Jump to content

Large bandwidth spikes graphed when CPU and/or GPU under heavy load


Recommended Posts

uTorrent 3.2.3 build 28705

I'm getting large bandwidth spikes in the speed graph when the cpu/gpu are under heavy load -- far above both the limits in place and my internet connection's limits.

Usually combination of heavy load and user interaction with uTorrent window.

See these graphs.

The first graph is 1 second interval, the second is 5 second interval and the third is 5 minute interval.

1 second interval Shows the latest instantaneous spike under load and user interaction with the uTorrent window. The spike's 1-second average indicates a download rate up to 976kB/s | 7.6 mb/s (limit set at 200kB/s) and an upload rate 1-second average of about 200kB/s | 1.6 mb/s (limit set at 47 kB/s)


5 second interval Shows the pattern goes back the same way for 5 minutes (when my user interaction began). The exact degree of the spikes varies, but effects download and upload to a roughly equally proportional degree.

Also note the massive spike about 7 minutes and 15 seconds ago (the first spike on the graph). It indicates a momentary download rate of (1,453 kB/s)*5s, or 7,265 kB/s second-moment-average. We could assume 200kB/s for 4 of those seconds, assuming this is a single spike, giving us a more accurate 6,465 kB/s spike's-second-average (50.5 mb/s!!).


5 minute interval Shows 3 other older events (or groups of events) with somewhat different patterns.

Ignore the drop in rate 1.5 hrs ago (other demand on network resources), but notice the up-to-several-minute sequence of events ending 3 hours ago.

The download rate surges to about (260 kB/s)*5m{300s} -- as high as 78,000 kB/s second-moment average -- or assuming this was a single momentary spike with normal download rate the rest of the period, the spike would have been 18,200 kB/s-average (142.2 mb/s!!). This is followed by a similar drop in bandwidth spike shortly thereafter that appears to roughly cancel out the first spike. This is different since none of the other spikes have this balancing action that we see here.

I could kind of understand the spikes under load if they all were balanced out like this one... this one of course may have a different cause.


Before you ask, some of my preferences...



My question is this:

A. Is the graph reflecting something that actually happened (even if data dumped to a buffer) -- is the graph reflective of a bug in tracking of bandwidth in certain conditions (load, seemingly gui-manipulation bound) -- or does it mean yet something else?

B. Should I report this as a bug?



Link to comment
Share on other sites


This topic is now archived and is closed to further replies.

  • Create New...