4.3.0 massive DPC latency

[Jonas H]

Hi again,

With 4.3.2 I had problems listening to Spotify in my host while the VM (with GA 4.3.3) was running, but this seems all gone now with 4.3.3. No problem when starting, using, pausing, resuming and then finally exiting a VM.

Brgds

Jonas

[Allen_ASU]

And here is 4.1.18.

Confused here... I'm guessing the comments are wrong but the names of the log files correctly indicate the VBox version, i.e. 4.3.2 (when you say 4.2.18) and 4.2.18 (when you say 4.1.18)?

You're correct. My derp. Sorry about that.

[Allen_ASU]

Okay, in the logs from Jonas H there's clearly a big improvement between 4.3.2 and the 4.3.3 test build. Both the maximum values and the average latencies are much lower.

With the logs from Allen_ASU, I see an improvement in the averages but not the maximums. I can't really tell which is "better".

Now here's the problem. On my machine, I can get a maximum latency of 8.5ms and a report that "your system appears to be having trouble handling real-time audio and other tasks" just by building VirtualBox, while VirtualBox itself isn't even running. Similarly on the first try of starting VirtualBox, I got a latency peak of 2.8ms, before any VM was started. When I started a Windows 7 VM with 2 VCPUs, the latency spiked at 4.5ms, which is bad but not nearly as bad as the latency incurred when building source code. The averages OTOH stayed low all the time.

The problem with the latency checking tools is that they don't (or can't) say what is responsible for the high latencies. When I was running VBox, I can't even say for sure if the DPC latency spike was caused by VBox or by whatever caused an even higher spike when building code.

I'll add that Allen_ASU's machine has much worse average latencies than Jonas H's despite being (admittedly an older/slower) Xeon.


So the real question is, does the 4.2.3 test build still cause music etc. to drop out where 4.2 did not?

4.3.3 still skips music, just not nearly as bad as 4.3.2.

The averages between all 3 versions (4.3.3, 4.3.2, 4.2.18) are all about the same. Approximately 50-80ms. The highest interrupts is where its interesting. Looking at the boot only logs, 4.2.18 shows the best by far at a peak around 2500ms. The 4.3.x logs are orders of magnatides higher, between 18,000-39,000ms. In just general usage, 4.3.3 is better than 4.3.2 and I think that is reflected in the DPC count. 4.3.2 shows higher interrupt counts than 4.3.3 which may indicate that 4.3.3 still has some issues with DPC latency, it just doesn't occur as often.

4.2.18 interrupt counts are much lower, spiked DPC latency are much lower, which results in a generally happy host system.

[jorgensen]

Here is my result with identical XP guest setups and updated Guest Additions.
Tested on the same computer using multiboot.
Version 4.3.3 far better than 4.3.2 which was in the red most of the time.
All WM starting point around 35 sec.

Windows XP host, Virtualbox 4.2.14 and XP guest.

Lantency 4.2.14.png (12.89 KiB) Viewed 1202 times

Windows 7 64-bit host, Virtualbox 4.3.3 and XP guest.

XP Lantency 4.3.3.png (2.72 KiB) Viewed 1202 times

Windows 7 64-bit host, Virtualbox 4.3.3 and SharePoint guest.

SP Lantency 4.3.3.png (2.8 KiB) Viewed 1202 times

[AndrewG]

What is the consensus on 4.3.3? Will it behave properly in the background with audio/video going on the host?

[Jacob Klein]

I was having horrible problems when running multiple VMs in BOINC, using 4.3.2. It seemed if I just ran 1, my system would only have very minimal stuttering, but if I ran 2 or more, my host mouse cursor's acceleration would make the cursor skip around uncontrollably, the audio would sometimes skip badly, and sometimes controls would paint/refresh slowly. Also, I believed I had narrowed the issue down to occuring only when the VM had: "Enable VT-X/AMD-V" checked.

Today, I found this thread, and I installed the 4.3.3 test version within the thread. It has improved things tremendously! I know we can't really "trust" Latency Monitor, since I'm using Windows 8.1 x64, but... Latency Monitor now shows the latency bouncing between a sustained 950 microsecond latency and a sustained 1975 microsecond latency, while running 3 VMs that are each doing complex single-CPU computation alongside 5 other non-VM tasks and 3 other GPU tasks.... whereas before (with v4.3.2), it was spiking and averaging to around 8000 microseconds.

I can tell you for SURE that the music no-longer skips, and the mouse cursor's acceleration is now fluid and accurate.
Kudos for making great headway into solving this problem. Looking forward to the 4.3.3 official release!

- Jacob Klein

[michaln]

I can tell you for SURE that the music no-longer skips, and the mouse cursor's acceleration is now fluid and accurate.

Thanks for the report! BOINC is probably an excellent test case, since AFAIK it does a lot of computations inside the guest.

Out of curiosity, what kind of mouse do you have? We've never seen any issues with that, but I'm sure it's hardware/driver dependent.

[Jacob Klein]

I have a Logitech G9 Laser Mouse, running the latest Logitech Gaming Software, v8.50.281 x64, with cursor acceleration enabled. The behavior I was noticing was that, while the cursor was in motion, the DPC spikes would hit, and the host OS would not properly account for the cursor acceleration. The result was that the cursor would skip either too far or too short in the direction it was traveling. My primary test case, to convince myself of what I was seeing, was to draw medium-sized rapid circles with the mouse cursor. When the DPC spikes were hitting, the center of the circle would shift, sometimes even more often than once per second. It was nasty and made the PC barely usable.

Regarding BOINC as a test case, I think you're right. There are only a handful of projects that support VMs at the moment, but the "big test" on my machine was to get 1 RNA World VM running, 1 Test4Theory VM running, and 1 Climate@Home VM running...all at the same time. So, all 3 VMs (each with VT-x enabled) were fully-utilizing a CPU each, alongside 4 other non-VM CPU tasks and 3 GPU tasks. I'm happy to report that I can do all that now, with the v4.3.3 VirtualBox Test version, without a problem.

[Allen_ASU]

I tired this build and things are looking much better in latency monitor. Haven't had any issues with audio on the host.

Thanks for looking into this.

[AndrewG]

I just installed the newest version, 4.3.4. Subjectively it seems fine. I haven't heard any stuttering during my normal use.