Quick Sync performance on HD3000

_Rick_

Diamond Member
Apr 20, 2012
3,952
70
91
I'm not sure where to post this, but since it's more or less related to the graphics unit, I'm posting here.

I've recently started to take a look at MSI Afterburner's video recording function, which now has support for QuickSync based encoding.

I set up my computer using the VGA-forced output to force activation on the iGPU, and after a reboot apparently QuickSync works.
But, at 2560x1600 I get a mere 10-11 fps at the lowest quality setting, and when using software encoding with the same codec (selecting "disabled" instead of the DX9 interface) I get 28 fps (with someone increased CPU usage).

Now, I suppose I wasn't expecting miracles, but I think I was expecting QuickSync to perform much, much faster than just using the CPU. This is what all the benchmarks suggest. Now, what might cause this, is the relatively big resolution, which may exceed what QS was designed to work on efficiently.

My request therefore is for someone with HD3000 to test their QS H264 encoding speed on uncompressed 2560x1600 (or 2560x1440 - the difference should be minor) and compare it with CPU encoding. Afterburner is one way to do that (set up QS-supported recording and use the built-in benchmark), but maybe there are other free encoders as well.

I suppose half-framing will give me usable FPS for recording games, but I'm mostly surprised by the abysmal performance, and whether there isn't a way to get more out of it, or whether that is all that QS is capable of.

Thanks!
 

videogames101

Diamond Member
Aug 24, 2005
6,783
27
91
Generally when working with Quick Sync (or any realtime encoding engine) you provide 3 of the following 4 parameters: target bitrate, resolution, framerate, and quality. The other is beyond your control. Quick Sync (or any realtime encoding engine) will drop the last parameter to the point it that it can meet your 3 targets.

Generally, resolution, framerate, and bitrate are specified. Leaving quality to be varied by the Quick Sync encoder. It will choose the highest quality encoding setting that can still meet your 3 specified parameters.

I'm not sure how MSI's implementation works, but from what it sounds like it's specifying resolution, quality, and bitrate. Which means Quick Sync can only alter the output framerate. It will lower framerate until it can meet your 3 specified parameters.

That's a really bad encoder implementation by MSI. Framerate should pretty much always be specified.

I would suggest trying to use OBS to do local recordings using quick sync. It has a proper implementation where framerate is one of your 3 fixed parameters. Although it's meant for streaming, the local encoding option works very well. You shouldn't have a hard time setting it up if you managed to figure out the headless display hoops Quick Sync has you jump through.

Also updating your iGPU drivers can often increase QS performance.
 
Last edited:

_Rick_

Diamond Member
Apr 20, 2012
3,952
70
91
There is a frame rate setting too, but of course that's only the output frame rate, not the encoding frame rate. The lowest quality setting will result in dropped frames at 30fps at 2560x1600.

Having said that, another test run just now gave me similar performance and CPU usage under both QS and software encoding. Yet it reports back each time that either was used to encode. This only makes it more confusing, as now I really can't tell one from the other, performance-wise.