virtualbox.org

Recently I tried 2 Mac Host with exactly the same guest ubuntu.

Host 1: MacBook 2017 (12 inch), 1.4GHz Dual-Core Intel Core i7 / 16GB 1866MHz LPDDR3 / 512GB PCIe-based SSD
Host 2: MacBook Pro 2017 (13 inch), 2.5GHz Dual-core Intel Core i7 / 16GB 2133MHz LPDDR3 SDRAM / 1TB PCIe-based SSD

Guest: Ubuntu 16.04, Base Memory 12288MB / Chipset: PIIX3; Enable I/O APIC; Enable HW Clock UTC Time; 2 CPUs; Execution Cap: 100%; Enable PAE/NX; Paravirtualization Interface: KVM; Enable Nested Paging: Video Memory: 128MB; Disable 3D/2D Acceleration; Storage Contoller: SATA; Enable Host I/O Cache.

My expectation: MacBook Pro is supposed to run the guest OS faster, smoother than MacBook. However, the result is opposite - the guest Ubuntu is running MUCH FASTER on MacBook than MacBook Pro.

The way I test the performance is running ROS application by playing a rosbag and launch rviz to visualize the data (if not familiar with ROS, just know the applications is very heavy to both CPU and memory load).

To be more specific, while running the applications, the guest OS on MacBook can still respond to desktop UI event (mouse, key...); regarding MacBook Pro, the guest OS just hangs there without any response; if lucky, sometimes it responds to a mouse click event after 30 seconds or even 1 min...

I also monitored the CPU and memory loading during the tests, and the results might explain why - CPU loading on MacBook is usually higher than 230%, while it remains lower than 200% and mostly < 150% on MacBook Pro; I don't remember the memory usage for now, but I'm sure the VirtualBox application always consuming more memory on MacBook than MacBook Pro.

My hypothesis is that the MacBook launch it's Turbo Boost and makes it always ON while I doing the tests; on the other hand, with some unknown reason, Turbo Boost is OFF when I run the same tests on MacBook Pro.

Is there any way to force my MacBook Pro to turn on Turbo Boost for VirtualBox? Or maybe my guess is wrong?

Please give me some hints. Thanks.

Could you please try again with the guest set only to a single CPU?

We're going to need a complete VBox.log, from a complete VM run, from both Macs:

• Gathering the necessary information:
  1. Start the VM from cold-boot (not from a paused or saved state).
  2. Observe the situation, i.e. run the programs that you run for a couple of minutes, enough to observe the differences?
  3. Shutdown the VM (force close it if you have to).
• With the VM completely shut down (not paused or saved), right-click on the VM in the VirtualBox Manager and select "Show Log".
• Save only the first "VBox.log", ZIP it and attach it to your response. See the "Upload attachment" tab below the reply form.

Remember, try to follow the same gathering evidence technique for both Macs...

@socratis,

Please find the attached logs, thanks for your help!

@Martin,

Thanks for the advice. I just tried, set only 1 CPU did not make difference.

Your logs look almost identical, so your concern is more than valid.

qiao wrote:The way I test the performance is running ROS application by playing a rosbag and launch rviz to visualize the data (if not familiar with ROS, just know the applications is very heavy to both CPU and memory load).

I have some reservations regarding your "benchmark". I would try other application to see what's going on.

Another parameter that you have to take into account is the differences in the resolution of the two Macs. Get out of FullScreen, try the windowed mode with the same resolution.

And finally, I took a look at the Robot Operating System (ROS) and "rviz". The latter seems like it's heavy-duty visualization application, so I would enable the 3D acceleration on both VMs. And of course the same non-FullScreen same windowed resolution.

qiao wrote:My hypothesis is that the MacBook launch it's Turbo Boost and makes it always ON while I doing the tests; on the other hand, with some unknown reason, Turbo Boost is OFF when I run the same tests on MacBook Pro.

Download these two apps from Intel. I have not tried the PowerGadget personally:

• Intel MacCPUID
• Intel Power Gadget

Also, here are a couple of additional monitoring applications, so that you know what's going on your system at any given time. Your preference may vary, but you can use either of them (I use all of them concurrently):

• MenuMeters (for OSX > 10.10).
• X Resource Graph.
• Memory Monitor.
• Built-in Activity monitor.