1990s stuff. I had an AlphaStation 400 which ran Windows NT using FX!32 which was like Rosetta 2. Performance was actually decent though the hardware was not price-competitive.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Instead of Rosetta 2, why Apple doesn't make a Windows x86 emulator?
- Thread starter UBS28
- Start date
- Sort by reaction score
1990s stuff. I had an AlphaStation 400 which ran Windows NT using FX!32 which was like Rosetta 2. Performance was actually decent though the hardware was not price-competitive.
As I’ve attempted to point out, there is a difference between offering up control of a whole machine, versus sharing time on a machine. In the latter case, there’s more work involved ensuring isolation and the like which makes things more complicated. Existing tooling for that sort of virtualization is either still in beta (Orka, VMWare), or just came out of beta without full feature parity (Anka) for ARM Macs.
Edit: I will add that at least in businesses, there’s been a move towards setting up CI pools using bare metal by buying or renting machines. So it’s also possible that the demand for CI-as-a-service on Mac isn’t as high, and so organizations that offer it aren’t moving as quickly. Which ironically would push even more teams to setup their own CI pools. These pools can then be used with places like DevOps or GitHub.
I did comment that I believe Azure DevOps and GitHub share their system images for their build agents. The GitHub engineers may even be the ones responsible for the Mac images Azure uses these days. But GitHub and DevOps are already offering the same functionality on their Mac runners.
But without knowing what GitHub/DevOps use to isolate builds running on their machines, it’s hard to say if it’s a case of not buying hardware, or if like Anka, they are spending time porting their virtualization tooling (or waiting for it to get updated).
If Anka has been able to deliver, I don’t think it’s a blocker, but it does suggest that it is one area that folks are going to have to spend some time working on in order to get it right. Anka in particular was already using the Hypervisor framework prior to the M1 transition.
Last edited:
"Agencies" and "studios"... Talk about a bubble.
Edit: they could court the scientific and technical markets like they do the "pretty picture" markets. They make generally excellent laptops and desktops, but they prefer your market to mine. Which is why Apple is very much absent in much of the technical and scientific computing space, even if their early OS X marketing targeted it. It's a shame, really. I'd love to deploy a large M1 Ultra cluster for local HPC. it's just that a bunch of "Studios" or "Minis" aren't going to cut it. No good switching fabric. No redundancy. Limited RAM (pretty sure no ECC, either). So, nope.
Great for small problems, I suppose. Still might get one for home at some point once my Intel MacBook Pro dies, but not for work. Our problem sets are far bigger than "agencies and studios".... It's the focus on "does it run PhotoShop!?!?!" that drove much of the derision regarding Apple. Burt (Rutan - I rather liked his sarcasm) liked Ashlar Vellum, but the rest of the industry.... did not......
Last edited:
I understand that those applications run on macOS. Wouldn't it be better for Apple to offer a tier 1 hypervisorl? How would Apple's CI/CD service work?
Although they have the software to do so, MacStadium doesn't offer CI/CD runners on Intel-based mac minis by the minute, and I doubt they will on ARM-based mac minis when Orka is not beta.
I hadn't been aware, looks like that's true.
However, I'm still not sure I'd agree with characterizing it as "hacking", as Microsoft itself endorses the concept by providing a tool to do the same thing:
DISM can do many things to Windows images in various formats, including driver injection. Microsoft needs to support OEMs, IT departments, and more who want to roll their own slightly customized version of Windows installers.
It does sound like Apple reverse engineered just enough of DISM to re-implement the small subset of its functionality needed in Boot Camp, but that's because DISM is a Windows based tool and Apple needed something that runs under macOS.
All this reminds me of a product Microsoft made to run windows on Mac before the switch to intel… on PowerPC…
I had a PowerPC 970FX (A G5 rev B dual CPU 2.7Ghz) with 8 Gb of DDR and a GeForce 6800 GT DDL… you could barely run office on that thing…
I had a PowerPC 970FX (A G5 rev B dual CPU 2.7Ghz) with 8 Gb of DDR and a GeForce 6800 GT DDL… you could barely run office on that thing…
Download Microsoft Virtual PC for Mac Version 7.0.2 Update from Official Microsoft Download Center
This update is recommended for all users who are using Virtual PC for Mac Version 7.0 or 7.0.1. This update improves overall stability of Virtual PC and allows Virtual PC to run on Mac OS X Version 10.4.1.
www.microsoft.com
They bought Virtual PC from Connectix. (At least the Windows version, I'm unfamiliar with the PPC version)
I watched a coworker fire it up one time and I couldn't believe that people actually bought the product. Way more productive to just buy a Windows laptop to run Windows.
Why get into the CI/CD business yourself when you have GitHub, Microsoft, Travis, Apple all playing in that sandbox, and folks are investing in renting the systems themselves to have personalized CI pipelines? Especially when one of the services you can offer is renting the machines to GitHub/et al?
As for Apple’s service, there’s multiple WWDC videos from last year demonstrating what it’s capable of. How Apple implements it though, is up to them.
As for tier 1, vs tier 2… A tier 1 hypervisor is effectively a mini-OS in its own right, just stripped down. Apple could build one, but they’ve shown little appetite for building Server-focused OS variants. See the slow decline of OS X Server from a separate SKU, to an add-on, to something that can be considered EOL.
Microsoft didn’t make that, it bought it. It was originally Connectix VirtualPC. Microsoft bought it to kick-start what would eventually become Hyper-V.
One of the things it was able to do on pre-G5 hardware was convert x86 load/stores into little-endian PPC load/stores, keeping memory layout identical to an x86 PC while also not having to do endian-swaps after the load. The 970/G5 couldn’t do it, and performance suffered noticeably.
Connectix is probably one of the few companies I watched sell off pretty much everything of value in the early 2000s. VirtualPC to Microsoft, Virtual Game Station to Sony, leaving them with webcams that were about to become heavily commodified. Whoops.
Last edited:
It was so slow you could go for a smoke and it would still load windows when you came back…
They still bought it and released version 7…
Edit.
Yup the last version by connectix was 6.1 I just checked… it was okay for some stuff but still terribly slow… hyper-v is in another class running the same hardware level instructions on an hypervisor through hardware virtualization… not on the fly code conversion
Last edited:
Yup, if anything, Connectix was smart to sell when they did. Ignoring the Intel switch that caught everyone off guard, the G5 was a perf regression big enough that the product was basically doomed on the Mac anyways.
Microsoft I believe was mostly letting the engineering team finish their development cycle before pulling them full-time onto the Windows version and starting into the work for Hyper-V.
Then again they had access to hardware level virtualization something that was impossible with the G5 and x86 code emulation / conversion.
And you make a valid point about Apple’s interest in hypervisors and the lack of interest in OSX server…
Last edited:
At work I had both a PC and Mac on my desk. But at home I bought Connectix VirtualPC 4.0 with Win 98 just to use a couple of proprietary Windows 98 utility apps that controlled misc devices over rs-232. There were no Mac equivalent apps for those devices. I also had to buy a USB to RS-232 converter for that purpose and getting them both to work together with the Win driver on the PPC Power Mac G5 (remember the liquid cooled system built by Delphi Automotive - that should have told you that you were in trouble) was an adventure.
One cool thing about it was years later when Apple switched to Intel and VM Fusion came out Microsoft transferred the Win 98 license to Win XP to run under Fusion for no charge.
Last edited:
Codeweavers CrossOver. Have your x86-64 Windows apps run in Rosetta 2 on Apple Silicon Macs and away you go!
A later version of Windows Virtual PC actually got pretty close to Hyper-V in performance. I liked it better than Hyper-V for awhile, as it had better UI performance.
I've never had good luck with crossover, either on the Mac or Linux. It never ran the stuff I wanted it to run.
Maybe you're talking about something specific when you say "scientific computing", but Macs are plenty popular in scientific research settings (at least in data science, neuroscience, experimental psych, and biology).
The combination of a proper UNIX-based OS + broad compatibility with commercial applications (e.g. MS Office, Slack, Adobe Creative Suite, etc.) makes the Mac an appealing option for a lot of us in the field. The only drawback is a general lack of drivers for specialized hardware from National Instruments and other similar companies (they have terrible Linux driver support too), but we have dedicated PCs (or Windows VMs) for those sorts of edge cases. The lab I'm currently in, which does research around the intersection of neuroscience & kinestheology (basically research on motor learning & rehabilitation), is about 70% Mac users across the board. My brother-in-law did a PhD in genetics and there were plenty of Macs in his lab too (they even had a PowerMac G4 with an old-style plastic Cinema Display in the corner when he showed me around)!
For engineering, I know software and specialized hardware support is a lot worse. Not sure how much of a chicken-egg problem it is though, since software like AutoCAD and LabView are given poor-quality ports missing key features, which understandably limits adoption in the field.
Google and Oracle offer employees their choice of Mac or Lenovo. Mozilla offered Macs when I did some work there about 12 years ago.
My son's oncogenomics lab provide MacBook Pros (they gave him an M1 Pro MacBook Pro 14 with 32 GB RAM in January). They have large Linux systems for large workloads and pipelines.
Macs were popular with the researchers at our kids' university a decade ago.
I hear that IBM likes Macs as well.
I've seen very few in the physics and aerospace fields. Very few. Going back decades.
That was the marketing early in the OS X era, but they haven't kept up with it. If they did and worked on that, they could own the technical workstation market. They choose not to. I could run part of my toolset on OS X, but not all of it, where with Linux I can. So, why Apple? Workstation-level M2 chips with ECC and not insulting RAM limits would be great. Will we see that? Will they actively support or encourage bringing higher-end CAD and CAE applications to MacOS? Doubtful.
Some of the tools, but not all, are available for MacOS. More of them are available on Linux, so that's where the center of effort has gone. Plus, a common platform from desktop to supercomputer is attractive. There's enough of a difference between Mac OS and Linux to make that an issue, as well.
I've seen only 3 colleagues in over 20 years who used Mac systems as their work daily driver. Windows for Office and Linux for the heavy lifting is where things have gone. Frankly, Office on Linux without going through a browser would be the environment Apple said OS X could be.... considering software support everywhere else..
Why is every post of yours on this website so bitter?
Last edited:
I worked at IBM and they had a bunch back in the 2000s depending of the department. IBM is also one of the creators of the power architecture… it’s still going strong today…
Because current solutions don't work. Apple doesn't need to get into the CI/CD business, they could play nice with cloud service providers by selling them M1 SOCs or letting them run macOS on non-Apple hardware.
After almost two years of Apple Silicon, there is no cheap CI/CD service to help developers test their apps, so Apple should step up and fix the situation.
I remember Genentech once (or maybe still?) used a lot of Macs, and Arthur Levinson is Apple's Chairman of the Board.
Anecdotally Ken Thompson and Brian Kernighan, two prominent computer artists (/s), use Macs.