Hacker News Reader

Comment by rasz_pl | Hacker News Reader

As a nerd I love numbers, 'Latency is minimal' doesnt do it for me. Display this http://tft.vanity.dk/inputlag.html or any other stopwatch, and record both screens. What is the bitrate and cpu overhead at hd/fullhd resolutions 30/60hz?

Some very slow MVA monitors lag as much as full frame behind input, so in extreme cases you could get equal delay on your laptop.

View on HN

For my system it hovers around 50-70ms[1][2]. Tested with an 800x600 window at ~7% cpu utilization on a Core i5. My desktop monitor is a Dell u2711, which seems to add about 15ms latency itself[3].

I'm not sure how much latency OSX or Chrome adds (at least a frame more than Firefox[4]); Mobile Safari seems to be a bit faster, as evident in the video. I don't have a second Windows machine for comparison.

Edit: When connecting on the same machine, latency is 2 frames (33ms) exactly[5].

[1] http://phoboslab.org/files/jsmpeg/jsmpeg-vnc-latency.mp4

[2] http://phoboslab.org/files/jsmpeg/jsmpeg-vnc-latency.jpg

[3] http://www.anandtech.com/show/2922/4

[4] http://phoboslab.org/log/2012/06/measuring-input-lag-in-brow...

[5] http://phoboslab.org/up/B2u6w.png

> 'Latency is minimal'

Tell me about it, I've been trying to gather info about a solution where I can record with a camera, probably 1080p (via ethernet or via hdmi to capture card) and use the frame information as inputs in a simulation with a latency around 20 ms and am unable to get hard data on the actual latencies. Nobody cares about input lag because practically no one records as a real time input.

Incidentally, if anyone knows or has experienced with a similar setup I'd be forever grateful.

Here's a dumb idea I just came up with: Attach a bright light that the computer can blink programatically. Write a program which: a) blinks the light, b) Blinks the light again 0.01 seconds after the camera sees the light blink, and c) repeat b a thousand times. d) Measure how long the process took relative to ten seconds, divide by 1000 to get your latency (including latency for sending the signal to the light, but hey, it's at least an upper bound).

Another thought that occurs to me is that people who work with music equipment (especially usb synths and such) care about latency a lot. It maybe worthwhile to hook into that scene and see what kind of insights can be gleaned. Here's a link to get you started: http://www.mathworks.com/help/dsp/examples/measuring-audio-l... Looks like they're basically using a feedback model as well.

Thanks. I've done stuff like that. For "whole system" lag it's also a good idea to record input lag with a faster camera and analyze the count the number of images/frames that it takes from the input to the output on screen.

My problem is that, from what I've seen so far, camera/capture card retailers sometimes don't and there doesn't seem to be a particular standard for those who do.

Thanks for the link. It's quite interesting.

Gamer care about input lag. Pro gamer on PC don't use wireless input devices, they prefer cable based ones for exactly that reason. The input lag is also what (afaik) almost killed Gaikai. Sony recently bought it for some older PS3 titles. https://en.wikipedia.org/wiki/Gaikai

I never played at the pro level but I did play competitively and I would never use wireless internet, mice, or keyboards. These days there are some great mice, keyboards, and headphones but most people I know won't leave it up to chance and always go with wires.

Yes. I'm aware they do and manufacturers of those devices often provide options and information but cameras are seldom used as input devices so they're not usually measured and that's a shame.

http://superuser.com/questions/419070/transatlantic-ping-fas...

would like to see latency numbers too. Gstreamer with h264 is about 150ms ish

Reply on news.ycombinator.com