TMPGEnc - bad frame problem

[psyklax]

Here's a frame I've captured from an MPEG2 video I encoded with TMPGEnc, for a DVD. Anyone else seen something like this? I've had the problem quite a bit, there must be a reason for it. It's very frustrating, especially when the video takes hours to encode and once or twice this happens.



EDIT: This has just happened again with a video from a completely different app, so it's not a TMPGEnc problem. Is it my video card? My hard drive? Processor? RAM? I just don't know. This is very puzzling and extremely frustrating.

[EddyH]

without any further info or investigation, i'd have a look at maybe the mpeg matrix setting... the only time i've had bizarre picture twisting with TMPG was because of that, though it didn't manifest until I used it on a standalone player...

that, or, if you're transcoding from XviD, it might be some kind of decoder problem. for me, Nero Vision 4 seems to really hate XviD, giving constant smearing corruption, as if the video had been downloaded on a very broken 56k connection...

(now i'm just trying to figure out which movie that is

*some cheating later*
1/ league of extraordinary gentlemen?
2/ one field is good; the other appears to be very corrupted in a very strange way, but some slight evidence of the original image is present (unless of course it's the further compression of the screenshot bleeding through), the pattern suggesting some kind of P or B frame cockup...

[manono]

Hi-

Is it my video card? My hard drive? Processor? RAM? I just don't know. This is very puzzling and extremely frustrating.

I had random corruption similar to that for a few frames every movie encode when I had a bad stick of RAM. You might test the memory using Memtest86:

http://en.wikipedia.org/wiki/Memtest86

[EddyH]

seconded, MT86 is a simple but awesomely useful tool. can even use it to narrow down other problems if it's not memory (e.g. locating to the motherboard/cpu if it errors persist across all chips you test, or to disc/etc if nothing bad happens at all inside MT (which really punishes the RAM) but consistently in normal PC use. <<<--- bitter experience

(hunt out the latest version of "MT86+" for preference, as the vanilla utility stopped development some years ago)

[psyklax]

God damn it... I ran memtest86+ for three hours, got 10 passes, no errors. Strangely enough, that wasn't what I wanted. At least if I had errors I'd know what the problem was.

Any tips on how to refine the testing? Stick with 1 particular test for example? Surely if both of you have had this problem thanks to bad RAM it must be the RAM (I think I have cheap RAM after all). Then again, like you said, maybe it's just leading me somewhere else... bah...

[manono]

Hi-

In my case, the errors popped up almost immediately when testing the RAM, and I quickly piled up hundreds of errors. Most of the screen turned red. I think that after 3 hours and 10 passes you're safe. But also back to square 1.

In a similar thread I saw recently, it turned out the problem was caused by the encoder (Procoder in his case) and switching to a different encoder (CCE) solved the problem. You said you got the same corruption from a different source besides TMPGEnc. I'm a little unclear about what that means, but you might try a different MPEG-2 encoder (HCEnc, maybe) to see if the results are any better.

The thing is, when I was trying to diagnose my problem, before I got the bright idea of testing the RAM, I'd get this corruption only a very few times every movie, and I'd have to watch the movie all the way through each time I tested out something else. Not much fun. So I sympathize.

But I did learn from the experience that Memtest86 is a great tool for testing the RAM if there ever are any problems, and also for testing the memory every time I tweak my overclocks. It and Orthos should be used by every overclocker. Are you overclocked, by chance? If so, maybe turn down the OC and try again.