our still-evolving Fling has been crucial, and thankfully we have some overlap in the teams’ history, meaning folks have exactly the right experience needed for this project. 

To support Fusion on M1 devices, while maintaining code and feature compatibility with our ecosystem, we are essentially bringing the core of these two projects together.  This is a much different task that simply shipping a single product like Fusion to say the least!

So, how’s it going?

Well, our initial assessments are going very well! For starters, we have VMs booting in a variety of Arm operating systems, and we are very impressed with the performance! 

Because of our kinship with ESXi, we have a major architectural advantage over our competition. ESXi is designed to be enterprise-grade, which includes security, resiliency and performance benefits that both Fusion and Workstation get to benefit from.

Here’s a couple of screenshots from my test desktop, a M1 MacBook Air with 8 CPU + 8GPU cores and 16GB of RAM:

You can see here that I have 7 ARM VMs booted at once… 2 are CLI only (Photon and BSD), the others are full desktops… each is configured with 4CPU and 8GB of RAM. 6 different Linux flavors and 1 FreeBSD… MacBook Air. On battery. No fans. Yep. 

Of course, just booting a bunch of VMs that are mostly idle isn’t quite a ‘real world experience’, nor is it the same as doing some of the stress testing that we perform in the leadup to a release.  Even with that said, and note that I’m using ‘debug’ builds which perform slower, So we’re very encouraged by our early results, and seriously can’t wait to get it on every Apple silicon equipped Mac out there.

Sounds good, so what’s the hold up?

While booting all that at once and it being usable (which it all has been in my testing) is an impressive feat in itself, we do still have a ways to go, and some challenges along the way.

For instance, the best Linux VM experience comes by installing VMware Tools, and by and large Tools are included with every Linux distribution. Currently, open-vm-tools are not readily available on the aarch64 (Arm) platform.

The ESXi-Arm project, in addressing this gap, currently has users build open-vm-tools from source itself. That works just fine for some people, but obviously not everyone is comfortable doing that.

Because open-vm-tools is such a key building block to support the experience we want to deliver for Linux VMs on every platform we support, we’re working with various Linux upstream projects to include the necessary kernel patches to support open-vm-tools and open-vm-tools-desktop on arm64/aarch64 architectures so they can be included in OS distributions. These changes will also benefit the ESXi-Arm Fling by not having to compile Tools from source going forward, so things should ‘just work’ out of the box, as users have come to expect.

So for now, while VMs are booting, we don’t currently have things like 3D hardware accelerated graphics, and other features that require Tools which Fusion users on Intel Macs have come to expect.

That said, even without hardware 3D and while using debug-enabled-builds, we are super impressed with how well things are performing, even against the GA release of our competition.

What about Windows?

Of course, users are expecting to run Windows in a virtual machine, much like we’ve been used to for many years now. With Windows on ARM however, this presents a unique situation, particularly as it relates to Licensing. 

The Insider Preview program says: “To install Windows 10 Insider Preview Builds, you must be running a licensed version of Windows 10 on your device.” And as far as we are aware, there is no way to buy a Windows 10 ARM license for a Mac with Apple silicon. There have been plenty of discussions on the topic from users and the media, and from the Insider Download Page, it reads:

With Windows 10 on ARM Insider Preview builds, you can create 64-bit ARM (ARM64) VMs in Hyper-V on Windows 10 ARM-based PCs. Creating ARM64 VMs is not supported on x64 hardware.

ARM64 VMs are only supported on devices that meet the pre-requisites:

You can see it doesn’t say anything about Apple silicon. We have reached out to Microsoft for comment and clarification on the matter. 

For the time being, our work has been focused on Linux guest operating systems, and we’re confident that if Microsoft offers Windows on Arm licenses more broadly, we’ll be ready to officially support it.

What about x86 emulation?

We get asked regularly about running x86 VMs on M1 Macs. It makes total sense… If Apple can emulate x86 with Rosetta 2, surely VMware can do something too, right?

Well, the short answer is that there isn’t exactly much business value relative to the engineering effort that is required, at least for the time being. For now, we’re laser focused on making Arm Linux VMs on Apple silicon a delight to use.

So, to be a bit blunt, running x86 operating systems on Apple silicon is not something we are planning to deliver with this project. Installing Windows or Linux from an x86 ISO, for example, will not work.

More personally speaking, I really don’t think the next era of Macs will be defined by “switchers” in the same way the previous one was. I expect this platform will be one to more rapidly introduce new experiences at the expense of cutting away from the past. Where we’re headed is anyone’s guess, but I am confident the direction we’re moving isn’t backwards.

That being said, we’re always looking to broaden our horizon with respect to use cases, so if there’s something we’re missing, give us (or me) a shout, we’re eager to hear about it.

When can we get it?

Ah yes, the burning question: When can everyone get their hands on a tech preview?

We’re working diligently to get VMware VMs on Apple silicon and making great progress as you can see, but we would be remiss if our product did something unexpected or unsafe to any computer it was installed on. Even for a Tech Preview there’s a good deal of QA/QE work still to be done as we continue to add code to bring features online.

We’re the leaders of virtualization in the enterprise with SDDC stacks like VCF because we consistently deliver a high measure of quality, security and performance across all our products. It’s not because we shipped first, it’s because we ship when it’s ready.

That said, the team is planning to deliver a Public Tech Preview of VMware Fusion for macOS on Apple silicon before the end of this year, and we can’t wait to get it in the hands of every Apple silicon Mac owner.

Wrapping Up

If you’ve made it this far, the team and I appreciate it! The world is a bit crazy right now, but there are great things coming. Thanks for your understanding and patience as we take the time to get it right amidst all of the challenges of today.

We really couldn’t be more excited about not only the development progress, but for life with Fusion on Apple silicon. The hardware is incredible, and if our early testing is any indication, users are going to be very happy running VMware virtual machines on the latest Macs of today the future.

Keep an eye out in the VMTN community or follow us on Twitter for our Tech Preview announcement in the coming months!

And please consider donating to the charities above to support our Indian friends and colleagues affected by the global pandemic. <3

May 3
– this blog was updated to reflect that the timeframe to deliver this offering is not currently impacted by the COVID situation. That said, for folks looking to help with the relief efforts in India, please check out the 2021 India COVID-19 Recovery Fund.


93 comments have been added so far

  1. I really need that x86 emulation and Windows, so there’s just no way I can buy VMWare fusion for my M1 MAC. It probably also means I wont be purchasing any more M1 Macs until I get that emulation.

    1. Thank you, Michael, for the update, although I have to say it’s disappointing. I have to use some highly specialized x86 based software for the industrial controls industry (think Rockwell, Siemens PLCs, etc). If I can’t run VMs on my M1 MacBook Pro, then I will probably go back to using a Lenovo Windows laptop and forego the M1 altogether, which would be a shame. I’ve been a loyal VMware customer for over 10 years, but this is disappointing and I’m going to try the Parallels release that was just announced.

      1. Parallels gives you Windows ON ARM which is very VERY different from Windows

        If you are dealing with software for industrial application please evaluate if driver and other stuff can run on Windows ON ARM…

        1. Actually Microsoft’s Surface X tablet is ARM. You can download arm versions of Windows to test. They even run on Raspberry Pi 4.

    2. Agreed with Bob above. I have a bunch of obsolete software running under Windows XP (!!), where the applications never were ported to Windows 8, let alone anything else. I have a windows XP environment which I have been running for two decades, and have faithfully been buying new Fusion licenses every few years to keep it running. I recently upgraded to Fusion Pro so I could export the VM and import it into a VMware Player on Linux.

      VMware Player on Linux kind of worked. But it’s so hideously unreliable (I’ve had to re-install Player three times, and Linux once, just to keep it working), that it’s not a viable long-term strategy. There is certainly a market for emulation of ancient operating systems on ancient hardware, to keep equally ancient applications working. It’s technically possible, from the other comments I’ve seen, it certainly looks like it should be financially viable.

    3. I am currently waiting to change my Macs as a apple and Windows developer I use x86 emulation to test on the Windows platform how it can be not viable I can’t see as there are many thousands like me. I switched back to VMWare from Parallels because of the improved cross compatibility with other VMware products on Windows but if they are not going to continue there is not much point in continuing with VMWare as I like many other don’t use Linux. I don’t see the point of Linux for Mac users the OS is already Unix based so most apps can be ported anyway.

  2. > “we’re always looking to broaden our horizon with respect to use cases”

        I agree with Bob Comer about x86 emulation and Windows, and disagree with “there isn’t exactly much business value relative to the engineering effort”. There are decades worth of legacy Windows apps, which is what made Intel Macs so exciting—a truce (now expired) in the Mac vs. PC wars. My use case is Quicken: I’ve looked into MacQuicken since its first OSX beta, and every few years since then; they’ve never seemed serious about data migration integrity.

    1. Would they not run with the x86/x64 emulation provided by Windows 10 for ARM?

      1. Thanks, I got a CodeWeavers trial license in 2008, but real Quicken performance, even on an Intel Mac, was so slow that it never actually converted my data, which was quite worrying. Data conversion seems a lot easier to simulate on the same processor architecture than, for instance, emulating UI code on ARM. (BTW, Hi back at you, Jim, and thanks for the UTM pointer. Likewise to Mazen for Parallels, which I used before I switched to VMware—from what I heard they were a lot more serious about quality.)

  3. I was hoping that VMware being the market leader in virtualization and cloud tech would have been the first to the table with x86/64 emulation for Apple Silicon. Whomever at VMware come to the conclusion “..there isn’t exactly much business value..” for x86/64 emulation should reevaluate that there won’t be not business value for Fusion to exist if it can’t support x86/64 emulation. UTM is only $10 for ARM only VMs, but I would be happy to pay more for x86/64 emulation in addition to ARM on Apple Silicon.

    1. UTM emulation of x86-64 works on an M1 Mac. It is reasonably usable if you turn on the force multi-threading option. It isn’t clear if that is safe though.

      QEmu 6.0.0 just came out today so I’m looking forward to seeing how well that works. If you aren’t aware, UTM uses QEmu internally for both its hypervisor support and for emulation.

  4. One business use case for Dev and Test has been legacy OSs in the past. VMWare Fusion has been on par with VMWare Workstation and ESXi in the past by allowing me to run OSs from DOS6.2 up to Mac OS Catalina. Why does MS licensing even have to enter the discussion when producing a VM. License usage is in the purview of the user not the VM Hypervisor manufacturer. Just produce x86 emulation and you will give people access to a huge library of legacy OSs and software that only runs on legacy OSs. You also skirt the Windows for ARM licensing problems you are simply creating by not having x86 emulation. The business case? LOSS OF MARKET SHARE. If you can’t produce a hypervisor capable of running ALL the popular architectures regardless of the underlying platform, then you become much less useful. It’s funny that you are touting the seamless ability for VMs to be experienced the same on all platforms but at the same time saying that if I move a VM from an x86 hypervisor to an ARM hypervisor that it just won’t be supported. This eliminates one of VMWare’s core concepts of Hardware abstraction that has been a PLATFORM for arguing VMs since the beginning. Again focus on every hypervisor supporting every architecture and VIOLA the licensing become a problem between the user and the OS manufacturer.

  5. I agree with previous comments about x86 support. For me the value of VMs is precisely to be able to look backward and use previous operating systems and programs. Even a very slow x86 emulation would be a good start and very valuable.

  6. One example: I am working with an Mac for about 10 years. Most of my business depends on VMs running on ESXi, which i test on my Mac and my ESXi servers. The performance of the M1 Macs is great. I need to replace my Mac in the near future and I always try new technologies to see how it can help in business. Should i switch back to a x86 Linux Notebook? Perhaps …, or get one of the last Intel Macs…

    Another example: Students gets M1 Macs In five to ten years they decide which technology should be used in the future. If they cannot run an x86 VM with VMware Fusion soon, they possibly will not think about to run anything on ESXi in the future.

  7. I think that we need the emulation for x86. In my case, I use home labs based with VMware, and I can’t find a Vcenter for Arm.?? On the other hand, I use vms for openstack in x86, how to leave all your old VMware environments?

  8. Parallels currently loving it. The performance on their VM on M1 macs is so great. they released the the tech preview months back. comone vmware. you’re getting too late to the party.

    1. Wow, I hadn’t heard of this, I am going to try it out. I became a macOS convert because of VMware and Win10 support. I was pumped to get the new M1 chip too, I bet you can guess how that ended up. The reason I use it is to run excel addons and other weird software that is Win exclusive (some only work on x86 of course) in my workflows so this may be the next step for me. I really appreciate you sharing.

      1. I tried both, VMWare & Paralells, today on my Mac M1. Both currently support only ARM OS Systems, hoping to run some VM’s from my older Mac (Intel 2012) I’m in the same situation, customers having x86 Industrial Software, even running on Win XP or Win7. Can’t test. Industrial & Custom developed Software is always more expensive than consumer Apps. What happens when Intel also decides to go ARM?

  9. I get the sense VMWare isn’t serious about the Mac market. I’ve been a customer for 12 years. But when I upgraded to macOS 11.3 a Windows VM suddenly started crashing. Rather than pay them $50 for support (that is apparently not included with my license) and hope that a future Fusion release resolves the problem, I downloaded the Parallels trial. So far it’s working great, so I’ll probably just stick with it.

  10. Will the final product still have a “Pro” incarnation that can connect up to ESXi & vCenter servers? That is a mission-critical function for me, so I really hope it doesn’t go away!

  11. I’m using a few WMWare Windows systems, w10. w7 and even server … I can’t rebuild those with a reasonable effort.

  12. Thank you, Michael, for the update, although I have to say it’s disappointing. I have to use some highly specialized x86 based software for the industrial controls industry (think Rockwell, Siemens PLCs, etc). If I can’t run VMs on my M1 MacBook Pro, then I will probably go back to using a Lenovo Windows laptop and forego the M1 altogether, which would be a shame. I’ve been a loyal VMware customer for over 10 years, but this is disappointing and I’m going to try the Parallels release that was just announced.

  13. I recently purchased an M1 Macbook Air, and rely heavily on Sparx Enterprise Architect for my daily work, Windows only. I have tried CrossOver Mac from Codeweavers earlier on my 2017 15″ i7 MacBook, and it left a lot to be desired. Tried it again on my M1 Air and it runs so much better here, thar I don’t think I will need a WM in the future.

    I know that CrossOver is a bit of a hit and miss, but I have no issues with Enterprise Architect.

  14. This is *really* disappointing and I hope VMWare will change their mind:

    –> We don’t plan to support installing or running x86 VMs on Macs with Apple silicon.

    My entire research infrastructure is currently intertwined with VMWare products and this is a deal-breaker that means I will need to look for an alternative and dump VMware Fusion in the near term.

  15. Blaming VMware Fusion for not being able to handle a Windows problem. You jam an entire new processor design into a Mac, Windows itself is on the strugglebus for ARM based CPU’s and this is on VMware? Trust me VMware will support it eventually once MS gets their act together. Besides all I hear is people complaining about Windows on a Mac. Hey why not just buy a Windows computer then? People you know who buy Mac’s use, you know. OS X. VMware has always been good about adapting to new technology and hardware. No one does it better. Someone might have a leg up right now, but in the long term VMware always comes out on top. Fusion for Mac isn’t exactly a large portion of VMware’s revenue so it makes logical sense for them to prioritize projects. If you are so up in arms (pun) about it buy a 2020 MBP with an Intel processor. Problem solved.

    1. Our blaming VMware is because they clearly state their disinterest in ever supporting Windows on M1. And how ridiculous is it to think buying another computer just to run an OS is a valid solution…IN THE FREAKING COMMENT SECTION OF A virtualization company blog.

  16. Please release an early VM Fusion / ESXi that can run at least mac VMs on M1 for automation testing.

  17. I will be convincing my company to move away from VMware and purchasing Parallels for it’s clients as we have a real need and it’s immediate because APPLE is selling the M1 and we are buying them…

  18. It’s clear VMware doesn’t care about M1 Macs. They’re the ones that know how many people are actually *paying* for VMware Fusion. If this was an important business decision, they would have made it much earlier. They’re capable of emulating or virtualizing pretty much anything they want – it’s their core business for decades.

    *Apple* decided to make their own custom silicon. What does VMware owe you to build an entire product line to make your workflow perfectly transition over? Go tell Tim Cook that you won’t buy any more Macs because you only really used them to run your non-Apple software in a pretty little package. I’m sure it will really cut into their bottom line.

    If you need x86 that bad, BUY A PC. All these threats of “well, maybe I’ll just switch over to something else” is exactly what you should do. If “performance doesn’t really even matter, as long as I can run DOS” is actually true, you should have no trouble finding x86 hardware that can run all your precious virtual machines.

    1. If you need an x86 that bad, BUY A PC?!?” – Sorry, most wintel HW builds are crap, with plastic pieces falling off after 6 months of usage. The SW isn’t much better, with bloatware added OEM as a “feature”. Don’t even get me started with the Lenovo BIOS that phones home to China, and “self-heals” if you make any changes (read: you try to remove their backdoors).

      Oh and the service. Dell used to be good, now they are crap. Thinkpad support has sucked since IBM stepped out of the picture…4 redirects for service, including one to an analog answering machine on my last go around 4 years ago. Never again.

      1. BS. Everyone has his own reasons to judge the inconvienience they have with the lack of x86 support. So don’t judge about other peoples issues with this problem.

  19. I understand you work on Fusion on M1, But your Parallels is already release their product on M1. Somewhere in 2021 is very large. Personally I will wait a couple of weeks and switch to Parallels if VMWare is not announce a more precise date. It would not be am easy decision because I support VMWare for a log time.

  20. I have exactly zero use cases for which a virtualization product that doesn’t support x86/x64 would work for me, since the only thing I need a VM for is for using hardware that has (Intel-based) Windows-only drivers. ARM Windows can’t load those, and they obviously won’t run in WINE, so I don’t really have a lot of options other than getting an Intel-based Windows computer. I’ve kept upgrading my Pro license for a while, but I don’t think it really makes sense to continue doing that if I won’t have any use for it after upgrading from my existing Mac.

  21. I can’t believe Apple isn’t more interested in helping you push this to completion faster. As someone who develops and tests enterprise software for macOS, running macOS VMs is a part of our daily life. If I can’t do that with M1 I’m not supporting M1. It’s as simple as that.

  22. Super disappointed that VMware is dropping the ball so far with x86 VMs on M1, which unfortunately is a deal-breaker for me for future purchases of Fusion and going all-in on M1s. Among other things, software protection mechanisms such as dongles and machine IDs don’t work in Wine and CrossOver, but work fine in VMs. I’ve preferred VMware solutions for decades, but if it can’t/won’t get it’s act together to port x86 VMs to M1, this will be a huge blow to many users, who will be pushed to investigate other options that will hopefully emerge. Pleaase reconsider this sorrowful abandonment of your Fusion user base.

    1. Totally agree. No x86 VMs, no future purchases for Fusion for myself and most of my professional colleagues. It’s not just Win support, but other vintage and legacy OS’s that our customers use and require support for.

      I choose VMware because I though they added value over the other VM offerings. Very disappointed.

  23. So how do I get a refund? I have been running a mac in a windows environment for over 10 years happily using Fusion. Mac died. Upgraded to M1 because Fusion was supported on Big Sur and M1. Restored from Time Machine. VM no boot. Ok cool. I get it. Wait it out for Fusion beta. Now no support for x86 and a year for possibly windows arm? I have no use for this product that I was only able to use for 3 months.

  24. Very disappointing. For my needs (test lab for Windows 10 SAC/LTSC and Windows Server) your product becomes useless on Apple silicon.

    > x86 emulation (…) there isn’t exactly much business value relative to the engineering effort that is required So there is much business value in supporting only Arm Linux VMs? Or the engineering effort is very low? But then, why the delay? It makes no sense to me…

  25. I’m going to have to side with VMWare here. As of late, Windows 10 performance when virtualized just doesn’t cut it for my use cases. I don’t use Visual Studio much anymore (Rider rules!), but when i did, I’d have several medium size projects open. That alone would kill a VM with 16GB of RAM allocated, even running on my iMac Pro. I realized i just have to pony up the money for a real Windows desktop. Since I got back into Windows based gaming, it was easy for me.

    I suspect many commenting here make pretty darn good money. I think complaining about something you can fix by buying or building a new Intel or Ryzen based PC is simply a waste of time. My Ryzen gaming PC is comparable to my iMac Pro with respect to most of what I can throw at it. There are zero interest promotions all the time. If you make big money, you can get a Threadripper :-).

  26. Interesting discussion. I’m currently running Windows 10 for ARM on an M1 Mac Mini 16/256 using Parallels. The W10 ARM VM has only 2 processors and 6GB of RAM allocated. Performance is blazing fast—significantly better than my late 2017 HP i7 16GB laptop.

    Of course I’m not using this for production and don’t have any x86 legacy dependencies to worry about. W10 ARM 64 is still somewhat buggy and not ready for prime time. But it seems to be getting close. Obviously I like being on the cutting edge and would love to test the tech preview of VMware Fusion. But I’ll probably have purchased Parallels long before Fusion M1 preview is available.

    It’s looking like the future is ARM’s for the taking with Apple silicon leading the way. Intel x86 is getting left in the dust. Strategic planning based on the future being much like the past has left both Microsoft and VMware behind the curve right now. Better catch up quickly!

  27. Simply said, VMWARE Fusion on a M1 Mac without x86 support makes no sense and is useless for me (and I thing for many others too). I’m sure someone will see the potential of the x86 support for all the x86 based VM’s that are in the world. I’ve used Fusion for over 12 years now, and I would like to use it for the next 12 years, but only with x86 support. So the first one that can offer x86 emulation for M1 Macs will win. I really think you should reconsider your decisions about that topic.

  28. Personally, I would pay 2x the price for VMWare Fusion to run x86 emulation on an M1 machine. I run engineering CAD programs that only run Windows, and for these use VMWare Fusion. It might be possible to run them in emulation mode on Windows for ARM, but there is a lot of unknown going on there.

  29. Indeed, VMWare Fusion running x86 emulation on ARM (particularly Apple silicon) would be fantastic. I couldn’t agree more—especially since Windows 10 ARM is still bleeding edge and even Microsoft is struggling with x86 emulation performance issues on ARM. Intel x86’s days appear to be numbered. As hard as it is to believe after decades of unrivaled desktop supremacy, it’s increasingly looking like those heavily dependent on legacy x86 apps (and there are many) are going to end up stranded on 14nm Intel x86 island while the rest of the world is lifting off with star fleet ARM. No doubt x86 island will be around for decades to come, but can it ever overcome the gravity of legacy compatibility to join the new age of processing architecture?

  30. I’m not sure VMware really understands its user base. For example, we use Fusion to prepare VMs which we then push to vSphere (ESXi). Considering that pretty much all of the current and upcoming server hardware is still x86/x64, I’m not sure what the value proposition is for running Linux for ARM under Fusion on Apple Silicon.

    Because the reality is that just because Apple is moving to ARM doesn’t mean the rest of the industry will. Despite being hyped for nearly 8 years as the next big thing in the server space, ARM has been pretty much going nowhere here (aside from a handful of token products, many of which have been EOL’d since). And while intel’s processors have fallen behind to some extend the same isn’t true for AMD and EPYC, which simply rocks. People naively ignore that ARM is just an IP core, it’s only a part of what makes a processor. And Apple’s M1 works so great because it’s not just a processor but a complex SoC which aside from the CPU cores also contains lots of other dedicated hardware blocks for graphics, AR/VR, ML, image processing and other stuff, all supported by a completely in-house developed OS over which Apple has full control.

    But that isn’t true for everyone else. Microsoft has a long track record of sucking at hardware, from it’s various unsuccessful attempts with Windows on phones to it’s ARM efforts, which all have been failures. None of the big server vendors really have their own unique OS platform for commodity servers, and none of the server OS vendors other than Microsoft (Red Hat, SUSE, Canonical) have their own hardware, let alone the capability to develop their own CPUs.

    So no, outside the Apple platform x86/x64 will be going nowhere for the foreseeable future, because no-one other than Apple has all the necessary capabilities (OS development. CPU and GPU development, hardware development, services) under one roof.

    Now considering all this, what value is there in a virtualization solution for Apple Silicon Macs which can merely run Linux for ARM? For our business at least, exactly zero. Even worse, if we can no longer create x86/x64 VM images we can simply push out to our servers on Apple Silicon Macs it means a lot of the main reason we’re invested in the (very expensive!) VMware ecosystem will cease to exist, which means vSphere’s days in our datacenter will be counted, too.

  31. ARM based Macs will be all over the place by the end of this year and starting next year pretty much all Macs will be ARM based.

    The potential sales of Fusion are probably as big as they ever could be if only x86 emulation is added.

    On another side – Windows 10 for ARM is basically already covered by Parallels at $79,99.

    It’s hard to imagine Fusion users to upgrade in the future, where they could not prepare x64 VMs for ESXi… So, basically Parallels Desktop that comes out like half a year later and costs more? Good luck with this.

  32. M1 is not designed for X86 emulation. Yes, by using UTM (qemu) you can run x86 VMs but the performance it terrible. so yeah. that won’t be worth even thinking. however I’d like if VMware could support windows10 as well. otherwise I’d go for parallels.

  33. M1 is not optimised for X86 VM emulation. Yes, by using UTM (qemu) you can run x86 VMs but the performance it terrible. so yeah. that won’t be worth even thinking. however I’d like if VMware could support windows10 as well. otherwise I’d go for parallels.

  34. Do I want to run i.e Windows 2000 on my M1 Mac? Yes.

    Parallels and VMware have always been virtualization, not emulation products. The CPU architecture changed, so it’s an impetus for MS to get their butt into gear regarding turning Windows 10 on ARM into a retail-ready product for VMware and Parallels. Besides, the user-mode virtualization in Windows 10 works pretty well – I didn’t have any problems with things like Office 2000.

  35. “Well, the short answer is that there isn’t exactly much business value relative to the engineering effort that is required, at least for the time being. For now, we’re laser focused on making Arm Linux VMs on Apple silicon a delight to use.”

    That is an astonishing comment. Either gross stupidity, or more likely marketing drivel.

    A significant percentage of the people that I know who run Macs, run either VMware Fusion or Parallels.

    And EVERY SINGLE ONE of them primarily needed VMware Fusion or Parallels to run Windows, or more specifically to run old Windows X86/x64 applications.

    A very small number run Linux too but most linux software will run on Mac so. All of those running linux need Windows too.

    But more importantly, why do they run Windows? To run old windows apps that aren’t available on the Mac.

    Bluntly, they are not going to buy Windows ARM versions of those apps. Even for ones where ARM versions exist.

    But this could impact Windows users Migration to Mac and Apple have a vested interest in stopping that from happening.

    The logical conclusion is that Apple, to protect the Mac market share growth,. will provide its own VM facility with emulation built in from the bottom.

    In which case the market for VMWare Fusion and Parallels goes away overnight.

  36. Many on here are making UNREALISTIC demands on VMWare Fusion for the M1. [The numbers used for this are guestimates – but reasonably guestimate] For Rosetta to work efficiency it has to be integrated with the OS that the applications are running on, and that OS has to be ARM based. What Rosetta does is it translates the code of the APPLICATION on installation/first run to what would be equivalent to less optimized ARM applications. From that point on the application effectively is a native application. This application may be let’s say 80 – 90% as performant as an application compiled directly to the macOS M1 platform. The vast majority of the code any application had to run, was already written to be M1 native already (the operating system). In a typical application it may do some branching, some calculations, but the largest amount of code that gets executed is calls to the operating system. So an x86 macOS application is translated to ARM, but inside that application it makes calls like ‘draw window’, ‘draw line’, ‘compute using metal’, etc. These already run a very large amount of ARM based code already written for the M1 and never needed to be translated since these calls are exactly the same for M1 as they are for x86. For VMWare to be able to emulate x86 code it effectively has to have a software implementation of the x86 CPU, and x86 version of Windows runs on that software CPU, and the applications in Windows is then run on that OS that is running on this software emulated CPU. All of this would have to be done on the fly since Windows and the applications above are basically one big black box to the VMWare emulator. So even if VMWare emulated x86 CPU, the performance would be absolutely abysmal (I would doubt that it would even come close to running at 25% of the performance of native — and likely much less)… The performance of Windows itself would likely be unbearable to anyone wanting to run that application — to the point that they would complain about performance and then have to get a native x86 computer to run it on. This means that any effort that VMWare put into this would likely take much much longer to implement (I would guess any solution would take maybe at earliest end of 2022 to implement), and the performance would be so bad very few people would use it as a solution. There is a reason that emulators tend to only be written for obsolete CPUs of 30+ years ago (MAME), the CPU performance at that time is both simpler and the performance of the CPU at that time was likely so bad in comparison to today’s CPUs that it would be bearable (and since the platforms no longer exist – there is no option if you want to preserve the legacy platform/software – i.e. MAME). Summary: If you want to run x86 Windows software on a Mac, buy an Intel based Mac today while they are still available. If you don’t want an Intel Mac (legacy Mac), then buy a Windows based x86 system (and use your VMWare Pro license for VMWare Workstation…. there is actually some quality Windows based x86 devices. What people here are whinging about here … sorry, but that boat has sailed and VMWare will never support x86 Windows on an Apple silicon based Mac.

    1. I’m not very technical but somehow I think the x86 problem is better solved by windows/apple. Fairly complex windows applications are often able to run via wine without taking much of a performance hit. Wine does it’s best to emulate windows binaries but its not perfect. In any case it proves a concept. x86 windows apps CAN run well on M1 hardware under the right conditions.

      If Apple & Microsoft would work together a Windows for M1 release could be made. (apple has mentioned this). Ideally it would have its own version of Rosetta 2. I doubt we’ll ever see it, but it could be done. It’s a win win for both companies.

      1. WINE is running x86 code – not translating anything…. it all depends on how efficient it is written on how well it is running.

        At the lowest level binary (executable code may look like this). x05 – and that one byte of information means JUMP for sake of argument… actually for sake of argument lets say it means JUMP RELATIVE…. so x05 xFE means jump 125 bytes into the future – this takes the minimum amount of time (1 clock cycle). VMWare emulation would have to look for specific byte contents then do effectively IF JUMP RELATIVE THEN do whatever… and now that emulated code would take much longer (let’s say for argument 10 clock cycles). So the emulated CPU in this case would be 10 times slower — or basically Windows would run 10 times slower.

    2. “So even if VMWare emulated x86 CPU, the performance would be absolutely abysmal”

      So please explain why Parallels can do it and with reasonable performance?

      What many people here who just say “fair enough” to VMware’s descision to not inlcude x86/x64 support in Fusion on M1 ignore is that running legacy x86/x64 programs has been the main reason why Fusion exists in the first place.

      Which is why Parallels already supports running x86/x64 VMs on Apple Silicon. Not at the end of the year. Right. Now.

      And yet people like you claim it’s impossible or unreasonable. You couldn’t make it up if you wanted to.

      1. Except Parallels didn’t do it. They don’t run x86 VMs, they run ARM VMs.

        1. And this is where I made the foolish decision to purchase the M1 Mac. As a Parallels user for over 5 years – I checked their website and found that M1 is supported AND native support blah blah… Only after purchase did I realise that the text “Now runs on Intel and Mac” was a Hyper link – which takes us to a table which has a footnote that OS supported is ARM based.

          Have been kicking myself for this – Stuck with an expensive device worth nothing for me as of now. In this time of pandemic with businesses doing not so great – this has been a very bad experience.

          Request VMWare to put the M1 support details on the main product page so others don’t fall for such.

          Regarding Business decision – Virtualisation on Mac has been the only way to have a complete development environment. Now if x86 is out of the way then i guess the visualisation need also disappears.

      2. “So please explain why Parallels can do it and with reasonable performance?”

        It CANNOT…. It can run NATIVE operating systems only. It can run a preview copy (which you do not have a proper license for) of an Windows/ARM on the M1 and it can run Windows 10 x86 on an Intel Mac…. but it CANNOT run Windows x86 on an M1 Mac. The Windows for ARM does not run any x86 software. Not even all of Microsoft Software is available for Windows for ARM.

        Microsoft I believe is working on something called Cloud PC (i.e. a PC in the cloud somewhere you can run software remotely — pretty much the basis for XBox Cloud) that might be useful for some people.

        1. Windows for arm can run x86 and the recent insider builds can even run x64 software It is even possible to run x86 games and legacy apps on windows arm via virtualization on m1 based macs, but: – microsoft still needs to improve x64x86 emulation integrated in windows for arm – it will not work for drivers – some gaming cheating protections are not working – it is still beta not production version – it cannot be legally bought – it will not work for software that is only working on windows xp I suppose that most legacy apps that support windows 10 will work, but testing is needed.

          The best current solution is to rely on vdi type of technology ( vmware horizon) which is fully supported on macos. You will stream x86 apps from servers to mac workstations. This kind of service can be implemented on-premise or you can rely on cloud services.

      3. “2. Running VMs on Apple M1 Mac computers in Parallels Desktop 16.5 requires ARM-based operating systems (OSs). Customers who install guest operating systems in Parallels Desktop virtual machines are responsible for making sure that they are compliant with each OSs’ end-user licensing agreement (EULA).”

    3. Apparently WordPress doesn’t like the clapping emoji I was trying to use in reply.

  37. Why do a lot of people buy fusion? They want Apple hardware and OS but need to run X86 Windows software for work. I have customers who want to move to a new M1 machine but can’t because they need a windows VM to run software that just about runs on X86 hardware. Getting a lot of companies to compile an application to run on ARM is near impossible. Some seem to be using code that’s 10 years old because it ‘works’. Please reconsider your decision to not have an x86 emulator for the M1 or I can see a lot of people moving on from VMWare and not coming back.

  38. After reading this I feel my purchase of fusion since for Mac has been wasted by the company that collected my fees. I run Fusion on 4 different machines so I can carry one VM to use on them. It would be a shame to lose that flexibility because I will drop all the licenses except 1 and just keep one x86 Mac for running Fusion.

  39. Parallels will never get my money ever again after the last stunt they pulled… with their BS subscription… I can’t wait for VMware Fusion to come out for M1… I hope they are making a much better product than Parallels is, and im ok with waiting… But hurry… We need you! 🙂

  40. I’m afraid I must agree with so many comments here. I used Fusion to keep my Windows 7 OS alive after I migrated to Mac. I check with Apple before buying a new Mac M1 to ensure Fusion would still work. Of course I made the critical error of not specifying I had a Windows 7 VM I wanted to run and feel like I was misled. The VMWare forums I also felt were a bit misleading as that talked exuberantly about the exciting Fusion coming out and failed to mention that old Windows VMs would not be supported. My wasted investment in 2 new laptops an 6 months of waiting and hoping for a release that would do what I need is just that, a waste. A review of Parallels website is not much better. They say they can run Windows but it’s only Windows 10 as an install and not an existing X86 VM. Damn shame. I’d pay double to get the functionality I need, I just can’t have it.

  41. I look to Fusion to be able to run legacy s/w not current hardware/os/app combinations. I do use it to trial new OS version like Big Sur and Monterey. If these functionalities disappear it is no longer useful to me.

    1. Yeah what about us folk still on Intel Macs… where’s the Tech Preview with formal support for Monterey and Windows 11 for us?

      I’m happy to report it looks like 12.1.2 works for now anyway.

      1. Is Windows 11 on Mac possible at all (x64 or m1)?

        It needs a TPM 2.0 hardware module. I am not shure if that can be emulated by VM-Ware?

        I have the Win 10 arm insider preview installed on my M1 Mac under ACVM (Qemu) and I cannot run the Windows 11 Update because of missing TPM 2.0 support.

  42. On the use-case for x86 Windows emulation: I’m surprised that you don’t see this as more lucrative that Linux (although I can see why Linux is easier and therefore probably right to prioritize). We prefer to use Macs for our developers but our main product is a Windows application so we make heavy use of VMs running a number of Windows test and development environments on developer machines. It is so productive having everything on one machine. We stocked up on Intel Macs but soon we will have to revert to buying Dells for developers if there is no “light at the end of the tunnel” visible for x86 VMware on M1.

  43. “Insider builds of Windows 10 ARM may only be installed on systems with a licensed version of Windows 10, which is currently not available on Apple hardware.” — What?))) What about Win 10 arm insider running on my mac m1 inside parallels?

    1. Installing Windows on ARM on an M1 device (whether in a VM or a ‘theoretical but non-existent bootcamp’) is still presently a violation of the Windows EULA.

  44. I don’t have access to your user base and their desired usage; however, one of the most significant factors in selecting WMWave Fusion was the ability to run “historical” VMs and wisely or not I expected VMware to continue supporting at least one generation of hardware switching. So while you may have correctly calculated the ROI on supporting historical x86 VMs, you are going to loose the goodwill of some percentage of customers including myself for not doing so.

  45. Wow, there are tons of people who want EMULATION, not just VIRTUALIZATION, so based on the amount of comments I see here, there appears to be a market for it. Personally, I would disagree that there is no “business value” in emulating. VMWare historically never really cared to completely support Mac users, and I feel they made that decision based on percentages of users running Windows on PCs compared to Macs. Why throw money at such a ‘small market segment’ when the ROI wouldn’t support it, right? That thinking based on old data and such a mindset is a shot to the leg of the company in terms of branding and market share in the future. Those people aren’t thinking far enough ahead. The power of the M1 chip changes everything. This is only the first version. Subsequent versions will undoubtedly be much more powerful, and even in terms of emulation, these chips would be able to run emulated systems without any significant lag. That’s the whole reason for having a product that will emulate as well as virtualize. Because of the current mindset, VMWare will only support emulation once MS *eventually* changes over to an ARM architecture. Only then will they create such a product because only then will there be a large enough customer base to support the ROI. It’s just that in the meantime, someone else will likely fill the void, and in my opinion as a marketing professional, then VMWare will have lost out, unfortunately.

  46. With full ARM support for Windows apps being literally 5+ years off, there is no excuse to claim that there is “not much business value” in X86 VM emulation. I run a VM with Windows XP that has my engine diagnostic software I use at work every day. There will never be an ARM update to those apps. If you won’t make x86 support then I have no choice but to buy an old PC laptop just for the job. No more VMware software purchases.

    Thinking x86 is not still valuable is extremely bad judgement. VMware not supporting it is literally the ONLY thing stopping me from buying a new Mac.

  47. “Death is very likely the single best invention of life. It is life’s change agent. It clears out the old to make way for the new.” -Steve Jobs

    From all the comments about trying to keep old pieces of software alive, it seems like it is time for all the old, legacy x86 baggage to be laid to rest. (at least properly rest it on Microsoft’s shoulders) Sure it might be painful to lose something you are emotionally invested in. Change is hard and unsettling, but it is a fact of life.

    My impression is that Fusion is looking forward on how make M1 viable for new development and modern technologies, not on how to provide life-support for the terminally ill.

  48. I understand the market, technical and licensing issues, but it is disappointing. Most of us who are using Fusion have older Intel Macs lying around that run windows pretty well in bootcamp. My plan is to do just that and remote to that system when I need x86, although I recognize this might not work for everyone. My personal option is a 2008 Mac Pro dual quad core XEON that runs Windows 10 pretty damn well on an SSD with 16 GB RAM or a 2012 Retina Macbook Pro.

  49. I’ll eventually land in this same mire. I don’t particularly care that VMWare is Virtualization not Emulation. I bought it to run Windows on my Mac. I use the Mac as my developer machine and want my solutions to work for everyone. So I implement something on the Mac and then can then flip to Windows to review my changes work ok there. This workflow has been great for years but sooner or later I’m going to have to replace my laptop and I’m really hoping VMWare/x86 Windows will be available so I can go to the M1. Maybe I’ll still be able to upgrade to another Intel MacBook, but that’s not a future-proof path. Much of my testing is around digital accessibility, which means apps like JAWS and NVDA on Windows. I can’t truly say something works accessibly if I can’t run those x86 Windows apps. Seems like something that will be solved in a matter of time whether by VMWare or some other company. I guess it’s been easy in recent years as the Intel monoculture made Virtualization a sufficient solution for most. Back in the day of 68K and PowerPc it was a much heavier lift to emulate x86 and it sounds like we’ll need companies to go there again. I’m not expecting fast performance, but I do need something that will work.

    1. Maybe UTM for macOS is the way to go as it can emulate x86 on M1 and virtualize ARM operating systems on M1 Macs. I did not try it personally, but maybe it’s the path to go. Is x64 also supported by UTM ?

      Emulation of x86 in VMware Fusion would really be great but I doubt it will happen. Hope that MS will officially support and sell licences for Windows 11 ARM on M1. With the built-in x86/x64 emulation, this can be the solution for many people.

  50. When can we get to try our hands on VMware for M1 Mac to use linux atleast

  51. All our admins and devs use Macs and will be on M1’s in a year, and our datacenters will be x86 for years. You might not see business value in x86 on Fusion, but your customers sure do.

  52. So, Parallels announced their M1 (and windows 11) version. Any idea when we can expect a M1 Fusion version, or should I switch?

  53. Our blaming VMware is because they clearly state their disinterest in ever supporting Windows on M1. And how ridiculous is it to think buying another computer just to run an OS is a valid solution…IN THE FREAKING COMMENT SECTION OF A virtualization company blog.

  54. Many Linux add-on package repositories provided by Linux software vendors aren’t offering aarch/arm in parity with their x86/amd repos. So even if I update my OSes to aarch/arm – I’m still pushed back to x86/amd, and my platform developers are already switching to aarch.

    Will M1 Fusion have nested x86/amd virtualization support? That is part of our software development process which was met by x86/amd Fusion. Does the Rosetta interface offer nested x86/amd virtualization, or how much hardware emulation would VMWare have to do ( performance hit is acceptable ) to retain this on M1 Fusion?

    Sorry if you have to be a hardware emulator again, VMWare. I’d already forgotten those days.

  55. I get the sense VMWare isn’t not kidding about the Mac market. I’ve been a client for a very long time. Yet, when I moved up to macOS 11.3 a Windows VM abruptly began smashing. Maybe than pay them $50 for help (that is obviously excluded with my permit) and expectation that a future Fusion discharge settle the issue, I downloaded the Parallels preliminary. That far working extraordinary, so I’ll likely stay with it.

  56. With so much focus on Intel x86 in the comments here (understandably so), it’s important to balance that with an awareness that it can change from dominant to legacy architecture more quickly than many might realize. Microsoft, AMD and even Intel themselves have all seen the handwriting on the wall since the release of Apple’s M1. Qualcomm is already promising to match or exceed the M1’s performance and efficiency. Obviously, that remains to be seen.

    I (lifelong Windows user) purchased an M1 Mac Mini earlier this year and, shortly thereafter, began running a Windows 10 ARM VM in Parallels (since VMware is so far behind on M1 support). Not surprisingly, W10 ARM was rough around the edges. However, things changed rapidly over the next couple of months.

    I was aware that Microsoft has been working on ARM support for x64 apps, I was surprised to see Windows 11 for ARM64 upgrade was available at the same time Windows 11 x86/64 arrived on the Insider dev channel. I was installing it via Windows 10 ARM update on day one! (No TPM requirement in the Dev channel.)

    Over the following weeks, apps (including Microsoft 365) and Windows 11 itself have become dramatically more stable and performant—to the point where I would already consider using it as a daily driver on a personal, non-production device. Since Windows 11 is running so well in a VM on Apple’s M1, my plan is to continue running it virtually rather than upgrading my Intel 7th gen core i7-based HP laptop.

    If Parallels is working so well, why am I here checking on the progress of VMware Fusion for M1? Corel is overly aggressive at trying to force customers into its expensive Pro subscription. If you purchase Standard and a major upgrade is released a few weeks later, you may already be outside the grace period and forced to pay for the upgrade like everyone else. Also, the license is only good for one device, so things get expensive very quickly for many who, for example, have a desktop and a laptop.

    The final straw for me is, it looks like Windows 11 required TPM support will only be included in the expensive Pro subscription. The 2-core 8GB RAM limit for Parallels Standard edition should be a sufficient differentiator. Attempting to gouge Standard users with a lack of TPM support is too much.

    I hope VMware can come to the rescue and continue to offer options that aren’t hostile to enthusiasts, students, and small-time developers.

Leave a Reply