DICE encodes 2-4-6 channel audio into a compressed DD signal on the fly output. Hence Dolby Interative Content Encoder. I got that much, but on my soundstorm codec (the newer Realtek ALC658 found in the Abit AN7 and NForce3 Boards) it can output 24/96 spdif. So in Foobar2000/Winamp5 I use the ASIO output plugin as well as the MAD Decoder (for both programs) set to 24/96 (it actually works).


I'm wondering is there any technical advantage of using my DICE output over audigy 2's analog out?

What i'm worried about is resampling, and if the soundstorm does bit-perfect output. Does DICE resample to 48khz like the realtek/audigy2 DAC's do? Or is the digital stream forked out before it hits the DAC and then DICE'd then out the digital out so it doesnt hit any resampling.

If DICE doesnt resample and does bit perfect out, then I can rely on my super nice decoder I got to do the digital to analog conversion and get better quality.

I hope someone has some answers I'd really appreciate some.

AFAIK, for two channel content you shouldnt be encoding to DD, rather allowing simple PCM passthrough, which should be bit-perfect, since very few audio peripheral products (none that I'm aware of that are not made by Creative Labs) resample digital output.

It's quite possible that precompressed audio being output using the hardware dolby digital encoder will sound worse than analog-out as a result of transcoding, hence you should probably turn off the DD encoding function in the driver when listening to two channel content.

Note that the NForce 3 platform does not possess DICE at all, only the NForce 2 platform does when equipped with the MCP-T southbridge. The Dolby Digital encoding happens in the MCP-T itself, before the digital signal is handed off to the CODEC, which in the case of SDPIF passthrough, does not modify the signal at all to my knowledge.

edit -- clarity

so your saying DICE is more of a gimmick than an actual improvement on a soundcard interface?

I know DD is compressed, but the fact that DTS is what, some 3x higher quality than DD but almost nobody can hear a difference even on super high end equipment must mean that the "transcoding" from 2 channel PCM audio to DD5.1 is seamless.

You make it sound like im taking a 320mp3, converting it to wav and re-encoding it but im not so sure that DD is that lossy. It'll probably seem like taking a 320MP3 and converting it to WAV then use a nearly lossless encoder and output that to the decoder for the DAC process.

It's targeted at gamers with external recievers. Positional audio through whatever expensive and esoteric sound system you might have...sounds like a pretty good idea, no? As an occassional gamer myself I can tell you that sound effects in gaming can be 48kbps Ogg Vorbis and not really spoil the experience.



The reason "no one" can hear the difference is because transcoding from lossless PCM is not that much of an issue: DD is transparent in most cases.

The problem with this scenario is it is not analagous to watching a film - using a lossy compressed course, decompressing it to PCM and then reencoding it in a different lossy format is much more likely to result in quality degradation because you're ultimately transcoding from a lossy source.

To my knowledge no film company captures their soundtrack as an mp3, then converts it to PCM, then recompresses it to DD. If such a company did exist, then perhaps we could viably compare DD to dts in terms of quality degradation (on one of their soundtracks) due to transcoding.

Also note that "quality differences" between DD and dts are going to be somewhat subjective. You can talk about the relative size of the data stream, which is not, of course, the same thing.

edited repeatedly for clarity, of course

For a decoder i have a Panasonic SA-HE100K Home Theater Receiver. Which is about 220 bucks, so I what I was really asking in my first post was weather or not the audigy2 sounds better than the digital out on my soundstorm.

It probably won't make any difference if you use digital pass-through, but it might with DD encoding enabled. It's likely the A2 has slightly better quality DACs than the reciever, though I doubt that will make a big difference.
The only way to know for sure is to compare, hopefully in a blind-testing situation so bias doesn't intrude

Honestly the biggest problems I've noticed with soundstorm systems have been when the outboard decoder is rather noisy for one reason or another (early logitech Z-680, cheap JVC reciever).

I had the z-680's but i have teh klipsch promedia 5.1's now.

the early z-680 have a hissing problem, which is probably what your describing, that's normal and if you call logitech you can get a replacement control pod with a updated firmware which removes the noise/hiss.

this is incorrect; practically all AC'97 compliant chips resample to 48kHz (see spec at intel site). This includes all nVidia Soundstorm implementations, edit2: see later postsas well as virtually any other onboard sound solution. Edit: some allow bit-perfect passthrough off non-audio streams, but these still do something to audio streams...

An exception was some CMI8738 chip, but then I'm not sure if it was AC'97 compliant, and also: it's analog performance is worse than terrible.

Edit: as to the comparison of the Audigy2 DAC with the receiver DAC, I think measurements in reviews suggest that the Audigy2 DACs are among the best. It is likely that analog out from the Audigy2 measures better, even if the audible difference isn't significant. One digital connection to the receiver is ofcourse much more convenient...

more tips: use the coaxial out if you can (unless there are ground-related problems), and like Audible! says: DICE is only convenient for gaming apps, for hifi PCM, switch to plain PCM output.

I do not believe this is accurate when referring to Digital passthrough regardless of what type of content is involved. Perhaps you have a link....

For analog it is completely accurate of course, the incoming digital signal is upsampled to 48KHz before converting to analog, provided it is not already at 48KHz.
There is no analog conversion in this case.

Edit:
In fact I believe I have found why our opinions differ on this matter. Your opinion appears to be valid for the AC97 1.X specification (pre-2000), while the 2.1 and 2.2 spec (2000) adds "variable sampling rate capability".

Check page 28 of the AC 97 2.2 specifications for some information on how variable sampling rates may be done. Also note that the current spec is 2.3.

Heck, I'll quote the part that caught my attention:

It seems somewhat clear (Intel also has a table below this section that shows setting the SLOTREQ bits to 0 or 1 allows for the sample rate registers to not be forced to 48KHz, in other words made "writable") that variable sampling rates can be supported in revision 2.2, which the ALC658 exceeds as it's a rev2.3 part.





Be very careful making sweeping statements about CODEC performance when the motherboard implementation/placement of the CODEC can cause the output quality to vary drastically.

I'm aware of the 2.3 spec; in chapter 5.10.1 you will find that the minimal spec requires only 48kHz spdif operation.
44.1kHz operation is optional, as it was in earlier versions of the spec.

Also, my opinion on the CMI8738 wasn't formed from reading on the web - I tried several implementations of it. It is cheap and allows true 44.1kHz operation, in digital too. These are rarely combined virtues.
But my no-name version, as well as the Terratec Aureon Fun, show terrible analog performance. You can get a taste of it, looking at the graphs of the Zoltrix built card over at Rightmark.org.

Note that I was reacting to this general statement only, which seemed to suggest that Creative were the only company resampling 44.1kHz streams.

In general, most AC'97 codecs support minimum requirements only.

If we're talking specifically about the ALC658: this is a rather advanced codec, with nice specs. From it's spec-page at http://www.realtek.com.tw/products/product...modelid=2003084 we see:

Forgive my brashness, but your earlier response seemed to imply that this was essentially unheard of.

Without a comprehensive listing of which products do and do not suport the variable sampling rate options that are contained within the previous specifications, it seems wholly inappropriate to suggest that the 2.3 compliant "high-end" Rtk 658 and other contemporaneous IC's do not support this function without proof.


Wasn't suggesting it was. I was suggesting that mediocre CODECs can sound terrible if badly implemented, and relatively acceptable when done properly. I have observed this firsthand with the canonical ALC650.


The rarity of the virtue is unknown without data. There is no data here to suggest it is either rare or common.

Perhaps you have a source.

edited spelling

You're right, there was a wrong statement I made. I've put up a strikethrough in that post.

I actually wasn't aware of combinations of soundstorm+alc658. I had only seen recent athlon-64 boards with this codec.

I don't understand your remark about rarity? All I meant to say was that I thought a 40EUR costing audio card, with digital in/outs and no resampling is something you don't come across all the time.

Ahhh! I thought you were refering to the AC 97 CODEC market specifically, not just
standalone solutions...




The thread starter mentioned it comes on his AN7 NF2 board Unfortunately I'll bet the board doesn't support the 7.1 output capabilities of the CODEC.


_________________________________________________________________
To summarize this extended thread: the ALC658 will pass-through nearly any sample rate you wish as a digital signal, but certain CODECs may not.
Be sure to check with the CODEC manufacturer for specifics.

To be more complete, talking of codecs, the ubiquitous ALC650 upsamples spdif, iirc. Unfortunately, the specs at the realtek site are inconclusive (just mentioning ac-97 compliance of the spdif port).
For the beefed-up ALC655 specs specifically mention only 48kHz output.

Forgive my incompleteness in every post, and half-understandable sentences... it's time to go to bed, and I will

source: http://www.realtek.com.tw/products/products1-1.aspx?lineid=5

edit: and it all started out, just because I feel sorry for Creative all the time - and thought it was unjust to have them hang for the resampling issues. It was Intel who came up with the weak AC'97 spec...

I strongly doubt that if the 655 is 48KHz out only that the 650 is not, but it's possible the 655 is actually cheaper to make.

We'd have to look up the CMI 9739(A) and 9738, but the only board I'm aware of that uses the older 8738 (Chaintech) does not implement the "soundstorm" DSP (APU) even though it features the MCP-T southbridge.

I'll make a convenient list for posterity-

NForce 2 Boards that should definitely support bit-perfect 44.1 digital output:
Chaintech 7NJL1-APOGEE: NO soundstorm/DICE, digital output header may be optional.
Abit AN7: soundstorm supported, optical in/out, 96KHz digital output supported, analog output reputedly poor quality.

Unknown:
XFX Mach4 (unknown CODEC)
All motherboards with ALC650 CODEC and digital output including:
Asus A7N8X Deluxe, Albatron KX18D ProII, DFI Ultra Infinity, DFI LANParty Ultra B, Gigabyte 7NNXP, Gigabyte 7N400 Pro2, etc.

CMI9739 data sheet found at http://www.cmedia.com.tw/doc/CMI9739%206CH...0SPEC_Ver12.pdf reports '48 kHz fixed rate sample sync' for the AC-link. I don't think it's easy to send 44.1kHz data embedded in a 48kHz stream, so probably this one is out too.

The more expensive CMI9761 and CMI9780 (supports 96kHz) have a spec saying
So even these doesn't offer 44.1kHz out.

To be more precise: you're listing boards that don't resample on SPDIF-out.

Bit-perfection is another problem. Just about all chips that are able to transmit 44.1kHz streams still come with win2k/xp drivers that mangle the stream. This is a totally different issue ofcourse.
For more info, browse the forums for 'kmixer'.

In short: this is a winnt5 component that allows for playback of multiple audio streams. the bad thing is, that it always dithers, even if only one stream is present. K
nown solutions are the use of kernel streaming, or using professional cards by Marian or RME, which have patches for it in the drivers.

That's an important point, bit-perfect isn't an accurate description unless you either bypass the kmixer or do not use a modern NT OS. Presumably the foobar2k kernel streaming output will be able to defeat signal modification in the majority of cases...

I just looked at a few Analog Devices specsheets.

To make the listing here more complete:
AD1985A, AD1981B. AD1980, and AD1888 support 44.1kHz on S/PDIF.
I didn't look at any other of their codecs, since those are rather old.

So my AN7 Realtek ALC658 isnt bad at all for digital out? if it does all those sample rates. I wonder if it's better using the soundstorm on my mobo w/o DICE (just pcm stereo) hooked up to my decoder and have my headphone amp hook up with my Grado's

It seems kinda hard to have an audigy2 zs and a soundstorm, and switch between the two without it being a pain in the ass, i wish i could do ctrl+alt+f11 for soundstorm and ctrl+alt+f12 f audigy2 for on the fly switching between music-movies/games.

Does anyone know is it possible to change the SLOTREQ function with WPCREDIT and WPCRSET with nforce boards? There is 2 audio controller device table, I played with them a little but awful lot registers..