View Single Post
Old 26th December 2012, 09:35   #16589  |  Link
madshi
Registered Developer
 
Join Date: Sep 2006
Posts: 9,140
Quote:
Originally Posted by 6233638 View Post
Unrelated to that, the optimizations with Jinc have made a big difference. With 720p or lower videos, I can use Jinc 3 AR for Chroma now, and Jinc 3 without the AR filter for 1080p content.
Glad to hear that.

Quote:
Originally Posted by Mangix View Post
I tried upping the settings to the point where it would start dropping frames on this GPU and with the "use a separate device for presentation" option enabled, i got lower rendering times but more frame drops(around 5 frames a second) whereas with it disabled, i got higher rendering times but less frame drops(3-4). Something is up.
Interesting. This is what I meant when I said that trusting rendering times alone isn't always telling the whole truth. However, 6233638 also has better real life performance with the option enabled compared to having it disabled. So it once again seems like the option behaves differently for different setups. Anyway, I'll disable it by default in the next build.

Quote:
Originally Posted by Mangix View Post
edit: i also have a question about the refresh rate switcher. when it switches to a custom resolution of eg. 48Hz or 50Hz, everything is fine. But when it switches back to 60Hz, it switches not to the normal 60Hz resolution but some special HDTV one.
I think this was reported by other Windows 8 users before. It doesn't seem to occur on Windows 7. You could create a debug log for me to check if I can see anything wrong. But IIRC last time I checked, madVR behaved correctly. I think it's probably a driver problem. I think that either the driver is telling madVR that the original mode was 59Hz. Or alternatively, when madVR asks the OS/driver to switch to 60Hz, the driver switches to 59Hz instead. One of those 2 things, I think.

Quote:
Originally Posted by ryrynz View Post
Madshi, thoughts on a general GPU performance impact bar graph showing along with the rest for each of the algorithms?
That's not as easy as it may sound. Performance depends on so many factors. E.g. I've found that on my PC Jinc scaling is relatively fast with very small scaling factors, but gets exponentionally slower with bigger scaling factors. Also the anti-ringing filter is so much slower on older generation GPUs than it is on newer generation GPUs. Finally, with v0.85.5 I just introduced special shader code for 2x and 3x scaling factors, which is way faster than with other scaling factors. How could I possibly show such effects with one simple graph?
madshi is offline   Reply With Quote