Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

tommiy

macrumors 6502
Dec 11, 2015
412
127
Done an install of ubuntu server and i hung on the install. Rebooted it after 20 min and it runs but always returns at the start to wanting to reinstall. Manually force it to next partition and it works. I'm just reinstalling to see if it does the same thing.

Update: Same thing on the reinstall. I wonder if it finishes on a default install rather than custom. Next try.

Update: OK the ubuntu server install fails at the end due to the fact it can not unmount the cdrom. Hit escape and enter.
 
Last edited:

Krevnik

macrumors 601
Sep 8, 2003
4,101
1,312
Done an install of ubuntu server and i hung on the install. Rebooted it after 20 min and it runs but always returns at the start to wanting to reinstall. Manually force it to next partition and it works. I'm just reinstalling to see if it does the same thing.

Update: Same thing on the reinstall. I wonder if it finishes on a default install rather than custom. Next try.

Update: OK the ubuntu server install fails at the end due to the fact it can not unmount the cdrom. Hit escape and enter.

20.04, or 21.04? I saw the same thing on 20.04, haven’t installed 21.04 from clean yet.

There’s also a crashing bug in 20.04’s version of libssl that was fixed in 21.04, that only seems to affect M1 systems, although it might affect any ARM64 CPU without ARM32 compatibility (this is a guess). So I not only hit the unmount issue, which Parallels doesn’t have, but I hit the libssl issue a second time because I forgot I had manually upgraded my 20.04 install to 21.04 on Parallels to deal with the libssl bug.

——

Bug-wise, I found a couple issues I haven’t seen mentioned yet:

* Importing a Parallels ARM VM gets converted to an x86 VM and refuses to run once converted. Whoops.
* I hit what appears to be a crash I didn’t see in the Parallels VM, but I’m not sure if that’s due to VMWare, or the fact that a couple bugfix releases of the server software I was using got released.

——

Performance-wise, VMWare is clearly doing something right here. I ran the same server VM (Minecraft w/Fabric + Lithium + Starlight + Carpet) on an i7 2018 Mac Mini in VMWare (Ubuntu 20.04) and an M1 Mac Mini under Parallels and VMWare (Ubuntu 21.04). Same server settings and world.

Performance as measured in ms per tick is fastest on the M1 under VMWare, then the i7 under VMWare, then the M1 under Parallels. In terms of power doing this: just having the server active with no connected clients on the i7 can cause it to jump from <10W to >30W, and I have to use auto-pause to let the i7 properly idle and keep the CPU under 50C, because the CPU seems to ramp up power states to handle the idle processing. Meanwhile, the M1 just kinda shrugs at the extra load, and stays under 40C when a client *is* connected. Harder to get power measurement from the M1 in this case though.

This is a sort of best case for the M1 though, since most servers can spin down to a true idle, while a game server can be doing some sort of game loop even with nobody connected. That game loop itself seems to be able to keep an Intel chip in a power state it doesn’t actually need for the load.

——

I think an M1 Mac Mini (or the upcoming higher end model) would make a very nice home server to be honest. Especially with how it handles scaling power consumption better than Intel/AMD for these sort of low load services I run. I attempted to run a small Ryzen system (after I upgraded my gaming rig) using ESXi, but it pulled 30W at idle, while the i7 Mac Mini pulls 1/3rd that, while the M1 even less. I can run two Minis at the same power budget as a single 6-core Ryzen system.

I do expect I’ll probably have a dual home server setup for the forseeable future though. Keep my 2018 for anything x86 I need, and let the ARM server take on what it can. I could probably save a tiny bit of power by moving my Pi Hole to a VM on the M1 at this point, and have one less physical device on the network.
 
Last edited:

jdb8167

macrumors 601
Nov 17, 2008
4,859
4,599
Meanwhile, the M1 just kinda shrugs at the extra load, and stays under 40C when a client *is* connected. Harder to get power measurement from the M1 in this case though.
Not really. Try man powermetrics
 

Krevnik

macrumors 601
Sep 8, 2003
4,101
1,312
Not really. Try man powermetrics

Note I said harder, not impossible. :)

Downside is that you can only get SoC power rather than say, the measurement off the primary 12V rail like you can with the Intel systems, which is a lot closer to what a kill-a-watt measures. While Apple's numbers tend to be conservative, they do rate the M1 Mini at 6.8W at idle, and based on powermetrics, it's not from the SoC which is using under a watt for the whole package. While I can say that the SoC on the M1 is pulling about 1-2W when players are active, that's not a complete/accurate comparison when trying to measure the difference in system consumption.

My i7's power consumption at idle is about 7-8W in real world measured at the 12V rail, while the CPU itself reports 1-2W. So while yeah, the M1 is clearly an improvement, it's hard to quantify how much without me hooking each up to separate kill-a-watts (which I don't currently have).
 

jdb8167

macrumors 601
Nov 17, 2008
4,859
4,599
Downside is that you can only get SoC power rather than say, the measurement off the primary 12V rail like you can with the Intel systems
You can if you are on battery on a MacBook. I have a bunch of command line tools that parse the ioreg battery stats that gives me the current battery power drain.

Edit: I use it in clamshell mode on battery to get the power used minus the display.
 

Krevnik

macrumors 601
Sep 8, 2003
4,101
1,312
Unfortunately, I don't have an M1 laptop to test that with. I'd still say that counts as "harder" when you have to do things like put it in clamshell mode, and rely on battery drain numbers rather than just have the thing reported in a way that tools like iStat Menus can then display. ;)
 

jdb8167

macrumors 601
Nov 17, 2008
4,859
4,599
Unfortunately, I don't have an M1 laptop to test that with. I'd still say that counts as "harder" when you have to do things like put it in clamshell mode, and rely on battery drain numbers rather than just have the thing reported in a way that tools like iStat Menus can then display. ;)
Point conceded.
 

mikeboss

macrumors 68000
Original poster
Aug 13, 2009
1,544
860
switzerland

Just Released: VMware Fusion 22H2 Tech Preview​




Fusion-TP-22H2-Testing-Guide

Tips/Techniques/Gotchas for the Fusion for Apple Silicon Tech Preview Revision 9 (30-June-2022)
 

jdb8167

macrumors 601
Nov 17, 2008
4,859
4,599

Just Released: VMware Fusion 22H2 Tech Preview​




Fusion-TP-22H2-Testing-Guide

Tips/Techniques/Gotchas for the Fusion for Apple Silicon Tech Preview Revision 9 (30-June-2022)
It looks like they are supporting Windows 11 Guests on Apple silicon. So I guess Microsoft is cooperating now since VMWare was adamantly against supporting Windows previously without Microsoft's support. This is very good news.
 
Last edited:
  • Haha
  • Like
Reactions: mi7chy and Darssh

Darssh

macrumors newbie
Sep 24, 2018
4
0
It looks like they are supporting Windows 11 Guests on Apple silicon. So I guess Microsoft is cooperating now since VMWare was adamantly against supporting Windows previously without Microsoft's support. This is very good news.
thank you for including download links 💞
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.