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.
8GB vs 16GB for Parallels / VMware
- Thread starter stuffx
- Start date
- Sort by reaction score
i was wondering the same but I also think no.
esxi is quite smart in how it tries to reuse ram used by the same program across different vm’s, for example if say you need 6GB for 1 copy of windows, 2 copies won’t need 12GB, but with protected and encrypted memory things that are not the hyper visor will struggle to know what can be reused across multiple vm’s.
that said VMware on M1 won’t be a hypervisor and won’t stop big sur from analysing what bits of a program are actively used and need to be in ram vs those that are seldom used and can stay in storage.
containers in big sur will be the most efficient way to run virtualised apps, but if you need to run x86 software and wanting to use VMware etc there may be significant other issues Due to M1 being ARM and having to translate x86 Mac OS apps via Rosetta, not sure how that deals with virtualisation like VMware.
esxi is quite smart in how it tries to reuse ram used by the same program across different vm’s, for example if say you need 6GB for 1 copy of windows, 2 copies won’t need 12GB, but with protected and encrypted memory things that are not the hyper visor will struggle to know what can be reused across multiple vm’s.
that said VMware on M1 won’t be a hypervisor and won’t stop big sur from analysing what bits of a program are actively used and need to be in ram vs those that are seldom used and can stay in storage.
containers in big sur will be the most efficient way to run virtualised apps, but if you need to run x86 software and wanting to use VMware etc there may be significant other issues Due to M1 being ARM and having to translate x86 Mac OS apps via Rosetta, not sure how that deals with virtualisation like VMware.
I have ran VMs on 8 Gb machines, and that was working well enough, provided I would not do too many things in the same time. So that might work.
But, I was using Linux VMs with 2 or 3 Gb of RAM. Not sure how much Windows requires.
I now have only 16 Gb or 32 Gb computers and it is much much more comfortable to run VMs now than it was when I was working with 8 Gb total.
Also, most of us pretty much always run one or two web browsers all the time, possibly with quite a few tabs open, and they consume more and more memory.
So, I would strongly advise getting 16 Gb for working with a VM, except for very specific use cases.
But, I was using Linux VMs with 2 or 3 Gb of RAM. Not sure how much Windows requires.
I now have only 16 Gb or 32 Gb computers and it is much much more comfortable to run VMs now than it was when I was working with 8 Gb total.
Also, most of us pretty much always run one or two web browsers all the time, possibly with quite a few tabs open, and they consume more and more memory.
So, I would strongly advise getting 16 Gb for working with a VM, except for very specific use cases.
I'm in the same boat exactly.
I'm toing and froing between MBA and MBP and between 8 and 16GB of ram.
I too plan on experimenting with VM's, Parallels or others.
I've pretty much decided it's going to be 16GB ram but still stuck between the two machines.
I'm toing and froing between MBA and MBP and between 8 and 16GB of ram.
I too plan on experimenting with VM's, Parallels or others.
I've pretty much decided it's going to be 16GB ram but still stuck between the two machines.
The Apple silicon macs work vastly different to other machines Where more RAM, 32GB+, is effectively required.
there is a chance that an M1 Mac with 8 or 16GB of ram would run more virtual hosts without performance penalty better than an intel machine with 64GB or more ram running equivalent Number of guests.
No. Here's why: Unless the Apple Silicon versions of VMware Fusion and Parallels Desktop support something similar to Hyper-V's Dynamic Memory feature, you'll always need to allocate exactly how much you need for each VM. 8GB, I don't care how efficiently used, is small. Unless runinng VMs is the only thing that you'll be doing and unless you're only running one VM at a time (with presumably 4-6GB of RAM), you're gonna want 16GB minimum.
I'm going to be buying another Intel MacBook Pro and I'll be configuring it with 32GB of RAM for this exact reason.
You are wrong. Apple Silicon Macs only work differently in terms of how RAM is managed and accessed. Virtualization's need for RAM isn't affected by optimizations to how RAM is managed and accessed, especially since traditional hypervisors STILL need to allocate RAM to guests for them to use. You will still need to allocate 2-6GB per VM (more or less depending on your workload) which will make that memory unavailable to the host operating system. Again, the fact that the GPU, Neural Engine, device controllers, or anything have simultaneous access to that unified memory doesn't negate that priciple.
If 8 Gig is ok for Intel Macs, what makes it poor for M1 ? If all you doing is ruining 1 VM, 8 Gig Mac is plenty. I think the swing to better performance as Apple touts the M1, automatically must mean you must have the best always...
I hate when we all side with that..
I hate when we all side with that..
8GB IS poor for Intel Macs. And it's especially poor for ANY Mac that is trying to run a virtual machine unless that virtual machine is either 4GB and the only process running or 2GB and is only one of a few processes running.
I think the key is "If all you are doing is running 1 VM" (not ruining
). In this case, it is likely that it works indeed.But as soon as you start a couple of web browsers with a significant number of tabs because you need them to test the VM or search about its configuration, or some meeting software like Zoom and cannot shut the VM off during this, then memory pressure will start to grow seriously.
If you need to give to the VM 3 or 4Gb of RAM, the right way to think about it is "will I do fine with 4 or 5 Gb RAM for everything else ?". Of course, if the VM is just 1 Gb, the question becomes a lot easier.
To sum up, trying to work with a rather low RAM configuration and using VM requires to know ones requirements very well.
There's no correct answer, it alls depends on what you're gonna do in Windows and what performance is sufficient for you. If you have nothing opened in macOS and just gonna open Notepad in Windows, then yes 8GBi is enough.
My 8GB M1 Air is often paging to the SSD just from having Safari browsers open. Virtualizing Windows on top of that obviously requires more RAM.
My 8GB M1 Air is often paging to the SSD just from having Safari browsers open. Virtualizing Windows on top of that obviously requires more RAM.
I run Windows VM today with 1,5Gb of RAM just fine.
32-bit Windows 10 has a system requirement of 1Gb of RAM for the 32-bit version.
It depends entirely on what operating system, what version of the operating system and what you want to do in the operating system and on the Mac.
I ran two Windows VMs on a 4Gb Mac without any problems between 2010 and 2012.
Windows 10 32-bit requires 1Gb of RAM. If you don't do any memory intensive task in Windows, you could probably get a way with 2Gb for the VM. Also use another browser than Chrome in both Windows and macOS.
6Gb for MacOS should be enough unless you are also doing memory intensive tasks in macOS at the same time.
32-bit Windows 10 is not the most ideal thing in the world. Nor is running any version of Windows so close to its bare minimum requirements. Plus, pretty much every 32-bit version of Windows since 7 has had memory leak issues of some variety.
So, yes, you can do it. SHOULD you do it is a different story entirely.
Plus, I'm pretty sure that, as far as ARM64 is concerned (which is primarily what the OP is concerned with), 32-bit Windows 10 won't be an option.
I've run 64-bit x86 Windows 10 with 2 GB of RAM. It DOES run, but the experience is crap. I'm sure the mileage is better on 32-bit x86 Windows 10, but given that the number of 64-bit Windows apps out there is increasing and that most people don't need to be running 16-bit Windows apps, it's not worth it when going with 16GB of RAM is a $200 upgrade, even with Apple Tax.
16GB would undoubtedly be better for any VM. Though I will say, it also depends on what you intend to do with it.
For instance, I know 8GB will be fine for me, as I only rarely use a Windows VM (I still have a Windows server, so if I need it, I use that mainly), the tasks I do occasionally perform in a Windows VM are so basic that a lack of memory isn’t really an issue. I can tolerate worse performance for 5 minutes worth of tinkering.
For instance, I know 8GB will be fine for me, as I only rarely use a Windows VM (I still have a Windows server, so if I need it, I use that mainly), the tasks I do occasionally perform in a Windows VM are so basic that a lack of memory isn’t really an issue. I can tolerate worse performance for 5 minutes worth of tinkering.
I think you have some things mixed up, but I’ll try to break it down. ESXi is a Type 1 Hypervisor that runs at the lowest level of hardware. VMWare Fusion is a Type 2 Hypervisor, and generally speaking while similar to ESXi, due to the way it’s getting access to resources is just different. Arguments can be on whether or not one is different than the other. VMWare Fusion (on Mac) is essentially using the same Hypervisor seen in VMWare Workstation. I know there are specific commonalities, but as someone thats worked in the industry for a number of years, I think it was worth spelling out.
In terms of what Virtualization looks like on the Mac... best I can tell, the existing Hypervisors simply don’t work. The impression I’ve gotten from posts I’ve seen is Apple is telling companies to simply use the system native Hypervisor.framework. I’ve used applications that use that method, and at least to me, on systems I’ve loved, Hypervisor.framework apps don’t do specific things that make me cringe to use Parallels Hypervisor. Parallels Hypervisor is actually very nice in terms of performance. But various other parts of the system suffer as a result of it. Instability (10x more likely to see crashes), performance (because Parallels goes so deep, it seems to be messing with how the kernal is doing scheduling, and/or resource management), and the like. I’ve heard peoples opinions that say otherwise, and thats fine. It’s an opinion, I’m just sharing mine.
The biggest thing that I wanted to correct is on Containers. Containers and Virtualization aren’t the same thing. They are both “abstractions” but the fundamentals are literally completely different. I could type in depth about my generalized understanding of Containers... but someone in here with far better understanding on this would chime in and tell me all the ways I’m wrong. Containers and Serverless is a HUGE goal of mine for 2021.
I’ll leave the containers conversation as this. Virtualization is about abstracting the hardware, and what runs on top of it. Containers is really about abstracting the operating system, and what runs on it.
In virtualization, Type 1 Hypervisor sit below the operating system, and it manages how operating systems above it behave, like ESXi, and to a much lower extent Hyper V. Type 2 Hypervisors, such as Parallels, and VMWare Fusion, sit on top of the operating system, and they need to try to get resources from the OS kernel to get resources for the virtualized operating system.
Hopefully that makes sense. I’m not taking any ownership if someone with a deep Container background wants to pick this apart. By this time next year though, I’ll be ready to take you to the mattresses!
16 GB minimum for running VMs. And it's sometimes a bit tight with my actual 16 GB MBP. Containers are a lot lighter, bit full blown VMs take much more RAM.
Absolutely not; go for 16GB. You need to keep in mind that with any VM software, you have to set the allocation of your system memory to that OS. It's not just available on the fly.
If you're VM-ing Windows 10, then ideally you would want to spit your RAM and allocate 8GB each to Windows and macOS.
Furthermore if you decide to use an external monitor, this will use-up even more RAM since the M1 has an integrated GPU.
If you're VM-ing Windows 10, then ideally you would want to spit your RAM and allocate 8GB each to Windows and macOS.
Furthermore if you decide to use an external monitor