Tweet
Hi Rob,
Thanks for the effort and time you took testing this on your system.
To answer your first question though. I resized in TMPGEnc by setting the output to 480*576. The frames sent to TMPGenc were sized 704*396. I left all the sizing in the beginning to TMPGEnc and just set it for source video to 16:9 and output video to 4:3.
You mentioned your memory usage did not increase? Mine was steady also on mostly around 225 to 275Mb. The available physical memory though in time dropped to on average 2 to 10Mb. I assume this was the same with your setup? Performance monitor from w2k also didn't show that. I only found that out after I used a more decent memory monitor, which is part of a cache manager, and plotted a graph on the real memory used by the system.
I've in the mean time also done a lot more testing and found that if I make certain changes the time it takes for the problem to occur is delayed. Examples of that are the known multiplication of 16 for the frame size with encoding. What you did already yourself, being resizing in Vdub to 480*576. Using stacked AVISource commands instead of one SegmentedAVISource command. Adding boarders to the 704*368 capture to make 704*576. I've also tried older and newer versions of TMPGEnc, Vdub and even the first version of AVISynth. All of these put together delayed the time the problem occurred, but it's still there. All of it put together almost for sure confinces me the cause is AVISynth. If it were something with memory management on the TYAN mobo it would not have changed, but instead stayed the same.
I've in the mean time also received an answer from Edwin v Eggelen, the person that currently continues development of AVISynth. He thinks there might be some sort of memory leak in AVISynth. He mentioned that he is about to release a new version where a lot of bugs have been solved. When I get it I'm gonna test it hoping it also solves my problems. If not he mentioned he would check into this problem. So I'm much more hopefull here.
I never thought that this dual CPU system would cause me such a pain. It all looked so promising on first trials I did with it. I had a reduction in encoding time to alsmost 1/5th, sometimes even 1/6th, compared to my old PII-450. A well, such is life. In Holland we have a saying, buying cheap is expensive. That's probably true in my case also.
Thanks again for your great help in specially setting up your dual PIII-450 to test this. It confinces me that at least I'm not crazy as there is some issue with AVISynth in combination with dual CPU systems.
Best regards, Leon
PS
Only for comparison I sometimes calculate back to Dutch guilders. It all seems so cheap now, so I really have to surpress the urge of buying more then I would normally do.
I was very disappointed on my first bankreceipt this year though. It was at an all time low. 
We'll get used to it though for sure.
Thanks for the effort and time you took testing this on your system.
To answer your first question though. I resized in TMPGEnc by setting the output to 480*576. The frames sent to TMPGenc were sized 704*396. I left all the sizing in the beginning to TMPGEnc and just set it for source video to 16:9 and output video to 4:3.
You mentioned your memory usage did not increase? Mine was steady also on mostly around 225 to 275Mb. The available physical memory though in time dropped to on average 2 to 10Mb. I assume this was the same with your setup? Performance monitor from w2k also didn't show that. I only found that out after I used a more decent memory monitor, which is part of a cache manager, and plotted a graph on the real memory used by the system.
I've in the mean time also done a lot more testing and found that if I make certain changes the time it takes for the problem to occur is delayed. Examples of that are the known multiplication of 16 for the frame size with encoding. What you did already yourself, being resizing in Vdub to 480*576. Using stacked AVISource commands instead of one SegmentedAVISource command. Adding boarders to the 704*368 capture to make 704*576. I've also tried older and newer versions of TMPGEnc, Vdub and even the first version of AVISynth. All of these put together delayed the time the problem occurred, but it's still there. All of it put together almost for sure confinces me the cause is AVISynth. If it were something with memory management on the TYAN mobo it would not have changed, but instead stayed the same.
I've in the mean time also received an answer from Edwin v Eggelen, the person that currently continues development of AVISynth. He thinks there might be some sort of memory leak in AVISynth. He mentioned that he is about to release a new version where a lot of bugs have been solved. When I get it I'm gonna test it hoping it also solves my problems. If not he mentioned he would check into this problem. So I'm much more hopefull here.
I never thought that this dual CPU system would cause me such a pain. It all looked so promising on first trials I did with it. I had a reduction in encoding time to alsmost 1/5th, sometimes even 1/6th, compared to my old PII-450. A well, such is life. In Holland we have a saying, buying cheap is expensive. That's probably true in my case also.
Thanks again for your great help in specially setting up your dual PIII-450 to test this. It confinces me that at least I'm not crazy as there is some issue with AVISynth in combination with dual CPU systems.
Best regards, Leon
PS
Only for comparison I sometimes calculate back to Dutch guilders. It all seems so cheap now, so I really have to surpress the urge of buying more then I would normally do.
I was very disappointed on my first bankreceipt this year though. It was at an all time low. We'll get used to it though for sure.
Comment