The first thing I noticed when selecting it in OBS, is that
it has support for XRGB format!
I squealed like a little girl.
Edit: I'm going to look through the whole thing one more time to confirm what I have said was correct, because this is a serious allegation.
Edit2: Alright, I'm going to stop the investigation here, because I have not found concluding evidence that shows bad reviewers are bad. Instead I found possible reasons as to why this was not noticed by the reviewers. So I'm going to dial back the tone a bit.
But bottom line is, the difference is real, and this is what you will see when you see OBS's preview vs Elgato Game Capture HD's preview.
Look at the 2 screenshots below.
The first one is a typical result you would get with a typical capture card capturing in the typical YUY2 video format using a typical capturing program (in this case OBS)
https://i.imgur.com/CWQTyQD.png
The second one is Elgato HD60 S capturing in Game Capture HD
https://i.imgur.com/WlhGxIM.png
Notice anything different between them? Specifically, the red circle which shows which song is selected.
This is the result of chroma subsampling. Googling shows multiple webpages with good explanation so I'm just going to link a random one
here. But to summarize, chroma subsampling has the most obvious effect on the red color, and in the case of YUY2 which is a 4:2:2 format, you lose half the horizontal resolution. (Previously I incorrectly said you would also lose half the vertical resolution, but I was thinking about the 4:2:0 format. Many video capture cards capture in YUY2 / 4:2:2, it is the video files that are usually 4:2:0 (or technically speaking, it is the encoder's doing).)
And how do we know that it is 4:2:2 subsampling? We just zoom in.
Notice in the left circle, although you can still see the individual pixels, the red content seems to be coming from rectangles of 2x1 pixels, showing it is 4:2:2.
Reasons why people may have not noticed the difference
Disclaimer: Note that most of this is unofficial speculation. Because trying to get official support for an open-source software is like trying to get official support from people who are working for free. Just look at OBS's official forum if you want to know what I mean. It is also for this reason I did not bother asking them about Razer Ripsaw's stuttering issue.
First, to address why you will not see the difference in the preview of OBS. The most probable explanation I have, is that OBS simply doesn't work properly with HD60 S's settings. As in, even when I selected 29.97fps, I still got 59.94fps. Clicking video settings shows an empty window where the settings should have been, and nothing shows up when I click crossbar. In earlier version of OBS (I used 18.0 at first), the sound doesn't even work. So you know when people mentioned HD60 S's compatibility issues with OBS? Yup.
The second explanation that I have but is only a guess since I can't prove this second explanation if I can't prove if the HD60 S is working properly because of the first point above: From what I could gather, OBS automatically subsamples when doing streaming, and the preview counts as streaming.
There is still some discrepency in this hypothesis. I remember reading that it would resample to NV12 (4:2:0), but the screenshot shows 4:2:2.
But either way, for whatever reason and in whatever way, the end result is still that with OBS, you will get the subsampled picture in the preview, while with Elgato's software you won't. And you can't use Elgato's software without HD60 S. So end of the day, you still need HD60 S to get better quality preview.
Next, to address why you may not see the difference in the output files. And this difference is the abovementioned concluding evidence that I failed to find. As I mentioned, video codecs usually use 4:2:0, but the resulting videos usually don't see such aliasing. This is because modern video codecs (as in DivX and newer) also have in-loop deblocking, and the typical non-deblocked picture is way more blocky than the effect of subsampling that this kind of defect can be repaired without issue. Furthermore the codec knows that the video is subsampled so it probably has even better algos just to reconstruct the original picture based on the subsampled content. Add that to losses in encoding and recovery in the form of deblocking and sharpening, and we get a result that looks like a circle even though the original source should have been subsampled.
The better reviewers (well, better than me) would use computer text and test patterns to prove whether the video is subsampled, but even then it only increases the chance of detection, due to the same reasons. You would see a blurer text, but is that due to subsampling or encoder losses?
And finally, the best method is to just capture lossless and decode lossless. But OBS does not support external codecs (People saying x264 and GPU-accelerated options are good enough yada-yada). I also tried VirtualDub, which has more settings including subsampling, but it ended up with me not understanding if there is any conversion going on. Maybe I just didn't try hard enough, but it was past 1am and I have a day job.
Screenshots... they are lossless and avoid post-processing, but I can't tell whether they are taken from the preview or the actual data. And OBS doesn't have screenshot feature anyway.
So I'm going to leave this mystery to the pros. But bottom line is, typical card in OBS = subsampled preview, HD60 S in Game Capture HD = nice preview.
