Running some tests in my VM with an application we are developing, the 6.1.22 version is about 50% slower than the 6.1.18 now with 6.1.24 its about 10% slower than 6.1.18. I've switched between them twice now and get the same results.
I've attached a zip of the log files. The VM guest has 8G from a 32G host, has 2 cpus and has VT-x, Nested Paging, PAE/NX, and Hyper-V paravirtualization.
The log zip file includes both the 6.1.22 log and the 6.1.18 log for comparison. Nothing was changed in the vdi and it is using the current guest extensions.
Virtual Box 6.1.24 slower than 6.1.18
Re: Virtual Box 6.1.22 slower than 6.1.18
Not sure if this will help. The slowness does not seem to affect all aspects of the guest. Boot is fast. But the unit tests for a product we are developing are slower on 6.1.22 than 6.1.18. The tests are mostly CPU bound. Almost no disk.
-
- Volunteer
- Posts: 5677
- Joined: 14. Feb 2019, 03:06
- Primary OS: Mac OS X other
- VBox Version: PUEL
- Guest OSses: Linux, Windows 10, ...
- Location: Germany
Re: Virtual Box 6.1.22 slower than 6.1.18
I didn't see anything obvious when comparing the log files given. But I can suggest two things:
1. If you do not need nested virtualization, you should disable it in the VM configuration and see if it makes any difference.
2. If you'd provide new log files from "better comparable" VM runs, I'd compare the statistics in the log files.
2. (a) Start the VM and let it run idle for approximately 3 minutes. Then shut it down from within the guest OS.
2. (b) Start the VM and let it run idle for approximately 3 minutes. Then execute the unit tests for approximately 30 minutes, and shut the VM down from within the guest OS.
2. (c|d) Repeat (a) and (b) with the other VirtualBox version. It is important to run the unit tests for the same amount of time, not to run the same amount of unit tests.
1. If you do not need nested virtualization, you should disable it in the VM configuration and see if it makes any difference.
2. If you'd provide new log files from "better comparable" VM runs, I'd compare the statistics in the log files.
2. (a) Start the VM and let it run idle for approximately 3 minutes. Then shut it down from within the guest OS.
2. (b) Start the VM and let it run idle for approximately 3 minutes. Then execute the unit tests for approximately 30 minutes, and shut the VM down from within the guest OS.
2. (c|d) Repeat (a) and (b) with the other VirtualBox version. It is important to run the unit tests for the same amount of time, not to run the same amount of unit tests.
Re: Virtual Box 6.1.22 slower than 6.1.18
I saw that 6.1.24 came out so I retried this and for my workload and configuration it is still slower than 6.1.18, though not as much as before (6.1.22)
I've attached logs from running the test load on both 6.1.24 and 6.1.18. Any ideas or suggestions are appreciated.
Herb
I've attached logs from running the test load on both 6.1.24 and 6.1.18. Any ideas or suggestions are appreciated.
Herb
- Attachments
-
- VBox.6.1.24 logs.zip
- (184.26 KiB) Downloaded 7 times
-
- VBox 6.1.18 logs.zip
- (82.57 KiB) Downloaded 6 times
-
- Volunteer
- Posts: 5677
- Joined: 14. Feb 2019, 03:06
- Primary OS: Mac OS X other
- VBox Version: PUEL
- Guest OSses: Linux, Windows 10, ...
- Location: Germany
Re: Virtual Box 6.1.24 slower than 6.1.18
Regarding the VM statistics, I see a similar difference as in VirtualBox vs Hyper-V AMD/Intel big performance difference, showing that vCPU0 nowadays counts its idle time in the cNsOther state instead of the cNsHalted state in earlier VirtualBox versions. But I cannot draw any connections to a performance drop.