eschultz Posted June 19, 2018 Share Posted June 19, 2018 Can anyone here confirm the 20-25% cpu usage for an idle Windows 10 VM, without HPET, that's just using VNC graphics (no GPU passthrough)? Quote Link to comment
Taddeusz Posted June 19, 2018 Author Share Posted June 19, 2018 1 minute ago, eschultz said: Can anyone here confirm the 20-25% cpu usage for an idle Windows 10 VM, without HPET, that's just using VNC graphics (no GPU passthrough)? Yes, my Windows 10 vm just uses the VNC vgpu and I had the 20-25% cpu usage until I turned HPET on. Quote Link to comment
DZMM Posted June 19, 2018 Share Posted June 19, 2018 On 6/18/2018 at 4:36 PM, DZMM said: Well, I've been using my PC now for 7 hours today and I've not had one crash and although I get the occasional freeze they're not as drastic as before - maybe a second at most. Normally, I would have had 2-3. Looking promising so far. If I make it through a couple of days with no freezes (kids will be online as well soon, so will get more data) I think @limetech might need to look into this and/or change the W10 VM defaults ok, had my first crash today which caught me out as I'd gone back to my old carefree ways and wasn't constantly saving stuff.... Had the kids on their VMs this evening and experience was bad - still got lots of freezes. I might rollback Quote Link to comment
Spatial Disorder Posted June 19, 2018 Share Posted June 19, 2018 47 minutes ago, Taddeusz said: Yes, my Windows 10 vm just uses the VNC vgpu and I had the 20-25% cpu usage until I turned HPET on. Yep, same here @eschultz Quote Link to comment
eschultz Posted June 20, 2018 Share Posted June 20, 2018 4 hours ago, DZMM said: I might rollback Your saying 6.5.2 was completely stable for you, just become unstable at 6.5.3? Quote Link to comment
DZMM Posted June 20, 2018 Share Posted June 20, 2018 4 hours ago, eschultz said: Your saying 6.5.2 was completely stable for you, just become unstable at 6.5.3? I'm pretty sure my problems started with 6.5.3rc1 - I posted about my experiences (including GUI mode not working anymore) in the release threads and sent a couple of bug reports. I'm going to rollback to 6.5.2 today to confirm. Quote Link to comment
NewDisplayName Posted June 20, 2018 Share Posted June 20, 2018 (edited) Didnt read everything, but i use a windows 10 Vm every day. Do you pinned the CPUs? I found when pinning the CPUs (so e.g. only use 2 out of 4) i had big cpu usage, after chaning it back to non pinnged (using all 4 cores) I can run all my stuff (many dockers and plugins and VM) with 10-20%. Anyway important is that ur vm has enaught RAM (atleast 4gb for windows 10 i would say) (and unraid also, depends on what u use unraid for) - but u should always have enaught free RAM, strange things happens when u run out of RAM. Edited June 20, 2018 by nuhll Quote Link to comment
DZMM Posted June 20, 2018 Share Posted June 20, 2018 3 hours ago, DZMM said: I'm pretty sure my problems started with 6.5.3rc1 - I posted about my experiences (including GUI mode not working anymore) in the release threads and sent a couple of bug reports. I'm going to rollback to 6.5.2 today to confirm. Ok, 2 hours into my rollback to 6.5.2 and I haven't had a lockup or a crash yet on my PC (Buzz). The other VMs (LEGO & Disney haven't been used yet and are just at the login screen. They're running at roughly 1/3 of a core (got 4 each) which seems sensible to me. As a sidenote, I've never really paid attention to cadvisor and I've just spotted it shows how many cores docker is using. It's showing an average of 2.5 for me, so I'm going to pin more of my dockers so that my cores are free for my VMs Quote Link to comment
NewDisplayName Posted June 20, 2018 Share Posted June 20, 2018 How u actually changed the HPET Thing? Set OS to other? One ppl in the thread linked said that this was just a quick fix and stopped working!? Hmmm after some tests, the option Other OS seems not working ...I've many Windows 10 1803 guest that comsume a lot of CPU on the PVE hosts ...Arghhhh Quote Link to comment
Taddeusz Posted June 20, 2018 Author Share Posted June 20, 2018 1 hour ago, nuhll said: How u actually changed the HPET Thing? Set OS to other? One ppl in the thread linked said that this was just a quick fix and stopped working!? You have to manually edit the XML. Quote Link to comment
NewDisplayName Posted June 20, 2018 Share Posted June 20, 2018 (edited) Only thingi found was: Quote <clock offset='localtime'> <timer name='hypervclock' present='yes'/> <timer name='hpet' present='no'/> </clock> Just change it to yes, or what? (doesnt make sense for me) edit: okay it seems to be right, i try it. edit2: Yea, it seem to reduce the idle (when im not logged in) CPU from 20-30% to 10-15%, but i didnt notice any performance increase, but it looks better... Watt is also the same. (75w) While logged in i would say about 5%, or no change at all. Does HPET effect anything? I read about it (https://en.wikipedia.org/wiki/High_Precision_Event_Timer) , but i guess, thats to technical for me. And i dont understand, some ppl say it gets activated anyway onle when a application needs it, doesnt matter what u put in bios (enable/Disable). So why does it make a difference in the first place? Edited June 20, 2018 by nuhll Quote Link to comment
DZMM Posted June 20, 2018 Share Posted June 20, 2018 10 hours ago, DZMM said: I'm pretty sure my problems started with 6.5.3rc1 - I posted about my experiences (including GUI mode not working anymore) in the release threads and sent a couple of bug reports. I'm going to rollback to 6.5.2 today to confirm. @eschultz rolling back to 6.5.2 has definitely solved my freezes and crashes. The kids are currently on their VMs for their PC time and we haven't had a freeze or crash - yesterday we were getting them every 1-2 minutes Quote Link to comment
isra007 Posted April 22, 2019 Share Posted April 22, 2019 Greetings, I been looking across all the different forums with no light on the same issue. I'm currently using a Xeon E5 2697 v3 passing thru 10 cores and their pair threads and an GTX 1070 to a windows 10 Pro VM. I literary tried everything from isolating cores, to changing HPET settings on the xml file. Any ideas on how to solve this issue. CPU utilization on idle on the VM's first core is between 35% and 45% reported by the unraid dashboard and the wattage is about 30 watts higher as well. Cheers... Quote Link to comment
NewDisplayName Posted April 23, 2019 Share Posted April 23, 2019 The display from unraid cant be trusted. Install nerdpack and with it htop and ssh into your server and then type htop to see REAL cpu % Quote Link to comment
isra007 Posted April 23, 2019 Share Posted April 23, 2019 Thanks for the reply, I ran the command htop and pretty much the same results. Quote Link to comment
NewDisplayName Posted April 23, 2019 Share Posted April 23, 2019 (edited) How much % your windows 10 vm has (windows task manager)? How much does unraid have (htop?) When idle? Also did you try hpet? Edited April 23, 2019 by nuhll Quote Link to comment
mdrodge Posted December 7, 2020 Share Posted December 7, 2020 (edited) On 6/17/2018 at 6:54 PM, Taddeusz said: Wow, thank you! That fixed it right up. Now back to a low rumble when idle. Just so people know. Go to the clock section and change hpet present to “yes”: <timer name=‘hpet’ present=‘yes’/> Is anyone able to explain this. I don't understand where the clock section is. I'm guessing you are saying edit the vm xml with this? Can someone give an example please. Edit/update Oh wait I see. The vm's xml has that string there currently set to no and you just change it to yes. Very simple. Edited December 7, 2020 by mdrodge I'm a idot Quote Link to comment
Taddeusz Posted December 7, 2020 Author Share Posted December 7, 2020 7 minutes ago, mdrodge said: Is anyone able to explain this. I don't understand where the clock section is. I'm guessing you are saying edit the vm xml with this? Can someone give an example please. Yes, edit the XML. Quote Link to comment
mdrodge Posted December 7, 2020 Share Posted December 7, 2020 Thanks for the quick response. It's not making a difference for me. Windows is reporting mostly idle but in the uraid tab I can see the first vm core sitting at 70-85% I have 2 vm's. I just tried it on the second because I'm using the first one. (Up until I found this thread i just assumed this was a inevitable part of vm life, like an Emulator overhead) Would be lovely to sort this though as it messes with the boost (I've had to disable boost in bios so I could actually get a real performance "boost" out of this) I'll show you what I'm dealing with. Cores 2-5 (and there multithread) are vm1 and cores 6-7 (+ht38-39) are vm2 I've already changed vm2's hpet to yes. Is there any other suggestions (Is it possible it only works if they both are set that way??)(i could try that but I'd have to find a different terminal first and that involves getting up lol.) Quote Link to comment
Taddeusz Posted December 7, 2020 Author Share Posted December 7, 2020 @mdrodge I've noticed with my Windows 10 VM, with which I pass-through a video card and USB card, that when it's idling it will often times have two of my cores going. As soon as I hit the mouse the CPU activity drops. I'm not sure what's going on there but I'm pretty sure it's some kind of internal Windows process that happens when the system is idle. Quote Link to comment
mdrodge Posted December 7, 2020 Share Posted December 7, 2020 (edited) 20 minutes ago, Taddeusz said: @mdrodge I've noticed with my Windows 10 VM, with which I pass-through a video card and USB card, that when it's idling it will often times have two of my cores going. As soon as I hit the mouse the CPU activity drops. I'm not sure what's going on there but I'm pretty sure it's some kind of internal Windows process that happens when the system is idle. I know windows definitely "ticks" and buildzoid did a YouTube video showing with an oscilloscope that it even caused boost to come on for ryzen cpu's. The effected vm has just fired up runs sonarr and has just fired up a download so it's busy now so that will effect my results. The vm I tried it on has no hardware apart from a harddrive passed through. The other vm has all the passthrough and I've not touched it yet because it is my terminal. Even with or without using the system if I look via my phone I see the same behaviour though so I don't think that's what is going on. It's an overhead of some kind, ot never go's away and the system becomes unresponsive if I tax it because it thinks it can use core 1 but it's already at 80% I tried to having an extra core in the cpu isolation compared to the cpu pinning to see if I could move that load to it and have the vm's unaffected by it but that didn't work. Edited December 7, 2020 by mdrodge Quote Link to comment
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.