Let's set this straight - No one can do 24p consistently well - Page 35

Anandtech did some testing and I believe that the ATI 6450 is the most accurate GPU they had tested as of a year or so ago.

The internal clock in the GPU is as much of a contributor to problems or the lack thereof as drivers, application software, etc. I actually have had zero problems with my 6450, my new Onkyo 818 on the other hand.....

The article is a good piece and explains a lot about issues with clock drift, etc, on PC video cards. I would recommend anyone who is having issues read it first as a good starting point for understanding clocking;

http://www.anandtech.com/show/4380/discrete-htpc-gpus-shootout/5

Agreed. Ganesh did a great job.

Good to know. I was considering buying some kind of passively cooled ATI card or new trinity APU hoping they offered similar a/v sync accuracy. Too bad it's not universal across all their new GPU's. Guess I got lucky with this crappy Dell OEM ATI card - they go for about $60 on ebay.

The 7570 I tested is a relabeled 6570 so maybe it offers similar clock performance to the 6450.

What people need to consider, and this may not be obvious from reviews, is that absolute 23.9760 perfection is not really the goal.

First, the refresh rate reported in programs like MadVR, MPC-HC, etc. is only as accurate as the PC timer used to time it. Try disconnecting your PC from external internet time synchronization and see how far it drifts in just a few days. So just because you see 23.9760 reported on your screen does not mean that's actually what you're getting.

Second, the sampling rate of your audio card may be slightly off from the ideal 48000 samples/sec as well. So it's entirely possible that 23.973 could give you fewer drops than 23.9760. This is why I only look at audio/video clock drift over time as opposed to absolute video refresh rate.

To prove the point, my Nvidia cards with custom resolution tweaking would report near perfect 59.9400 out to many more decimal spaces than the ATI card which reported 59.9388. Yet in practice, the ATI card a/v clocks stayed in sync much longer than the Nvidia on 59.940 content - probably because their HDMI audio clock was also slightly slower to better match 59.9388 video refresh.

I wonder what the "enforce smooth video playback" option in the ATI control panel does? The help just says "Enable this to ensure smooth video playback with no dropped frames". Why would anyone turn this off unless if has some side effects?

I know this is probably a REALLY dumb question but.....how exactly can you tell if you are NOT getting "True" 24p playback? Will it be mainly judder? On my HTPC I am running an ATI Radeon 5870 and locked in at 24p. I seem to never have any issues with losing synch either. I'm NOT running reclock (stopped a LONG time ago) and not running any additional add-on filters, just MCE player (with the occasional Total Media and WMP).

Well, as the extreme enthusiasts we are, we NEED to know that we are milking EVERY single ounce of performance out of our HTPC's. It just HAS to be the best that it can be:)

If you don't know if you are getting 24P and have been satisfied with the performance I don't know why on earth you would start screwing with it. Many of us run 24P on our HTPCs because we are already accustomed to the way 24P film content looks on our displays, having used for example blu-ray players in the past and set them to 24P out.

I've never seen a clear explanation for smooth motion, I expect that it does frame sampling which tries to hide judder and frame drops. I turn it off, just like all of the other video processing garbage in the ATI control panel.

Well, as I run a 400 Disc Blu Ray changer set to 24p as well, the results compared to 24p output on my HTPC are very similar in terms of motion. So, I guess I'll assume I've done something (even if by mistake) right.

You would see the frame drops, it's pretty obvious if you have a problem

IIRC when it works properly (some GPUs have had issues with this setting in the past) it allows the drivers to slave to the audio clock by disabling features like VP and decoding every frame making it possible to "watch" videos that would otherwise be a hot mess on underpowered HW.

Does anyone know which would be the "lesser" evil?
Option 1) Core i5-2500k with HD3000 running @ 23hz (imperfect 24p); or,
Option 2) Core i5-2500k with HD3000 running @ 60hz (3:2 pulldown)

My projector (Sony vw95es) also has Motionflow that I do like set at LOW. Would that make option1 or 2 any more advantageous? I'm using base WMP and don't appear to have any lipsync issues with either option, FWIW.

I prefer 1. You will get occasional frame drops instead of always on judder.

Thanks babgvant. I've been watching movies with option1 WITH motionflow set to low. I always wondered- does using frame interpolation in this capacity (23hz/imperfect) hide the dropped frame? I do notice the typical artifacts one might see with FI in this mode (as one would see in any other mode), but was never sure how it would affect (positively or negatively) on the dropped frame situation.

Why not try both and report back...

Has anybody tried the latest Nvidia drivers - 306.23? I just set up a GTX 670 and encountered the 24p bug with XBMC, and I'm curious as to how to fix this with the latest drivers... Or is it the same as the fix in the OP's first post - set it to 23.972?

The latest drivers work just fine, and improved video-related behaviour quite a bit in some parts (especially with madVR), but i don't think anything on the timings changed - just stick to custom resolutions if it bugs you.

I would guess that it would help mask issues with frames dropping because the device is creating "fake" frames to smooth out the display of film content at a much higher actual rate.

Ok, so I have been reading through various posts on this thread about getting the settings right.

I have a HTPC (mpc-hc, lav, madvr) - with an Nvidia GTS 450 - and a Panasonic AE-4000U Projector.

I get a dropped frame every 10-30mins and I only notice it every now and then... but I notice it.

When I set my refresh rate to 23Hz, I get 23.977 in mpc-hc - when I try to tweak it with a custom refresh - it will not accept 23.976 as a value when I 'test'. I have to set it to 23.975 - and when I do that I get 23.974 in MPC-HC.

I find dupe frames bother me more than dropped frames - so i have been sticking with 23.974.

Is there a way to tweak this closer to 23.976?

I'm not familiar with that particular GPU but depending on how accurate it is, you might be close to as accurate as possible.

Yeah, I got 23.97583 with my GT430 and called it a day. With that setting I usually geat either drops or repeats on the order of 15 to 20 hours. Withe the exception of initial loading my playback is very fluid.

+1

Sammy - I am getting 1 dropped frame every 8.6 mins - so roughly 6-8 per hour. How can you be at almost the same refresh rate and be showing 15-20 HOURS between drops? Or did I mis-read you - and you are getting 15-20 dropped frames per hour of viewing?

Also - I have turned off all 'frame repeat' options on my projector since that can cause 'soap opera' effect. Would turning frame repeat back on fix the dropped frame glitch that I am noticing?

Also - would the 'soap opera' effect of frame repeat be almost non-existent because my refresh rate is so close to the real rate?

Well I'm using a 120Hz LCD and not a projector. Don't know if that has anything to do with it. Some files are better than others too I've noticed. I've seen dropped frames down to one every 20 or 30 minutes too but most of the time it is hours.

Ok, so apparently I am not going to get the video card any closer than what it is now.

So I guess I need to focus on if there is something that can be done on the projector end.

There are two options I see in the manual for the panasonic ae4000u:
Frame creation (with 3 modes)
Frame response (with normal and fast options)

Which option could possibly help with my issue WITHOUT causing the 'soap opera' affect that I am trying to avoid?

@lordx2 - what are you using to measure the frame drop?, is it mpc-hc. I'm not sure how accurate these stats actually are - but I think sammy may be using madVR to generate his. I know I do and I see similar results to sammy - though perhaps more in the region of 4-5 hours reported.

If you don't already and can, I would be inclined to try madVR (ctr-j on playback will provide stats details) and LAV Filters, certainly for me they proved the better combination. So much so that I moved to JRiver for my playback needs (they preconfigure madVR and LAV as their playback engine) - though it isn't free.

Keyboard input "CNTRL J" brings up the stats screen that is different when using LAV / MadVR than when using EVR (CP).

The 4 - 5 hour range is pretty typical, the 90th percentile or whatever. The outliers are the 20 hours and 20 minutes that happen with some files so it must have to do with the encoding as well. Try different files and see what you get.

SOLVED:

Sammy - found the solution - and it was so easy - I got 23.97591 with 2 clicks.

Funny enough it is on a thread that you were part of:

http://www.avsforum.com/t/1417683/madvr-intel-igp-and-ram-speed/60

Page 4 - post # 71 and #77 - thanks to Renethx