Parallels Hypervisor vs Apple's Hypervisor?

Same Problem here: Since Mojave 10.14.4 and later PD is sucking up all resources on my iMac Pro. Windows 10 (brand new VM, just couple of days old) is coming to a halt. Multiple sessions with support didn't help anything so far. My problem is that I am also using Docker, so I need nested virtualization...

 

Also during startup or any IO-heavy activities (like loading a VS project), i sometimes see "System Interrupts" task taking up to 100% of the CPU (I allocated 4 of 8 CPUs to PD). This seems to be a sign that some driver is causing heavy interrupt load on the system, which would explain why it's so slow...

 

I don't buy it.

I have switched to Apple hypervisor in both my Macbook Pro 15" and my iMac 27" and have noticed much better performance as well as better resource utilisation.

Using Parallels I experienced longer delays, especially to compile code and install softwares. Besides, my CPU usage went to the roof all the time, and the RAM pressure was also much higher. This caused more heating and higher fan speeds, noise, etc, without giving any performance boost.

Using Apple hypervisor both my machines run with much better performance, and much quieter too. The CPU and RAM only go up when required, otherwise they stay very low and quiet. The Mac seems to be doing a better management of resources resulting in faster and more efficient workstations.

 

I've been running a brand new MacBook Pro 15" for a few months now, running the latest MacOS and Win 10 with the latest PD. Even when not running Windows apps, my Win10 CPU usage was massive, and my fan ran crazy high 100% of the time. It was so concerning that I would shut down the VM and only run when needed (I used to keep it on in the background all the time).

After doing some reading here, I switched to the Apple Hypervisor.

Only about 15 mins into my test, but so far the difference is very noticeable. Fan not going full out. VM booted up just as quickly.

So far, I am Team Apple Hypervisor.

 

Just to add my similar experience - I had been running Parallels Hypervisor since day one of my iMac Pro, performance better than I'd ever hoped in Windows until 10.14.4 and then everything slowed down to unusable speeds. After some reading (e.g. https://forum.parallels.com/threads...access-since-update-to-mojave-10-14-4.346849/) I switched to Apple Hypervisor and it's much improved. Still feels like performance is only about 80% of pre-10.14.4, but at least it works. Every time there's a macOS or Parallels update, I've given it a go on Parallels Hypervisor but the issues prevail - back to Apple HV for now!

 

Hi people. A lot of people said a lot here. Just to make it short:
MacbookPro 15" (2019) with 2TB SSD, Core i7 and 32GB RAM, OSX 10.15.3. Parallels Business version, VM has got 6 VCPUs vom 12 available and 16GB RAM from 32GB available.

Until today i had really, really big troubles with my Windows 7 pro (now Windows 10pro ). Problems means : After a while of usage of the VM and lots of used programs my VM got slower and slower everytime. At the same time the system fan was starting to make heavy noise. The only thing what helped a little bit was to start the VM new every day. And to keep an eye on tasks which take more than 10%of CPU usage and which were maybe not 100% necessary for the ongoing work. But is that the behavior i bought an 4.500 (!!!) Euro MacBook for ?!?!? I was really not amused, but i couldnt find any help somewhere. My feeling was that this behavior was going more and more worse with every new Parallels version. Some days ago my Windows 10 VM was not usable anymore after 2-3 days of usage. I always had to restart.

Now the solution:
I was so desparated that i checked out every single setting and parameter in my VM. Everything was correct and along the recommendations of Parallels.
THEN i saw this discussion thread here and i thought : Hey, why shouldnt i change the Hypervisor to Apple and test it ?! And i can only say : That was the best thing i ever could for my VM. Since i changed to the Apple Hypervisor, my VMs behavior was totally different than before. I can work for many days and without restarting my VM. Its totally quite (the fan) and i do not have the problem like before that even some simple applications suck nearly all of my CPU capacities. So actually my problem is solved and i can only recommend everyone with similiar problems to try changing the Hypervisor for a day. The ONLY negative thing i have felt is that Windows 10 takes a little bit longer foor booting, but that is totally acceptable. When the system is running well, you DONT have to reboot your system ...

The last thing i have to say that i am really wondering about this. Shouldnt the Parallels Hypervisor not be much better than the Apple one ?
I hope i could help someone with similar problems. Please let me know.

 

Very similar situations to most users here. I got my 2019 16GB MacBook Pro. Need to run some CAD software on Windows 10. The memory pressure remained on the edge before turning yellow or red. I was worried that my laptop cannot run more tasks, other than the CAD and browser. And thinking whether 16GB was a right purchase for me.
Yet, out of curiosity, I switched the hypervisor to Apple, and I notice the memory pressure is SIGNIFICANTLY lower. Lower than 1/3, if viewing from system monitor. And now I am happily using my CAD in "parallel" with the many Mac apps.
What I have observed is that, using Parallel's hypervisor, many memory of Windows are "Wired Memory", meaning that it cannot be swapped, or compressed. This may be contributing to the memory pressure that its having (maybe not a issue at all? maybe using wired memory ensures a good performance? Who knows). While using Apple's hypervisor, all memory of the VM is "App Memory". So basically delegate resource management to macOS itself. Maybe not a bad thing to have...

 

The Parallels hypervisor is unusable for my application. I have a 2012 Mac Pro with 12 cores and 64 GB Memory. I could not get a Win10 VM working with the Parallels Hypervisor. It's broken. The Apple hypervisor makes a word of difference.
Do NOT even ask about AV. It's irrelevant. MS's AV is mandatory in these environments anyway.

 

I think, everyone should experiment with the hypervisor and choose what works best. Parallels hypervisor, on my system, boots linux about 1 sec faster, than the Apple hypervisor. The Apple hypervisor needs much more memory. I gave the linux machine 8 GB. After startup, the parallels hypervisor claimed 600 MB for my Linux system. The Apple hypervisor claimed 2.03 GB for the same machine.

 

Something fishy going on.

Apple Hypervisor
* macOS Activity Monitor shows
** negligible memory pressure
** 9.9GB for FreeBSD
** 178Mb compressed memory
* Parallels Resource Monitor shows
** Mac 2.2GB
** FreeBSD 3.5GB

Parallels Hypervisor
* macOS Activity Monitor shows
** significant memory pressure
** 343Mb FreeBSD
** zero compressed memory
* Parallels Resource Monitor shows
** Mac 6.8GB
** FreeBSD 4.7GB

 

There seems to be a severe bug in the Parallels hypervisor on machines with a large number of cores. I've done some experimentation with Geekbench 5 on a 16 core Mac Pro.

The Parallels hypervisor doesn't scale well much beyond about 8 assigned VM cores (equating to about 4 physical cores). The multicore score maxes out at about 3000-3100, and then degrades, as I assign more cores to the VM.

In contrast, if I use the Apple hypervisor, the VM scales well, getting a performance boost all the way up to 24 (out of a possible 32) assigned VM cores, getting a multicore score of about 10800. After that point gains diminish before degrading. Switching off the adaptive hypervisor gives a further boost to give a multicore score of 11334, although whether you would want to do this really depends on your load in the host OS. This is much more inline with what I would expect from a VM - a marginal loss 5-10% from not running on 'bare metal'.

This performance gain in the benchmarks translated to real-world usage. Testing with Visual Studio, using a third-party multibuild tool, it took about 5-6 minutes to build a 300 project build with the Parallels hypervisor. With the Apple hypervisor that time was brought down to 2 minutes. Boot time was not affected (as suggested in previous posts). I didn't time that but it was quicker.

 

In my retina iMac from late 2014 all maxed out, I have kernel panics with the Parallels hypervisor. Does not happen on the last retina MacBook Pro from 2015. I think I notice better performance with Apple on the iMac but worst on the MacBook Pro. Nevertheless I now use Apple hypervisor in both machines. I cannot afford to have random kernel panics which often screw up the VM requiring me to restore hundred of gigabytes. As I work with embedded software development tools, I am tied to Windows.

 

I wasn't aware of the Apple Hypervisor so how do you start it and create a new VM.

 

OK, I now understand its an option in Parallels and not one I have tried. Will take a look. My 2016 MacBook Pro with 16GB RAM and Catalina suffers various freezes and in the worst case the machine never becomes usable and the only option is a hard power off which I hate doing. I've tried so many configurations of RAM, CPU etc but none seem to be ideal as over time Parallels just seems to not want to release resources and kills the MacBook Pro performance.

 

What I have found is that Apple Hypervisor works better overall than the Parallels version. Particularly in the area of resource sharing between the two.

However, if you upgrade OSX say from High Sierra to Mohave you MUST reinstall Parallels tools. Failure to perform this step will cause a number of programs to hang with the beachball of death, etc.

 

In my testing with Parallels 15 on MacOS 10.15, I've seen a similar pattern, though not as pronounced. A Linux VM was said by Parallels to use 1.5GB under HyperKit, 2GB under Parallels, while Activity Monitor gave 3.5GB and 1.6GB respectively.

I also did some measurements of disk performance, since it affects my use case (development with native and shared directories). Test was tarring and untarring a 1.8GB tree with 13K files. I found that:

• For within-VM trees, the performance with the parallels hypervisor was around 5% faster than hyperkit.
• For shared directories, parallels was around 20% faster.
• Relative to native MacOS disk performance, the VM with parallels hypervisor gave comparable performance for within-VM trees, and was about 2x slower for shared trees.
• Multipass (for those of you interested) was slower than Parallels on within-VM by 20% and slower on shared by 5-10x.

My conclusion was to stick with the Parallels hypervisor for now (and to use multipass only for non-performance critical tasks, convenient and lightweight though it is).

 

I am running a 2017 iMac with Win 11. I run the Mac off an external SSD drive because the Mac's caching does not work with the Parallel's file size. On the Win 11 VM, I'm running Visual Studio along with SQL Server. Visual Studio has been crashing a lot, so I switched to the Parallels hypervisor, and so far, it's running clean. Seems faster, too.