HARDWARE VS SOFTWARE ENCODING CARD

[BV2001]

Hi
I currently own an ATI AIW 8500 and use this to capture movie from recorded VHS. I wonder what is the real different between hardware encode ( ie, Dazzle or Win TV PVR ) and the software ( ie, AIW ). Does the hardware capture image better? Like if I am using dazzle II to capture DVD MPEG2@10000Mbit and AIW DVD MPEG2@10000Mbit using WinDVR. Which one will have a better capture? both at DVD Resolution
thanks

[kalyan]

you should read other posts also man.

hardware encoding is better.

[BV2001]

hey bud,
I know
I had done captures with my AIW and it looks excellent. I never own a hardware encoding yet so I don't know if it is better. I know that in order to get the maximum quality for VCD, oh, I convert to VCD. I always cap to mpg2 then reencode it. I will do the same for the dazzle II if I own it. Just wonder, capping alone without reencoding, does the dazzle II make a big gain comparing to the AIW.

[zap4]

Here is one article I have found on the subject: http://www.aardvark.co.nz/pvr/pvr7.shtml

[miamicanes]

Actually, the line between hardware and software encoding is kind of like the line between hardware modems and winmodems -- there are the two obvious extremes, and a grey area in the middle.

Pure software-only MPEG-2 encoding, relying upon nothing more than the might of a fast CPU, is still kind of a pipe dream. 2+ GHz CPUs are fast enough to make it work JUST often enough to frustrate users with an occasional dropped frame or crash to ruin an otherwise successful capture. 600+ MHz CPUs can occasionally get a good capture, but only if the user is willing to settle for low-compression I-frames.

At the other extreme are all-hardware encoders. The good ones are ungodly expensive, and the cheap ones produce output that's either big or mediocre (occasionally both).

In the middle lie the ATI cards, and probably a few others (Matrox, maybe?). They rely on software to handle the MPEG assembly logic, but have a few DSPs at their disposal to do things that aren't conceptually "hard" to do, but take lots and lots of time. Like discrete cosine transformations, motion detection, etc. The net result is a card capable of high-quality captures with sub-optimal compression that falls about halfway between the best TMPGEnc can achieve and raw, unwashed I-frame-only captures of equivalent quality).

Less than three weeks ago, I passionately hated ATI. However, now that my card (AIW 128 pro) finally is working the way it's SUPPOSED to work (with neither ATI's help nor blessing, but that's another tirade), my opinion of it has gone up significantly.

It's a shame that standalone utilities like TMPGenc can't take advantage of the Rage Theatre chip's capabilities (at least, not yet). If it COULD, transcoding HuffyUV-captured video into MPEG-2 would probably take somewhere between half and twice the clip's actual playing time, rather than 4-10 times the playing time like it does now.

-------

Side note -- for those wondering about the winmodem analogy, the same situation exists. At one end are all-hardware modems which work well, but cost way more than they're worth if you've got broadband and only use it to send faxes and 2-3 times a year when the cable/DSL goes down. At the other extreme are the all-software winmodems that literally use a cheap soundcard chip to digitize the other modem's sound and force the CPU to slough through millions of fast fourier transformations to decode it into a bitstream; they suck, and should be avoided at all cost -- the PERFECT gift to give people you positively HATE Finally, there are the hybrid modems (mostly Lucent chipset) that have a hardware DSP to do the grunt work of turning the other modem's audio output into a bitstream, but use the host CPU for logic and control. The DSP adds practically nothing to the final cost, but makes the final product a million times better.

[SHS]

BV2001 some fact Hardware vs Software.
Hardware can do VBR mode where software can't.
Hardware give a lot lower CPU usage and as miamicanes point out software is a CPU hog.
miamicanes the rage theatre chip is nothing more then iDCT+MC decoder it not an encoder.

[q1aqza]

At the other extreme are all-hardware encoders. The good ones are ungodly expensive, and the cheap ones produce output that's either big or mediocre (occasionally both).

I disagree that the cheap ones are mediaocre. On a PVR250 (cheap at around £140) at high bitrates (7000 +) you would be hard pressed to tell the difference from software encoder. Ok you would notice it on your PC but burn on to DVD and play thru your TV and there's not alot in it. Even at lower bitrates the PVR250 produces good DVD quality.

[GenePool_]

Also, you can do VBR encoding in software.
PowerDirector and Ulead poducts will let you do this.

[BV2001]

thnks guy
My system is AMD XP2100+ and MSI KT400 board with 1G DDR333 Ram.
I did try to download the capture DVD Sample for Dazzle and WinTC PVR and they are great
Just don't know if it is worth to switch from AIW to PVR or Dazzle now. My main goal is the quality. I dont' mind to cap in DVD resolution and reencode to VCD. But QUALITY is what I am really looking for.
What would you guy pick???
AIW
dazzle II
Win TV PVR 250???

[SHS]

GenePool_ Software can not do REALtime VBR encoding due to way VBR works how ever off line encoding yes.
I go for WinTV-PVR 250 the Dazzle DVC II don't a lot of VIA motherboard
You can go to Dazzel forum and ask you self about them

[miamicanes]

miamicanes the rage theatre chip is nothing more then iDCT+MC decoder it not an encoder

As I understand it, the chip works both ways, but prior to MMC 7.6 it was only used for decoding.

I have to admit I haven't viewed the source code or chip data sheets for the AIW Radeon 9700, but there's plenty of evidence to suggest that the newest AIW's hardware-encoding capabilities are kind of like the new features introduced with Windows 98 -- already present (or capable of being tweaked into activation) in the older ones anyway.

Prior to MMC 7.6, MPEG-2 capture at DVD resolution was flat-out unusable, even at 1.8GHz, and even MPEG-1 was unreliable and dropped lots of frames. Now, with 7.7 (same hardware), I can do realtime caps at full DV-1 resolution at maybe 40% CPU usage, and I've heard that even 800mhz PCs can pull it off reasonably well. Either ATI pulled off one hell of a software optimization project on the CODEC, or some previously-untapped hardware resource is now being utilized.

My personal (semi-conspiracy) theory is that ATI's management put all their development resources behind the Radeon's 3D capabilities in mid-2000, and had MAYBE one or two developers who maintained the capture drivers as a secondary focus. Let's suppose some genius in management decided to lay one of them off as a cost-cutting measure, and the other later quit... leaving ATI with NOBODY in their development team who REALLY knew how the capture subsystem worked. A few guys who'd touched it a few times and could graft on a small patch here or there, but nevertheless lived in mortal fear of doing anything major that might have caused the whole thing to break, and had to put up with their bosses griping the whole time about them spending "too much time" on the capture drivers instead of working on More Important Things (like 3D).

At that point, there was probably a code branch that used the hardware to assist capturing, but for whatever reason it wasn't included in the release builds. Eventually, ATI's management probably figured out that they had a major crisis on their hands, desperately told 2 or 3 other developers to "study" the code and try to figure it out, and held a few months of meetings before finally committing the resources to pull the driver subsystem out of the mothballs and dedicate a couple of programmers to its active development -- the initial result being MMC 7.6/7.7, and upcoming 8.x with capture-time processing and effects.

[SHS]

Yes from what I read the only new rage theatre 2000 chip like what found on AIW Radeon 9700 but not any of the older rage theatre chip and new 2000 chip only help out 10% of the encoding.

"You can do realtime caps at full DV-1 resolution at maybe 40% CPU usage" Gee when I Max out my PVR 350 720x480 12MBits/sec with VBR on 0% to 5% at best on 1.6Ghz with SDRAM base motherbaord and I'm also using Hardware MPEG2 TVout decoder and I can play UT2003 wail it recording in the background let see you try that.
Romor has it there planing for all new AIW but no word on weather not it be getting full HW encoding and decoding in the first place which never happing just like the miss MPEG-4 recording.

[Donny661]

@BV2001

Wouldn't your best bet be to capture in huffyuv or mjpeg then convert to vcd? I don't do any captures, but why would you capture to a highly lossy format such as mpeg2 when you could capture to a lossless or less lossy format then convert it and you would get better quality. The only reason I think you would want to capture at DVD resolution and bitrates in mpeg2 would be if you wanted to take that captured video and right away burn it to a dvd-r without having to go through the time consuming step of reencoding.

[drewson99]

except that newly captured mpeg2 files (at least from my ATI 7500) aren't standards compliant, and will not play when burned.

anyone else have this happen?

[W_Eagle]

Hello pal,

@BV2001
Wouldn't your best bet be to capture in huffyuv or mjpeg then convert to vcd? I

Yes, this is what I do with ATI AIO 7200. I get great results and full control over fitting feature movies into 2 CD-Rs in CVD format.

Take care

[BV2001]

I don't know but from my experience I actually get better result when capture in MPEG2 and reencode into VCD . Huff is ok but not as good as mpg2 to vcd, and I use Windvr to get the result. Capture with the same method using MMC and MMC turn out worst.

[Bizuser]

Drewson99

except that newly captured mpeg2 files (at least from my ATI 7500) aren't standards compliant, and will not play when burned.

I have the same issue with mpeg2 caps; they are non-compliant.

However, several people here have posted that the MMC 7.7 software is now compliant. Are you using MMC 7.1? (MMC is the software that ships with the ATI card). Miamicanes and others have tried to load 7.7 with some success on the older cards. I haven't tried it yet.

Not only does it not burn to SVCD format, it also falls apart if you try to cut it in TMPGEnc.

I do my captures in avi format using Huffy or MJPEG codecs. But if all you want to do is capture a 1 hour show and burn to disc, going through the encoding stage is a royal pain and 'should' be bypassable!

[Bizuser]

Hey, Miamicanes - did you go out and buy a newer ATI AIW card? You recently gave me tips on how to load MMC 7.7 on an AIW Rage 128 system. Have you since bought a Radeon?

[tjdmobile]

In line with SHS -

I have the Hauppauge pvr-250 AND the ATI AIW Radeon 32DDR. There are a couple of things to consider here, first being: Hauppauge card offers much better image clarity in real-time encoding. Hauppauge has superior VBR captures over MMC.

As example, I record a LOT of daily TV, and I use 352x480 Mpeg-2 video. I have capture rates of 1700 vbr (which is VERY accurate). I have a max bitrate of 4000. This produces video without macroblock and quality I never saw from the ATI card.

PLUS: There are now some outstanding 3rd party applications available for the PVR-250/350.

My ATI card is now just for TV-Out.

TJD

[TBoneit]

except that newly captured mpeg2 files (at least from my ATI 7500) aren't standards compliant, and will not play when burned.

anyone else have this happen?

I use theATI AIW 7500, to capture DVD but haven't even loaded the ATI MMC. I capture through other software from that ATI, Sonic MyDVD that came bundled works fine and using it I can autoset chapter points into the dvd capture. Total burn time for a full DVD is approx 1 hour. That includes selecting the video, saving the project, selecting the drive to burn to, the software planning audio, video, generating menus, transcoding and burning. I use the AIW for caps from sat TV, I use a hardware encoder to cap from tape as the SOnic is to sensative to dropouts and stops capping if a dropout occurs. But then I have no auto chapter and no quality settings which I get with the AIW, Thats why I use both. Using the Sonic S/w I can choose 3 quality levels of capture. Approx 1 hr. 2hr & 3 hr.

Good Luck