Minimum CPU for on-the-fly MPEG-2 software encoding?

[MrMungus]

Hello again!

Been wondering about this. I've been testing various ways to do video capture ranging from MPEG-2 hardware capture cards to Firewire capture with on-the-fly software MPEG-2 encoding.

My new computer (3Ghz P4) can easily handle on-the-fly software MPEG-2 encoding (only 25 to 35 percent CPU utilized while doing it). What I'm wondering is if anyone knows what the minimum requirement is for that? Can a 2Ghz CPU do it? A 1Ghz?

One reason this is an important question is because when people ask (here and elsewhere) what is required, my answer is always, "How fast is your computer?" since I think it's better to have an analog>Firewire converter (ADVC-100) than it is to have a hardware MPEG-2 encoder (Hauppauge, etc.) since the analog>Firewire devices give you more flexibility (ability to capture to AVI *or* to MPEG-2 via on-the-fly software encoding).

So I'm hoping to nail down an answer to this question.

Comments? Experiences?

[SaSi]

I hate to disappoint you, but I don't believe it's really possible to determine a minimum CPU for on-the-fly MPEG-2 encoding.

It simply has to do with the s/w used to encode. Some capture/encode tools seem to need much more CPU power than others.

In general, using an efficient tool, perhaps more than 2GHz is enough.

[MrMungus]

I hate to disappoint you, but I don't believe it's really possible to determine a minimum CPU for on-the-fly MPEG-2 encoding.

It simply has to do with the s/w used to encode. Some capture/encode tools seem to need much more CPU power than others.

In general, using an efficient tool, perhaps more than 2GHz is enough.

Hmm. But surely there is a range in which most software encoders fall? MPEG-2 is a standardized format; how much range could there possibly be if all encoders are encoding to the same specs? (except, perhaps, differences in CPU utilization occuring due to VBR vs. CBR and/or overall bitrate and/or audio codec used)

[offline]

all encoders are encoding to the same specs.

What you are referring to is mpeg2"output". How this output is created depends on the encoding engine and method employed - and this varies greatly.

Ask them what video package that they want to use. If they do not know, get them to look up for themselves, or suggest a package. Then go with the recommended system spec's of the software maker (not the minimum.)

[rkr1958]

FYI. I have an AVER-TV tuner which also captures to MPEG-2. On my old computer I could capture TV shows to MPEG-2 at 360 x 240 and 3000 kbps with my 733-MHz PIII. The resultant quality was as least as good and in most cases better than VHS.

[FOO]

Here's a data point. I have a 2 GHz AMD. I usually run it at 1.2 GHz because
except for encoding , who needs it , and it gets hotter than hell.

I regularly capture movies at 720 x 480 MPEG2 4200 kb/s with an
ATI AIW 8500 in slow mode. No problems
If I try 5000 kb/s it drops frames a little.

If I crank it up , it goes 8000 kb/s easy

[SaSi]

[quote="mrmungus
Hmm. But surely there is a range in which most software encoders fall? MPEG-2 is a standardized format; how much range could there possibly be if all encoders are encoding to the same specs? (except, perhaps, differences in CPU utilization occuring due to VBR vs. CBR and/or overall bitrate and/or audio codec used)[/quote]
Standardized, yes, specific, no.

Encoding at 720x576 takes a lot more time than encoding at 352x288. Theoretically should take 4 times as much based on the area ratio.

Encoding in MPEG2 involves motion estimation. This helps reduce bitrate. Using a low setting in motion estimation or disabling it alltogether can yield tremendous differences in encoding speed. Tmpgenc seems to range 16:1 between the lowest and highest setting. (Quality is another issue).

And all that applies to any encoder application, so it depends on what anyone is looking for when encoding.

I take MPEG2 to be a "final result" format. To be used within the VCD/SVCD/DVD medium. As such, I believe that this encoding step should be as elaborate as practically possible to ensure quality is not lost while not wasting file size.

Unless capturing to MPEG-2 using a hardware encoder device (like a Deck DVD recorder, an MPEG-2 capture card, etc), I don't believe that capturing via WDM and encoding to MPEG-2 yields any significant advantage. Typically, the captured material needs some extent of editing. Then, the captured material is typically captured at a relatively high bitrate CBR mode. This is a waste of bitrate so a re-encoding or transcoding step would yield the same quality at a lower bitrate.

So why bother with MPEG-2 encoding during capture at the first place?

Incidentally, I find that most of the questions about real time encoding are a result of using Tmpgenc. People try it, it is slow, so they try to find ways to reduce this lengthy step. Eventually, the idea of capturing in the final format comes up.

Then we have the questions about MPEG-2 editing. Capture at MPEG-2 and crop/trim the capture without loosing the precious encoded file.

Don't you think?

[MrMungus]

all encoders are encoding to the same specs.

What you are referring to is mpeg2"output". How this output is created depends on the encoding engine and method employed - and this varies greatly.

Ask them what video package that they want to use. If they do not know, get them to look up for themselves, or suggest a package. Then go with the recommended system spec's of the software maker (not the minimum.)

I mean creating an MPEG-2 file versus creating an AVI.

I don't understand how the CPU requirements can vary so widely as to not fall into a certain range. If the input is standard (analog video) and the output is standard (MPEG-2 at whatever bitrate) what's the difference between software encoders such that you can't say that it generally falls within a certain range? I mean, if I use encoder X vs. encoder Y, what would they be doing different to account for such a huge difference?

[MrMungus]

FYI. I have an AVER-TV tuner which also captures to MPEG-2. On my old computer I could capture TV shows to MPEG-2 at 360 x 240 and 3000 kbps with my 733-MHz PIII. The resultant quality was as least as good and in most cases better than VHS.

Yeah, one of the cards I first purchased for testing was an Avermedia card but I couldn't use it at the time because I didn't have the new computer yet...and I'm talking about CPU for 720x480...

[MrMungus]

Encoding in MPEG2 involves motion estimation. This helps reduce bitrate. Using a low setting in motion estimation or disabling it alltogether can yield tremendous differences in encoding speed. Tmpgenc seems to range 16:1 between the lowest and highest setting. (Quality is another issue).

Always or just when doing VBR or something like two-pass CBR?

Unless capturing to MPEG-2 using a hardware encoder device (like a Deck DVD recorder, an MPEG-2 capture card, etc), I don't believe that capturing via WDM and encoding to MPEG-2 yields any significant advantage. Typically, the captured material needs some extent of editing. Then, the captured material is typically captured at a relatively high bitrate CBR mode. This is a waste of bitrate so a re-encoding or transcoding step would yield the same quality at a lower bitrate.

Not always. Thanks to TMPGEnc DVD Author you don't have to edit anything, just pick the parts you want written to the DVD and which parts you want to skip/clip.

So why bother with MPEG-2 encoding during capture at the first place?

Speed (edit: and quality; been reading up on 4:2:0 vs. 4:1:1 lately...). And my captures have been comparable to their source. I should probably point out, though, that I'm not trying to squeeze more than 1 to 2 hours onto a DVD. If I *do* do a direct-to-MPEG-2 capture and the resulting file is too big for one DVD, I shrink it with DVD Shrink.
[/quote]

Incidentally, I find that most of the questions about real time encoding are a result of using Tmpgenc. People try it, it is slow, so they try to find ways to reduce this lengthy step. Eventually, the idea of capturing in the final format comes up.

Then we have the questions about MPEG-2 editing. Capture at MPEG-2 and crop/trim the capture without loosing the precious encoded file.

Don't you think?

Well, like I said, editing isn't always necessary (TDA) and my results have been very good, good enough to not even consider the lengthy TMPGEnc step because when I compare the original (tape) to the resulting DVD, I often can't tell the difference. Maybe it's more a question of speed vs. size vs. ease and everybody here puts a different priority on each of those factors... But hey, I'm *always* willing to listen.

[lordsmurf]

Click on COMPUTER DETAILS by by name ... I use these to encode MPEG-2 on the fly, no problems.

[MrMungus]

Click on COMPUTER DETAILS by by name ... I use these to encode MPEG-2 on the fly, no problems.

720x480? What's your CPU utilization % while capturing?

[rcguy]

I use Pinnacles Studio 9 to capture and encode MPEG-2 and the Studio handbook says your cpu must be a minumum 2.2Ghz

[MrMungus]

I use Pinnacles Studio 9 to capture and encode MPEG-2 and the Studio handbook says your cpu must be a minumum 2.2Ghz

What's your CPU load like while capturing?

[lordsmurf]

For the record, 720x480 sucks on an ATI AIW card since the card is made for 704x480 and 352x480

... but captures in 704/720x480 give me a CPU usage of about 60-80% ... which represents the hybrid hardware/software

... add any VideoSoap and you can use 60-90% depending on the strength and type of VS filter ...

Compare this to true software-only encoders like PowerVCRII and WinDVR that use 90-100% CPU.

Captures in the 352x480 range use maybe 20-30% CPU on the ATI AIW.

[SHS]

Real Hardware MPEG2 Encoder has no CPU load while capturing 0% other wirte the data to the harddrive.

[bottle-necked]

Viewing TV with MMC = 2%-5%. Average 4% cpu.
720x480 MPEG2-DVD 8m, 6avg = 10-17%. Average 14% cpu.
352x480 MPEG2-DVD 4m, 3.9avg = 8%-11%, Average 10% cpu.
352x240 MPEG1 1.85CBR = 4%-9% = Average 7% cpu.
352x240 MPEG1 VCD default = 4%-6%. Average 5% cpu.

Programs active:
Outlook Express
IExplorer
Windows Task Manager
Online w/Cable Modem
29 active processes
PF Usave = 169MB

No Soap...

[MrMungus]

Real Hardware MPEG2 Encoder has no CPU load while capturing 0% other wirte the data to the harddrive.

I know, but I was asking about software MPEG-2 encoding because I'm wondering what the minimum CPU is that one would need to have to do software encoding. I know I can run a hardware MPEG-2 encoder on my 750 Mhz CPU here and a 500 Mhz CPU at work and I can do software MPEG-2 encoding easily on my 3Ghz here, but I was wondering what the minimum is since I don't have any computers betwee 750 Mhz and 3 Ghz.

[lordsmurf]

Hardware MPEG encoders still use up 10-30% or more CPU too, because it still uses system resources like the capture programs, RAM, HD, etc (some of them bypass CPU, but I forget which ones).

So, no ... hardware MPEG is not 0% either, just about half or less.

I'm looking at the big picture, not just what the card does.

[SHS]

Not really but then againe my tired in to REAL Hardware Decoder know as REALmagic Xcard and oh you not have any high resources with Pervierw mode turn off unless you have a WinTV-350 output to TV you be make used of it Hardware decoder and most you ever saw up 10 to 15% on a lower end PII 300a Celeon.
Heck not long ago I took old Intel P133 Dell system and pop in my 250 and Xcard and must I ever saw with was up to 25% all done in REALtime.
Yes it will depend on capture programs but good one should be in ide mode while capture card dose it things in tell other wize told to stop by the capture programs.
mrmungus it mostly depend on the REZ you plan to use and weather not super clean signel and really soild input source min 2.0GHz for doing both field 480.

[ktnwin]

neoDVD can capture DV to MPEG-2 in real time using an Atlhon 1.2 Ghz.
Tried it, and it works.

[SatStorm]

I can capture realtime mpeg 2 @ 352 x 576 CBR (any bitrate), no preview, with a duron 700 using mainconcept 1.4.1, no problems
I can capture realtime mpeg 2 @ 704 x 576 VBR (any bitrate) with a duron 1200 using mainconcept 1.4.1, no problems.
I can capture realtime mpeg 2 @ 704 x 576 VBR (any bitrate) with an athlon xp 1700, using mainconcept 1.4.1. and the built in Noise Reduction filter, no problems.

With my Athlon 2600 XP I can capture at any framesize, VBR (any bitrate) , with the built in noise reduction filter using mainconcept 1.4.1 and at the same time cut mpeg 2 with mpeg2vcr or serf to the internet....

I love mainconcept 1.4.1 for realtime mpeg 2 captures. Not that I use this fuction so much (I'm a fun of the avi capture/ post proccessing / mpeg 2 encoding rute), but it is nice to know that I have a fast alternative for the time I'll need it.

[Jester700]

This isn't cut & dried, because many softwares have a quality slider that adjusts...well, SOMETHING. I don't know what the parameters are, but basically you set the slider higher and you get better quality, at the same chosen bitrate - BUT - you need a faster CPU. Set the slider too high and the CPU can't keep up, and after the cap is "done" it will still be encoding from a buffer. On my Athlon 3000+ I can use a setting of 30 (the range is 1-50) on MainConcept's encoder.

With Ulead's codec, I could do realtime on the Athlon 1700+ I had then, but my setting was "medium" or something. It looked terrible. With my faster CPU, the MainConcept one looks pretty good, but not quite as good as my hardware card (Hauppauge PVR250)

[ronnylov]

I remember when I had an AMD Athlon 1000 MHz and I got frame drops when capturing 720x576 PAL MPEG2 on my ATI AIW Radeon but 480x576 SVCD captures was OK. But overclocking to 1133 MHz made it possible to capture 720x576 MPEG-2 without frame drops. Now I have changed CPU to a AMD AthlonXP 2400+ (2000 MHz) and I have around 60 - 70 % CPU utilization capturing 704x576 MPEG2 with all settings maxed and a light video soap filter enabled in ATI MMC 8.8.

[MrMungus]

Hardware MPEG encoders still use up 10-30% or more CPU too, because it still uses system resources like the capture programs, RAM, HD, etc (some of them bypass CPU, but I forget which ones).

So, no ... hardware MPEG is not 0% either, just about half or less.

I'm looking at the big picture, not just what the card does.

I did a test last night. I connected the ADS Instant DVD MPEG-2 encoder to my P3 750 Mhz (through a new USB 2.0 PCI card) expecting the CPU usage to be relatively low. I was surprised to see that the CPU for an 8Mbps capture was running at 85 to 95 percent CPU (and for 4Mbps, 70 to 80 percent). I was surprised because when I tried the Hauppauge PVR-250 card on the 750, I think the CPU was a lot lower (I'll have to test that again).

[SteMan]

LordSmurf said:
Compare this to true software-only encoders like PowerVCRII and WinDVR that use 90-100% CPU.

---

I currently use WinDVR and it only uses 60-70% CPU.

[MrMungus]

LordSmurf said:
Compare this to true software-only encoders like PowerVCRII and WinDVR that use 90-100% CPU.

---

I currently use WinDVR and it only uses 60-70% CPU.

I think it depends on your CPU...which is why I'm asking. What CPU you got?

[bottle-necked]

Viewing TV with MMC = 2%-5%. Average 4% cpu.
720x480 MPEG2-DVD 8m, 6avg = 10-17%. Average 14% cpu.
352x480 MPEG2-DVD 4m, 3.9avg = 8%-11%, Average 10% cpu.
352x240 MPEG1 1.85CBR = 4%-9% = Average 7% cpu.
352x240 MPEG1 VCD default = 4%-6%. Average 5% cpu.
No Soap...

MainConcept 1.4 trialware. First run new install.
Default MPEG-DVD preset using provided MC MPEG encoder.
720x480 8m, 6avg, 3.2min
48k MPEG1-layer 2 256k audio
Interlaced

Computer/programs about the same as MMC test.

**************
Viewing TV (sat) at idle in MC 1.4 = 0%-2%. Avg 1% cpu.
720 capture (as stated above) = 33%-50%. Avg 40% cpu.

Quality about the same or better then MMC, but larger file size (per/min). About 98minute DVD compared to ATI's about 123minutes.

Hope this helps ya.
Good luck.

[vhelp]

mrmungus,

Also, you have to factor in that these real-time mpeg encoders have different
"mpeg encoding algorithems" (MEA) and as such, will reflect what a given
CPU setup will handle in % (or cpu resources) these days.

And, since there are a sworm of PC makes and models, and mobo's and
cpus and chipsets and bios and driver versions and software capabilties,
it's not wonder that very few low-end setups will pass with no hint of issues.

But, it's important to realize the the MPEG "MEA" is the key factor here.. not
to mention cpu speed too.

For real-time MPEG ...

But, the number one thing to look for first, is to prioritize the following:
* capture card
* mobo and chipset
* mobo/chiset drivers
* hd (though not as important w/ output source is to be mpeg)

All these must work in concert together, and w/ no issues.
.
.
* Then, the software real-time mpeg should be looked at for maximum speed
...and quality.

Also, remember that these SW real-time MPEG, that in order for them to
work well via software, is to cut corners. You can still get good results, but
just think for a minute, software like CCE ( $2000 ) would not work in this
setup, (if you want the quality it gives out) And, even TMPG would not as
well, too. So, there is some meat that is shaved off (or cut corners) to get
the results, but w/in reason or that can fool the eye to some degree.
Mind you, everyone's eyes are different too

But, perhaps several catagories should be used when comparing best pc
setup.
* CVD (352 x 480) ... would work best under these cpu xxx's
* DVD (720 x 480) (or 704 x 480) ... would work best under these cpu xxx's

-vhelp