Minimum PC for Direct MPEG2 Capture

[jchunter]

Trevor,
Digressing a bit from the main question, Googling around got me to the following comments regarding the ATI AIW Video card at http://www.amazon.com/gp/product/B0000A ... 2&v=glance

One User Comment: "However there is a serious problem with the TV-tuner application. Namely, the TV viewing and recording software is not compatible with Hyperthreading. As most Intel Processors over 2.5 Ghz or so provide hyperthreading (turned on by default) this is a serious issue. "

A second user added: "The idea behind a video capture card is that you can pause, rewind and record live TV. By far the best way to do this is to buy a separate video capture card with hardware encoding. This way, the card converts the video signal into a format that can be saved on your hard drive (mpeg2).

Otherwise, your CPU must do the work. That's what ATI All-in-Wonder cards do. While ATI makes great video cards, integrated cards like the AIW cards rely on the CPU to do all the dirty work."

[Trevor Andrew]

Hi John

I connect my camera via Firewire. Dv-Out
I select the Capture Tab
I select Capture Video
Vs recognises my camera and switches it to pause, the video can be seen in the preview window.
I select Format Dvd, or Vcd, Svcd, Mpeg
I select the ‘Options Cogwheel’
I select the ‘Video and audio capture property settings…
I see the Video and audio capture property setting window
I see the Quality slider across the middle of the window.

If I select Dv, I capture to Dv-Avi, no quality slider is available.

I was using a capture card with my GS400 as a test to check on dropped frames.

Trevor

[jchunter]

Trevor,
Assuming you are using VS9 (as am I) the only difference is that you are selecting DVD PAL Format, and I am selecting DVD NTSC.

Two-Dogs,
Where do you live? PAL country or NTSC?

John

[Trevor Andrew]

Hi John

This is the capture window i see:-



Trevor

[jchunter]

Trevor,
The screen behind my gear icon is completely different: no slider, no properties window, template selection is a dropdown, ...
John

[2Dogs]

Two Dogs,
This morning, I captured mpeg2 from my Panasonic GS400 NTSC widescreen video at a frame size of 720x480, VBR = 8000kbps, lower field first, Dolby audio.

My VS9 capture property window does not have a quality slider, so I'm wondering where you are setting the quality. In any case, I set the propject properties to match the capture properties and pushed the Quality slider to 100% in the Edit Tab (Video Studio sometimes copies property values from here).

I carefully monitored the load on the system using the using the performance tab in the XP Task Manager. The memory usage was 403MB and the CPU averaged 45% with peak usage at 65% and a low of 40%. Both CPU graphs hovered at mid range throughout the capture.

My CPU (3.0 GHz.) is only marginally faster than yours. I am beginning to suspect that the Video Studio code may favor Intel hyperthreading over the Athlon (although it shouldn't)... You said that enabling HT made no difference. Does your Task Manager show two CPU performance graphs?

Edit: Rereading your earlier post, I see that you have 512 MB of memory and this may be making the difference because you may not have enough real memory to absorb the shocks of peak capture loading. Why not add another 512MB and see how this works...

John

Appreciate your further input John!

Regarding the cpu utilisation - you get two graphs if you have HT enabled, but there are few programs out there that manage to get both "halves" of the cpu running at 100%, so your 40 odd to maybe 65 peak seems fairly representative.

If you disable HT in bios and rerun the test, you'll see the (single) cpu running consistently at almost 100%.

My own experience with my machine is that HT has virtually no effect on VS8 or VS9 trial, either in capturing or rendering performance, but by contrast, enabling HT slows down other programs, such as Spybot (freeware anti-spyware tool) appreciably.


I'm coming to the conclusion that the answer to my original question is P3.2c and up for Intel, and I'm still waiting to hear from an Athlon XP 3200 owner.

[Trevor Andrew]

Hi John

Hi John

File / Preferences / Capture Tab / Un-tick ‘Allow access to capture device’s settings’.

Afaik this box is unchecked as default.

Trevor

[jchunter]

Trevor,
Wow! Unchecking "Allow access to device Settings"in Prefs, changed my capture display to the one you are seeing. My FIRST screen had been the one that you got when you clicked the "Advanced" button.

So, I shoved the Quality slider to max and started capturing a long stretch of video to the same properties as before. This time, the memory usage started at 404MB and slowly increased. The tape got to the end at 13 minutes and 11 seconds. At this time the memory usage had increased to about 1150MB. I say "about" because I immediately got a "Flushing Transcode Buffer" message for about 5000 frames and that number started decreasing. 5000/30 = 167 seconds of video had not been transcoded from DV into mpeg. It took about 3 minutes to flush the buffer completely. BTW, the CPU utilization peaked at 75% with a min at 45%. The right hand CPU averaged about 80% and the left hand one averaged about 30%.

So, if a 13 minute capture gets 3 minutes behind the tape, then it is capturing 23.1% slower with a 3GHz CPU. A quick and dirty estimate says that 2dogs would need a CPU that runs that much faster or 3.69GHz to capture an extremely long video. (all other things being equal, which they never are...).

It is also clear that if you hold the CPU speed constant, more memory and a larger pagefile would allow a longer capture. In my experience, a capture file longer than 10 minutes is twitchy to edit, so I don't see this as much of a problem.

If I get some time I will try a longer capture later and also try enabling write caching...

Another worthy question: "Is the resolution any better when a resolution chart is captured at max quality?"
John

[2Dogs]

Two Dogs,
This morning, I captured mpeg2 from my Panasonic GS400 NTSC widescreen video at a frame size of 720x480, VBR = 8000kbps, lower field first, Dolby audio.

My VS9 capture property window does not have a quality slider, so I'm wondering where you are setting the quality. In any case, I set the propject properties to match the capture properties and pushed the Quality slider to 100% in the Edit Tab (Video Studio sometimes copies property values from here).

I carefully monitored the load on the system using the using the performance tab in the XP Task Manager. The memory usage was 403MB and the CPU averaged 45% with peak usage at 65% and a low of 40%. Both CPU graphs hovered at mid range throughout the capture.

My CPU (3.0 GHz.) is only marginally faster than yours. I am beginning to suspect that the Video Studio code may favor Intel hyperthreading over the Athlon (although it shouldn't)... You said that enabling HT made no difference. Does your Task Manager show two CPU performance graphs?

Edit: Rereading your earlier post, I see that you have 512 MB of memory and this may be making the difference because you may not have enough real memory to absorb the shocks of peak capture loading. Why not add another 512MB and see how this works...

John

Thought I'd fill you in on my latest findings, which contradict just about all I've said about Hyperthreading so far!

Although HT makes no noticeable difference to direct to MPEG2 capture or rendering an avi to MPEG2, I find that it gives a whopping 20%+ benefit when rendering an avi with scrolling titles and audio, and overlays.

You get to see both halves of the cpu running much closer to 100%.

Admittedly for most projects, titles and audio won't take up that high a proportion of the timeline, but the benefit is sufficiently significant to cause me to now leave HT enabled.

[jchunter]

I did some long captures: one with a page file of 1GB and another with 2GB. Unfortunately, the transcode buffer flushes when it gets to (approximately) 1GB regardless of the page file size. Evidently a hard-coded default value. When I enabled write caching the buffer actually filled a bit faster. Go figure...

Finally, I shot 10 seconds of a resolution chart at the end of the video tape and compared a capture at 70% quality with a capture at 100% quality. The result: lines merge together at the same point. the resolution is identical.

John

[Ken Berry]

John -- I have been trying to stay out of this one. But once again congratulations on your perseverance. And in fact you only confirmed me in my view that by and large, we can leave the quality/speed slider where it is set by default. I am not sure that the extra time involved in the production process when you set the slider to 100% -- let alone the other complications like overtaxed CPUs and transcode buffers etc -- is worth the effort, especially when the final difference is not really distinguishable to the naked eye, which is, after all, what we all use to measure the worth of our product!!!

[jchunter]

Thank you, Ken. Maybe we all learned something.

I remember that someone else did a different kind of study (measuring file size, I think) on the effect of the quality slider when making video files and he concluded that it made little or no difference.

I was surprised to learn that a single preference value setting has such an effect on the user interface for capturing. The two screens that one gets by unchecking "Allow access to device Settings" are highly redundant except for the Quality slider. I think it is bad form to change the user interface unless there is a very good reason because it just makes a complicated interface even more confusing.

The fact that Preference settings are not saved in the Project file makes it very strange when a user works on different projects with different preference settings. One can open a project and be confounded that it behaves quite differently than it did the last time it was used.

Cheers,
John

[daniel]

[Without wishing to interupt your AVI work or bore you to death, might I just ask if you closed down all unnecessary programs and services to achieve the 87%?

Hi

Please read again second line of my statement (Win2000 as it boots) :-)

The only precaution I took was to defragment the destination drive due to my IDE 166 being not lightning fast, I did not want to add search times.
(Any way disk throughput is related to the 7000kbps, not to the quality setting, but experiments are experiments so get it all right...)

As a side note to all willing to add to the testing: I think what we check here is only relevant to non-hardware-assisted encoding (plain conversion cards) via analog capture cards to MPEG.
I see some people mixing in DV capture via firewire, that is irrelevant, any machine can do that at whatever setting, this is a simple data transfer and the performance is USB2 for everyone; that's all.

The subject of the test as I understood it, is basically VHS transfer capacity to PC via a straight converter card, with the genuine interrogation of what processor would keep 100% for several minutes without dropping frames.. Any correction welcome.

[2Dogs]

[Without wishing to interupt your AVI work or bore you to death, might I just ask if you closed down all unnecessary programs and services to achieve the 87%?

Hi

Please read again second line of my statement (Win2000 as it boots)

The only precaution I took was to defragment the destination drive due to my IDE 166 being not lightning fast, I did not want to add search times.
(Any way disk throughput is related to the 7000kbps, not to the quality setting, but experiments are experiments so get it all right...)

As a side note to all willing to add to the testing: I think what we check here is only relevant to non-hardware-assisted encoding (plain conversion cards) via analog capture cards to MPEG.
I see some people mixing in DV capture via firewire, that is irrelevant, any machine can do that at whatever setting, this is a simple data transfer and the performance is USB2 for everyone; that's all.

The subject of the test as I understood it, is basically VHS transfer capacity to PC via a straight converter card, with the genuine interrogation of what processor would keep 100% for several minutes without dropping frames.. Any correction welcome.

Hi Daniel!

In fact my original topic was direct to MPEG2 capture, using a MiniDV camcorder with a firewire cable. (although I'd unwittingly jumped in on a thread started four days previously by GeorgeK, for which I apologise again)

Despite many protestations on my part, many people seemed to think that I was advocating the direct to MPEG2 capture approach rather than the "recommended method". This is not so. I mostly work with avi clips now that I have been enlightened, having read the "sticky". I'd originally made the mistake of reading (and believing) the VS8 manual. Importing project files in the share step seemed like such a great idea! Direct to MPEG2 capture and Smart Render - fabulous! In fairness to VS8, my fisrt few DVD's of school plays, musical recitals and so on all had better finished quality than "rival" productions or tests with the likes of ArcSoft, Roxio or WinDVD Creator 2.

I find from my own empirical and no doubt subjective tests that there is a noticeable difference in output video quality when you capture at 100% rather than the default 80% or whatever VS decides for you. This may well not be apparent on resolution charts - the principal difference is with fast moving and changing sections. (I just can't get my dogs to run slowly in the videos, and they're often on grass!) Though I'm not familiar with the use of resolution charts, they sound to me more like static images.

In any event, the real point of the exercise was to evaluate the performance of different, commonly available platforms with the use of VS. It would have been nice to learn how Celeron and Sempron systems perform.

In any event, I believe the ability of a machine to capture directly to MPEG2 at 100% quality is a useful measure of it's performance running VS. Perhaps it will be less so when even common budget systems can achieve it.

By contrast, if you compare rendering times, there are so many variables.

For example, my own setup will render a sample of my avi footage to constant bitrate 8000kbps MPEG2 at approximately 1.7 x real time.

If I add scrolling titles and an mp3 audio track, it goes up to 2.8 x real time.

Adding a chroma-keyed video overlay to all of the above (multiple insertions of the VS9 sample avi V15.avi) slows it down to 4.5 x real time.

You can therefore see that there are too many variables involved if you try to compare project rendering times.

At the end of all of this, it's true to say that the faster the cpu, the better. More RAM seems to help the program generally by all reports, though it probably doesn't have much influence on render times. I won't be able to confirm that for myself until I buy more RAM, though I can check out running with only 256Mb. XP is pretty RAM hungry, though, and 512Mb is generally acknowledged as a practical minimum.

The big surprise for me was the effectiveness of Hyperthreading, which I'd previously thought to be only "hype". Now I don't feel too bad about laying out 500 bucks for my pc two years ago! Interesting also that Intel will drop HT on future cpu's.

I will therefore continue to capture to AVI using WinDVD Creator 2, which seems to work flawlessly with my particular camcorder. I can then also use some of that program's great transitions and filters (two of my favourites bieng "video wall" and "rain", and use VS for adding titles, sound, overlays and output to the best possible quality MPEG2.

I really hope that VS10 will combine the best of Ulead's and Intervideo's programs!

Talking of titles, that will lead me to my next posting.... but I'll be sure to do a thorough search of the archive first and avoid a repetition of my earlier transgressions.

Thanks to all for their input and tolerance on this one, I did manage to glean most of the info I sought!