Hi all,

Recently, while watching an MPG2 I converted using TMPGEnc, I noticed a funny glitch that apeared at a single frame. It looked like a row of horizontal white lines, but only on one frame. When I checked the original AVI-file, the glitch was not there. I compressed the movie using TMPGEnc 2.57, W98SE, 5.5Mbps CBR, 720x576 PAL, floating point DCT. Then I decided to repeat the conversion using W2K, but same version of TMPGEnc, exactly the same settings, and the same input. This time the glitch was gone, perfect output. And the binary was different... exactly the same length, but after about 20KB, the files began to differ (the first 20KB were the same). Finally I repeated the same conversion a third time, now again using W98SE. Amazingly I got a perfect output this time, looked the same as the W2K output. But the binary differed a third time! And it is not only a time stamp/header difference.

I don't know enough about MPEG2-compression, but I'm wondering if it's normal that the same source with the same settings can result in different output using the same converter? Are there any random numbers involved? Is there any way to explain the glitch? Can this be a memory leak and I can expect to avoid them using W2K? Or do you have to expect such glitches just by not having full CPU power available? Of course you might say it's wiser not to use the CPU while doing MPEG2-conversion... still, this sounds a bit esoteric, and I'd like to know why parallel preocessing can influence the outcome of a conversion. Or am I just going slightly mad? :-)

Thanks for your help,

Klaymen

PS: My hardware is a 900MHz Athlon with Asus A7V motherboard. Never experienced IRQ- or instability-problems with it so far.