Спасибо

I checked some of the other ressources that you mentioned (btw thanks, I really wished I had known about these before I decided to buy DXtory). However it seems that there is no such thing as a perfect screen recording software. All of them seem to suck in one way or another. Fortunately, they all seem to suck in complementary ways.

Shadowplay seemed very attractive, until I found out that it cannot be installed separately, but only with Nvidia Experience, which first of all seems not to be compatible with my GT 220 and secondly seems to have the nasty habit of trying to automatically update your graphics card driver. This can be fatal for trying to run old games, a lesson I had to learn the hard way when after an update of my driver I couldn't play Take No Prisoners anymore.

OBS (open source and in active development !!!!) seemed fantastic, however it only runs on Windows >= 7 because of DirectX 10, so that's no option for me.

Raptr apparently requires you to set up an account for them, and the client displays ads. No thank you.

Just downloaded an tried Afterburner. It did impact performance (I use it with MagicYUV compression) noticeably but not dramatically during conversions, or to put it another way, it is most definitely not as horrible as Fraps but DXtory still seems to perform better for this particular game. Edit: It appears the performance drop was actually due to MagicYUV. When recording using lagarith, Afterburner seems to perform as good/bad as DXtory.

DXtory gave me some unexpected troubles recording Fallout. The first three videos recorded perfectly, that is, the end result after rendering and compression was perfect (no av-sync issue or anything). But for some reason, for the fourth video the video is playing back at maybe twice the normal rate. This is even more curious, since I recorded everything in one session, i.e. there is only one source video. Maybe the problem lies in rendering (which I do with Blender) and the interaction of ffmpeg and the video codec (lagarith). However in Blender the video plays back perfectly normal. So I really don't know which part of the chain is responsible here. But my guess is, the implementation of the lagarith decoder in ffmpeg is faulty (as is the decodec for Fraps btw). Edit: As it turns out, this problem has nothing to do with DXtory. The same problem appeared with Afterburner too. I re-recorded everything from the fourth video onwards, and again only the fourth (i.e. first) video suffered from this strange video-playing-at-twice-the-rate-effect. Comparing files I noticed that the .blender file for the fourth video was 3 kB smaller than all the other ones. Apparently it got corrupted when blender crashed, but blender did not recognize this corruption and rendering this broken file had this unexpected effect.