Parallels tech preview for M1

[UCDHIUS]

OK.. Who is pissed off that the now retail version of the standard Parallels 16.5 has a core allocation limit of 4?

When the whole time, for the technical preview had the option up to (8 or all cores..)

I just installed 16.5 retail and i'm maxing out all 4 cores with something as simple as browsing the web with chrome.

Increasing core limit throws up the error "please upgrade to the pro version"

I honestly think Parallels should get rid of that artificial limitation for these arm macs.

 

[Krevnik]

OK.. Who is pissed off that the now retail version of the standard Parallels 16.5 has a core allocation limit of 4?

When the whole time, for the technical preview had the option up to (8 or all cores..)

I wonder if they are doing this to avoid allocating efficiency cores to the VM? I’m not sure how well those actually play with virtualized OSes in practice. It might have been something they were experimenting with initially but then decided against?

 

[UCDHIUS]

I wonder if they are doing this to avoid allocating efficiency cores to the VM? I’m not sure how well those actually play with virtualized OSes in practice. It might have been something they were experimenting with initially but then decided against?

I think it's more to do with making money on the "pro" version especially with the "Please upgrade to pro version for more cores" notification. Oh well maybe if we get enough people to "complain" they might change the policy for M1 macs? Chances are slim though lol.

 

[haralds]

OK.. Who is pissed off that the now retail version of the standard Parallels 16.5 has a core allocation limit of 4?

When the whole time, for the technical preview had the option up to (8 or all cores..)

I just installed 16.5 retail and i'm maxing out all 4 cores with something as simple as browsing the web with chrome.

Increasing core limit throws up the error "please upgrade to the pro version"

I honestly think Parallels should get rid of that artificial limitation for these arm macs.

I use the Pro version but in general, rarely allocate more than 2-4 cores - even on machines with plenty more. Apart from specific jobs, I have not seen any speed impact with more cores.

 

[UCDHIUS]

I use the Pro version but in general, rarely allocate more than 2-4 cores - even on machines with plenty more. Apart from specific jobs, I have not seen any speed impact with more cores.

I used the "reclaimed" space button and cpu usage has gone down. We'll see how long that lasts.

 

[shimpster]

Edit: Misread that it was the retail version that wasn't allowing over 4 cores....I stand Corrected

Im running the Pro Subscription License for Parallels and its showing me I can go up to 8 Cores on my ARM Win 10 VM

 

[haralds]

I use the Pro version for work on all three of my machines - Mac Pro, Intel MacBook Pro, and MacBook Pro M1. Still have a Vmware Fusion Pro license, but am converting. I like what I see. Fusion does not recover space on macOS guests for one. The Developer menu on Parallels pro has nice features. And it is released on M1.
We'll see when VMware Fusion for M1 ships. The only way to outdo Parallels is if they release Intel emulation.

 

[leman]

OK.. Who is pissed off that the now retail version of the standard Parallels 16.5 has a core allocation limit of 4?

M1 only has four CPU cores for anything that matters. What's even the point of allocation efficiency cores?

 

[jdb8167]

M1 only has four CPU cores for anything that matters. What's even the point of allocation efficiency cores?

Well in my testing all 4 efficiency cores seem to add up to about one more performance core--maybe a little less.

 

[Gerdi]

We'll see when VMware Fusion for M1 ships. The only way to outdo Parallels is if they release Intel emulation.

The biggest tangible "benefit" of Intel emulation would be much slower program execution - both because emulation is much slower and you cannot run ARM programs anymore. Do i miss an argument here?

 

[Internaut]

M1 only has four CPU cores for anything that matters. What's even the point of allocation efficiency cores?

As far as I can tell, they’re the only cores that get used most of the time (at least for my workflow). They get used when they needed, for things like Lightroom Enhance Details or running up a virtual machine and making everything look like bare metal performance?

 

[haralds]

The biggest tangible "benefit" of Intel emulation would be much slower program execution - both because emulation is much slower and you cannot run ARM programs anymore. Do i miss an argument here?

It would allow to run older versions of macOS, if implemented fully. This is sometimes required to debug product on older versions of the OS. On Intel VMware I have VMs all the way back to Tiger that I occasionally pull "off the rack."
Just a couple of days ago I used Sierra and Xcode 7 to look at unit test results of one of our computational libraries to compare the output to a current version.

 

[bobcomer]

The biggest tangible "benefit" of Intel emulation would be much slower program execution - both because emulation is much slower and you cannot run ARM programs anymore. Do i miss an argument here?

Yep, big time. Not everything Windows runs on WOA, not to mention if you need or want some other x86 OS to run. (Like 32-bit MacOS perhaps)

 

[jdb8167]

It would allow to run older versions of macOS, if implemented fully. This is sometimes required to debug product on older versions of the OS. On Intel VMware I have VMs all the way back to Tiger that I occasionally pull "off the rack."
Just a couple of days ago I used Sierra and Xcode 7 to look at unit test results of one of our computational libraries to compare the output to a current version.

The only complete AMD64 emulator that I’m aware of that runs on the M1 is QEmu. I’ve managed to get it running via UTM running Ubuntu 20.04. It is very, very slow. I’d assess it as unusable. I don’t know if it would be good enough for this use case but it would be frustrating to use for any extended session and I’m sure that Ubuntu uses less resources than MacOS. In its current state I think you would have to “Hackintosh” the virtual machine to get MacOS running too.

 

[Gerdi]

It would allow to run older versions of macOS, if implemented fully. This is sometimes required to debug product on older versions of the OS. On Intel VMware I have VMs all the way back to Tiger that I occasionally pull "off the rack."
Just a couple of days ago I used Sierra and Xcode 7 to look at unit test results of one of our computational libraries to compare the output to a current version.

I really do see the point of running older version of Mac OS, thanks! I guess it is still not something the average user would need.

 

[mikeboss]

I think OS ARM native stuff will come along eventually but running something like SQL Server or Oracle natively on ARM is what I am worried about for now.

SQL Server on ARM -> https://threedots.ovh/blog/2021/04/sql-server-on-arm/

 

[bobcomer]

The only complete AMD64 emulator that I’m aware of that runs on the M1 is QEmu. I’ve managed to get it running via UTM running Ubuntu 20.04. It is very, very slow. I’d asses it as unusable. I don’t know if it would be good enough for this use case but it would be frustrating to use for any extended session and I’m sure that Ubuntu uses less resources than MacOS. In its current state I think you would have to “Hackintosh” the virtual machine to get MacOS running too.

For occasional work, it would at least for what I want, the trouble with it is that the network stack isn't good enough. A better emulator wouldn't be *quite* as slow either, but still slow.

 

[UCDHIUS]

Edit: Misread that it was the retail version that wasn't allowing over 4 cores....I stand Corrected

Im running the Pro Subscription License for Parallels and its showing me I can go up to 8 Cores on my ARM Win 10 VM

View attachment 1758883

Yup, it still the same policy they have had for the intel macs for years, on the "normal" version vs pro as a "perk".

 

[bobcomer]

Yup, it still the same policy they have had for the intel macs for years, on the "normal" version vs pro as a "perk".

And that policy also carries a subscription model for pro, so you pay the price every year, rather than a perpetual license.

I have the pro version for my Intel Mac, mainly because it has the cores to actually use more than 4, but for my M1, not a chance, it can't even do 4 cores without getting too hot, and I never allocate all the cores of the machine to one VM no matter what base OS I'm using or architecture. (It's an MBA)

 

[bobcomer]

For occasional work, it would at least for what I want, the trouble with it is that the network stack isn't good enough. A better emulator wouldn't be *quite* as slow either, but still slow.

Check out the other thread about emulation settings, it actually makes it a lot better. I even got a VPN to work now too, so my comment about networking is a little less of a problem.

 

[Krevnik]

As far as I can tell, they’re the only cores that get used most of the time (at least for my workflow). They get used when they needed, for things like Lightroom Enhance Details or running up a virtual machine and making everything look like bare metal performance?

I actually wonder how scheduling for more than 4 vCPUs works on these systems. How aware does the guest OS have to be to use them effectively? Is the Apple hypervisor helping with scheduling so that the guest OS doesn’t have to be aware?

 

[bobcomer]

I actually wonder how scheduling for more than 4 vCPUs works on these systems. How aware does the guest OS have to be to use them effectively? Is the Apple hypervisor helping with scheduling so that the guest OS doesn’t have to be aware?

I'm curious about that too.

 

[Gerdi]

I actually wonder how scheduling for more than 4 vCPUs works on these systems. How aware does the guest OS have to be to use them effectively? Is the Apple hypervisor helping with scheduling so that the guest OS doesn’t have to be aware?

Thread scheduling is totally owned by the guest OS.
That having said, ideally the guest OS also needs to be aware of the heterogenous nature of the cores in the M1. The Windows scheduler for example is aware of the big.LITTLE nature of the SQ1/2 SoC in the Surface Pro X.

The background here is, that Windows using UEFI/ACPI to configure itself and cope with architectural differences. UEFI/ACPI is the central element, which enables the same binary Windows distribution to run without changes on many ARM platforms with very different architectures. So part of the Information UEFI/ACPI provides is the number of cores and how to handle the clocking and power modes. UEFI/ACPI provides interfaces in 2 forms: tables and procedures. The procedures are implemented architecture-independent in a language called AML.

 

[hunkster]

OK.. Who is pissed off that the now retail version of the standard Parallels 16.5 has a core allocation limit of 4?

When the whole time, for the technical preview had the option up to (8 or all cores..)

I just installed 16.5 retail and i'm maxing out all 4 cores with something as simple as browsing the web with chrome.

Increasing core limit throws up the error "please upgrade to the pro version"

I honestly think Parallels should get rid of that artificial limitation for these arm macs.

Didn't know about the 4 cpu limit for the one-off version until you said it.

This might be more important when the newer Macs get released this year with more cores.

So the only way to run with more cpus is to go the subscription route?

 

[Lights87]

Is there any risk that (assuming i get window pro) that microsoft eventually decides not to develope an official windows arm AND removes access for all existing insider previews?

I'm somewhat ok if they decide to never develop an official version, but wouldn't want to get a new M1 mac with parallels only to find in a year or two that i can't use even old versions of windows ARM.