Video Cards - support for OpenGL

[JonB256]

Both of the computers I use have Radeon HD video. While currently (2016.27) they seem to be working, it appears that the nVidia users have better support for the newest OpenGL libraries.

So, is there a big speed difference anyone has noticed with the update to OpenGL ?

If so, is it a big enough difference to warrant a video card update?

and, if so, what level of nVidia would see the best cost/performance ratio?

note, I'm concerned about future updates and stability, not just speed.

[Gilrock]

I really don't think a new video card is going to give you much performance boost if xLights is your only program.  Now if you're a video gamer its another story.  Based on what we just talked about yesterday the resolution of the timer we are using in Windows is doing more to hurt our performance than anything.

[JonB256]

The last game I played was a Zombie game and it creeped me out too much to play. Not a gamer.  I'll sit tight for now until New PC time.

[Phrog30]

I really don't think a new video card is going to give you much performance boost if xLights is your only program.  Now if you're a video gamer its another story.  Based on what we just talked about yesterday the resolution of the timer we are using in Windows is doing more to hurt our performance than anything.

Gil, if you have time can you explain a little more about the timer you are referring to? Just curious. Thanks.

James

Sent from my SM-G900V using Tapatalk

[Gilrock]

When you hit play in xLights the rate of the frames is controlled by a wxTimer object being fired on a regular basis.  We attempt to tell it to fire at the rate you define you're sequence at so for a 50ms sequence you should see 20 frames per second.  Keith just added code to measure the actual FPS you are getting and we noticed we are only getting about 16.8 FPS instead of 20 and it's due to that timer.  We are really getting about 60ms for a 50ms sequence and 30ms for a 25ms sequence.  A 100ms sequence looks pretty close.  This time division also control the speed at which we would send data out the network if you were using xLights to play the sequence.  So although all this work to improve OpenGL is probably going to benefit us somewhere down the road it probably didn't fix some of the immediate things we thought it would like the folks that were having issues with strobes not firing at the proper rate.

Now everything I just talked about is in regards to Windows.  You move over to a Mac and it's a completely different story.  Over there it seems like you can actually get 40 FPS on a 25ms sequence if Dan comments out the line that displays the time and FPS.  When he added back in that one line of code to display those two labels it lowered his FPS to 30 FPS.  We aren't seeing the same effect in Windows so the two OS's are really very different animals.

[Phrog30]

When you hit play in xLights the rate of the frames is controlled by a wxTimer object being fired on a regular basis.  We attempt to tell it to fire at the rate you define you're sequence at so for a 50ms sequence you should see 20 frames per second.  Keith just added code to measure the actual FPS you are getting and we noticed we are only getting about 16.8 FPS instead of 20 and it's due to that timer.  We are really getting about 60ms for a 50ms sequence and 30ms for a 25ms sequence.  A 100ms sequence looks pretty close.  This time division also control the speed at which we would send data out the network if you were using xLights to play the sequence.  So although all this work to improve OpenGL is probably going to benefit us somewhere down the road it probably didn't fix some of the immediate things we thought it would like the folks that were having issues with strobes not firing at the proper rate.

Now everything I just talked about is in regards to Windows.  You move over to a Mac and it's a completely different story.  Over there it seems like you can actually get 40 FPS on a 25ms sequence if Dan comments out the line that displays the time and FPS.  When he added back in that one line of code to display those two labels it lowered his FPS to 30 FPS.  We aren't seeing the same effect in Windows so the two OS's are really very different animals.

Thanks Gil, great explanation. Can I assume that if we export/play via fpp the timer issues are irrelevant?

James

Sent from my SM-G900V using Tapatalk

[Gilrock]

Yeah that's why we've always pushed using an FPP because we knew Windows is not very accurate for real time operation.

[Gilrock]

But I also have to believe we might be able to make it better in Windows now that the issue is more obvious.  Otherwise the variations are most likely affecting the synchronization of how you place effects with regards to sync'ing with the audio.