by Sergey Samarin
08/25/2003 | 03:57 PM
Integrated sound hasn’t received much attention until recently. It simply existed, and that was all. It was one of the items in the mainboard features list and hence was viewed as such. However, the integrated audio subsystem has been steadily evolving and pushing out of the market all entry-level sound cards. Now, it is approaching even professional products.
Some prophets are already trumpeting the imminent death of all add-in sound cards as a genus, pointing at the sound solution from NVIDIA. But while NVIDIA is a direct competitor to the dominant force of the consumer market, which is of course Creative, VIA Technologies with its improved mainboard controller, Envy24PT, aspires for the professional market sector (I could remind you that the previous version of this controller – ICE1712 – appeared in a number of low-end professional cards).
Integrated sound is attacking; you can see this with a naked eye. So, this article is going to describe the combatants and pit them against each other!
In the beginning was ISA. ISA begat PCI and died soon after. With the arrival of PCI, there appeared mainboards equipped with integrated PCI sound controllers, which, however, never really caught the spotlight due to their high cost. These sound controllers settled down in add-on sound cards. Still, the very idea of integrated sound was in the air. Soon, Intel took up this idea and suggested that the CPU could be made to process audio content. As the central processor involves only half of its potential power in most tasks, it could handle the additional workload well enough.
Thus, the only standalone element left was the switch unit for connection to speaker systems and external audio sources. Intel embodied this idea in a set of standards and specifications, and it was taken by the companies that manufactured chipsets and sound codecs. The serial digital interface between the controller and the switches was named AC-Link; it is a part of the AC'97 standard. Intel managed to unify the digital part, setting it apart from the analog one. Besides the connection to sound codecs, the AC-Link interface communicates with AMR and CNR ports (VIA calls the latter ACR). However, these connectors never grew popular and are now rare guests in modern mainboards.
Intel's specification for the South Bridge (Input/Output Controller Hub as the company calls it) lists AC-Link support as an obligatory feature of all company's chipsets. VIA and SiS followed the suit. However, there is always somebody who feels oppressed in the framework of standards and specifications. This time the innovator was NVIDIA that decided to change the South Bridge and implement the audio processing unit (APU) in some chipsets of the nForce2 family. We will talk about them later in this article.
It's evident that the sound quality of the integrated audio depends on the analog part rather than digital (that is, on the AC'97 codec chip). According to the AC'97 standard, the codec consists of two functional units. The first is a codec control circuit and hardware interface to AC-Link that distributes audio data. The second is the analog part that includes mixing and amplifier circuits and supports inputs and outputs. The two units collaborate via special converters that translate signals from digital into analog format and vice versa. The set of functions every AC'97-compliant codec must have is described in detail in paragraph 2.3 of the specification. I guess there is no need to enumerate every standard function and specification item. It wouldn't be an easy read then.
It may seem that such a strict standard should make the driver universal, that is, operational with any “compliant” codec. However, some of the specifications describe numerous requirements, which are simply recommended for further implementation. So, it is the codec manufacturer that decides which recommended functions should be included. That's the main reason why we see several AC'97 codecs available in the market. For example, the codec chip may include the amplifier cascade, frequency filter unit (to control timbres), stereo-base enhancement unit (3D Stereo Enhancement), digital (S/PDIF) output support units and so on. According to the AC’97 standard, the codec is controlled by user commands entered along the GPIO bus (General Purpose Input/Output). This command-entry ability implies that you can vary codec settings from a software control panel.
Any codec must support codec cascading. For example, some devices achieve multi-channel support by coupling several stereo-codecs. Each codec is assigned a unique ID to “know” its audio content. All expansion cards with rear satellites and subwoofer connectors are made in this way.
Note that mainboards don’t practically please us with interesting and highly functional codecs. The major determinative for each manufacturer is very often the cost of the codec.
Well, if there are popular brands of cigarettes or cars, there should be popular codecs, don’t you think so? Having studied the audio insides of many contemporary mainboards, I arrive at a conclusion that the today’s top codec is ALC650 from Realtek. This is the chip used by an overwhelming majority of mainboard makers.
Realtek ALC650 is the most popular codec
ALC650 is an 18-bit full-duplex six-channel AC’97 codec developed specifically for PC-based multimedia systems. You may think this presentation rather suggests not very high sound quality of the chip, its orientation at the low-end market sector. That’s not quite right, as the today’s low-end differs dramatically from what we saw just yesterday. Just look: the specified signal/noise ratio is 90dB, which is not bad for a “mass-market” codec (you can see the spectrograms for ALC650 in the “Performance” section). It supports three pairs of audio outputs with independent volume control. The codec is equipped with its own amplifier (50mV/20Ohm) for headphones, which shortens the list of electronic components and, accordingly, the overall mainboard cost. Of course, the integration of the amplifier into a single DIP-package is only profitable from the manufacturer’s point of view.
Among negative points of this approach, I can name extra heat dissipation and worse sound quality because of the reciprocal EMI produced by the elements inside the package. However, ALC650 has two weighty advantages from the mainboard makers’ point of view: low cost and compatibility with an infinite list of chipsets. The codec works with Intel 810/815/820/845 as well as chipsets from VIA/SiS/ALi. Now, let’s discuss its features in detail.
Basic characteristics of Realtek ALC650:
The Realtek ALC650 codec is the strongest product in the market. For some other chip to achieve the same or higher popularity, it has to offer more functions and higher sound quality for less money. Which is quite implausible. However, new codecs do come out, while the competition issues are resolved by using new chips in the mainboards developed by codec makers themselves. A glaring example is VIA Technologies.
People at VIA have already realized that ordinary users use their computers for entertainment rather than text typing. The growing capabilities of PCs in audio recording and playback, in special surround effects for games, etc. pushed VIA to create its own audio division. This company, known mostly for its chipsets, has released so many audio-related projects in the last two years, that it would suffice for a lifetime of any other establishment. So, VIA bought up ICEnsemble, a sound controllers developer, and founded a special department to be engaged in audio projects.
For the VIA Apollo KT400 chipset, the company developed a combination of software/hardware audio solutions under the name of Vinyl Audio (“Vinyl Audio Six-TRAC” for VT1616 and “Vinyl Audio” for VT1616A). The new South Bridge, VIA VT8235, along with the full communication package, also includes six-channel Surround Sound AC’97 support via the digital AC-Link interface. VT8235 doesn’t have its own APU as NVIDIA’s nForce2 and thus loses to it in many service capabilities. Moreover, the lack of hardware audio processing results in overall system slowdown.
VT8235 South Bridge is the major element
of VIA’s “audio mosaics”
Any South Bridge is just a digital part of the integrated audio subsystem. The analog components, on the other side of the Bridge, are responsible for the sound quality. Any manufacturer wants to use his own electronic chips, so VIA developed an AC’97 codec aka VT1616. It completes the chain from KT400 and VT8235 to the audio inputs/outputs. The codec includes both: an ADC and DAC to compete successfully with the hero of our time, Realtek’s ALC650.
Basic characteristics of VT1616:
VIA VT1616 audio codec
VT1616 can process six audio channels at a time and with 18-bit definition. VIA specifies a SNR of 97dB for the codec, if it is used as a component of an add-in sound card (in this case, VT1616 is supposed to work together with the VIA Envy24 audio controller, which we have already discussed in detail in our previous articles). In the “noisy” environment of the mainboard, the SNR reaches only 90dB (although it is quite good for a low-end codec). To make the signal purer and reduce the heat dissipation, VT1616 doesn’t include a headphones amplifier. This allowed VIA marketing people to call this architecture “CoolAmp”.
Simplified flow-chart for VIA VT1616 audio codec
VT1616 can perform hardware channel down-mixing, thus taking this workload from the CPU. Again, marketing people at VIA don’t play lazy and invent names for everything. I checked this out: this technology is called DualMax (humble, and rather vague). In simple words, DualMax is a volume sound technology: multi-channel content is virtualized to be sent to headphones or a pair of speakers, which is important for a gamer, in the first place. It should be acknowledged that this function is implemented on the software level in most standard audio codecs. The manufacturer claims that the hardware approach is better: the sound is purer and more saturated, and not at the expense of quality. Moreover, VIA vows the mainboards with VT1616 can produce a higher-quality sound than those with nForce2 and mass codecs. They don’t say explicitly what mass codecs they mean, but it must be Realtek ALC650, I assume.
Although the CPU took up nearly all sound processing tasks with the advent of integrated audio, they never stopped working on an auxiliary sound processor. It’s all because the CPU turned to be not fast enough for such tasks as encoding Dolby Digital streams, three-dimensional sound positioning with math1ematics-heavy algorithms and so on. It was especially evident during DVD playback. NVIDIA was quick to come up with a solution: an audio processor unit (APU), integrated into the South Bridge of the nForce chipset. For today, NVIDIA’s APU is a worthy rival to the Creative Audigy2 sound card and even surpasses it in some functions.
NVIDIA’s supply met the demand from Microsoft. The corporation chose nForce APU for its X-box console. It’s not just good luck, since the sound processor has a powerful hardware part and supports numerous software interfaces used in modern games: DirectSound, DS3D, EAX 1.0 and 2.0, I3DL2, ASIO and OpenAL. As you see, only EAX 3.0 (EAX Advanced HD) support is missing, but this proprietary API is only available for the Audigy2 card today.
Unfortunately, nForce-based mainboards only work with AMD processors. My question to an NVIDIA representative about NVIDIA’s plans concerning any nForce chipsets for Intel CPUs received the following answer: “NVIDIA is considering this, but not for the near future”.
It should be noted that not all nForce chipsets feature an integrated APU. In NVIDIA’s terminology, the chip that processes audio/visual content and is responsible for communication is called “Media and Communication Processor” (MCP). So, the South Bridge, nForce MCP, with the APU and SoundStorm units, is bridging the gap between computer games and reality. As a part of the chipset, the APU uses five internal processors for work with audio streams. Such complex hardware solutions are evidently intended for hardcore games and multimedia fans. The NVIDIA website has the following table listing the main features of the NVIDIA APU and SoundStorm (I put it down as it is, with comments following):
NVIDIA nForce/nForce2 APU
DSP or HW-Accelerated 2D (Stereo) Voices
DSP or HW-Accelerated 3D Voices
64 3D Voices
64 3D Voices
Per Voice Parametric EQ**
Occlusion and Obstruction
Near Field Effects
EAX2 and I3DL2 Reverb
Global Effects (Reverb, Chorus, etc.)
2, 4, or 6 Speakers
Software and Utilities
H/W Dolby Digital Encode
* - The submixer is a unit to mix up several voices into one. The result can be sent to the voice processor for 3D-positioning and applying special effects;
** - The per-voice parametric equalizer is embodied in hardware. It means that with DS3D or DLS, the audio processor can correct (compensate) the frequency characteristic of each of the processed voices (many EAX 2 functions are implemented this way, although not all of them in hardware).
So, as I have said, some versions of the South Bridge (MCP) don’t have the APU (although all versions resemble each other). Today, there are three modifications:
nForce2 MCP-T South Bridge
The distinguishing feature of these three modifications is their price. The price is the sum of all license fees for the involved technologies. Thus, Dolby required licensing. Network technologies (3Com Ethernet and FireWire) have their cost, too, and they make the MCP2-T modification more expensive. It’s clear that a mainboard based on the basic, “barebone” nForce MCP chipset would be a standard product with an AC’97 codec for sound output and the CPU for sound processing. But the price is all-important, so many mainboard makers prefer to use this simplest MCP variant. Other functions are usually available in “deluxe” mainboard versions, like ASUS A7N8X. The package reads that the product supports Dolby Digital, Dual LAN, Serial ATA and IEEE1394, while the standard version lists them as “optional”.
The Deluxe version of ASUS A7N8X
ASUS A7N8X back panel
The audio subsystem of A7N8X is a mosaic of an APU, SoundStorm/Dolby Digital Encoder and six-channel Realtek ALC650 codec. Buying the mainboard with SoundStorm, the user will find the following additional connectors at the back panel:
As you see, the A7N8X mainboard carries all of them at its own back panel. There are also mainboards that come with brackets for the extra audio connectors.
Now, let’s try to make it clear what the audio processor is on the hardware level. Here is the APU flowchart:
So, the nForce2 APU consists of four sections:
I guess the purpose of at least three sections is evident. Let’s dwell on the last one aka DICE. Encoding into the Dolby Digital format is no mystery, in fact. All film studios do it when recording soundtracks for movies we watch in cinema houses or in the home DVD-theater. However, software processing of this content as performed by various players consumes a lot of system resources. Moreover, Dolby Digital audio should be “pre-prepared” to be reproduced along with the video. When CPU resources are insufficient, you will have the sound lagging behind the picture: that’s the tradeoff of the software processing. So, that’s where the word “interactive” in DICE comes up. Roughly speaking, this APU section is responsible for fetching “on the fly” the digital stream during DVD playback, converting it into Dolby Digital 5.1 format and outputting through the digital S/PDIF output onto an external hardware decoder. Here, the APU from NVIDIA is much more powerful than Audigy2 from Creative. The audio card cannot yet decode Dolby Digital in real time and only pre-encodes this digital content.
We may assume that such a multi-functional integrated audio-subsystem from NVIDIA should have a flexible and detailed control panel. The manufacturer tried to make it up to the mark, by taking advantage of the positive experience other companies had collected in this field. The Control Panel is a well-done product; it is informative, offers fine-tuning options to set up the system according to your own needs.
The Control Panel window is divided into six tabs.
nForce control panel
The Main tab consists of three sections (Speaker Output Levels, Equalizer, Input and Output) and carries the basic sliders to set up the signal levels, equalizer parameters and so on. Note that the sliders don’t set up sound volumes, but change the amplifier coefficient (frequency characteristic) from 0dB (lowest position) to infinity (highest position). As a nice addition, there is a per-channel volume indicator that tells whether there is any signal at all. It shows peak levels in orange and overloads – in red. The equalizer allows loading any preset from a long list.
Speaker Setup window
The Speaker Setup tab allows configuring the type of your speaker system and choosing analog or digital output. It’s here that you set 3D sound as well as volumes for each channel. Besides, you can test every satellite of the system by turning on the test signal.
The options gathered in the Surround Settings tab require some comments:
MIDI sound setup window
The MIDI tab is…well, for MIDI content. The Downloadable Sounds section allows loading/unloading DLS and SoundFont sample banks. Any changes can be listened to right here, on a test MIDI-file. The drop-down menu below lists surround sound presets and allows customizing effects. This section is nearly the same as the one on the Environment tab (see below).
As for the sheer number of options, the Environment tab is the most powerful one in the whole Control Panel. You can immediately check out any changes you make using the sample file. The drop-down menu lists numerous presets for you to choose. There are three tabs at the bottom of the window: Routing, Source Levels and Parameter Editor. Here, you can also create and save your own preset using the following effects: Chorus, Compressor, Distortion, Echo, Flanger, Gargle and ParamEq.
Source Levels: the tab where you can set volume levels for different sound sources
Parameter Editor tab for selected effects
The Parameter Editor is for advanced tweakers. The user receives access to hardware registers, DSP and mixer settings as well as to the memory allocated for the audio processor. Any changes are applied in real time, so you can set up the audio subsystem by ear.
Here you can select external applications for further work with the sound
The Application tab stores shortcuts to various multimedia applications to quick-launch them. You can add new shortcuts or remove unwanted ones.
The Information tab displays the version of the installed driver and Control Panel, and the DirectX version. The advanced info section informs you about the current APU workload and the number of processed voices. So, here you control the driver installation and learn what your APU is doing down in the hardware level.
As you see, the Control Panel for the nForce APU provides various settings, which only testify the hardware power of the audio processor. In the next, “Performance” section of this review, I will try to compare the work of the computer in resource-hungry applications with and without the APU involved. You want to know what difference it brings? Read on!
Two mainboard are going to participate in today’s testing: Soltek KT400A and ASUS A7N8X. The first mainboard features a VIA VT8235 South Bridge and an AC’97 VIA VT1616 codec. The second one is based on the nForce2 MCP-T with the integrated APU SoundStorm and an AC’97 Realtek ALC650 codec. Of course, the mainboards are based on different chipsets, and we cannot compare them directly. So, I tested an add-on Creative Audigy2 sound card to be a reference-point for our analysis. Here are the testbed configurations:
The Comanche4 Demo game test is quite hungry for system resources. It gives out very precise results (the second run produces just a one-tenth fps deviation). So, I ran the test with and without sound. The first test round reflects the overall system performance, while the second helps to evaluate the slowdown you receive when sound processing is enabled.
In the diagram below you can see how well the audio processors of the integrated sound subsystem and the standalone sound card work. The graph tells that NVIDIA’s APU and Creative Audigy2 are peers in terms of hardware power. The performance drop is negligible. Once again, the results were taken on ASUS A7N8X mainboard.
The lack of an audio processor in the VT8235 South Bridge resulted in overall system slowdown. This time, an add-on sound card with a fast DSP onboard is quicker than VIA’s solution (4 fps in Commanche4 Demo is quite a lot).
3DMark03 offers an air battle scene and counts up average fps rate. Just like with Comanche4 Demo, we witness a noticeable performance drop when we enable the sound.
Unfortunately, 3DMark 2003 found the audio subsystem from VIA unable to reproduce 60 sound sources at a time, but the general tendency is clear enough.
The tests we perform with SpectraLAB version 4.32.17 follow the methodology explained in our previous articles. Here, we are comparing the analog parts of the two audio subsystems. It is Realtek ALC650 in ASUS A7N8X, and VIA VT1616 in Soltek KT 400A.
First come Realtek ALC650 spectrograms:
44,100Hz/16bit, 1kHz tone signal
Determining IMD at 44,100Hz/16bit
48,000Hz/16bit, 1kHz tone signal
Determining IMD at 48,000Hz/16bit
Now let’s have a look at VIA VT1616 spectrograms:
44,100Hz/16bit, 1kHz tone signal
Determining IMD at 44,100Hz/16bit
48,000Hz/16bit, 1kHz tone signal
Determining IMD at 48,000Hz/16bit
The ALC650 codec showed the claimed signal-to-noise ratio, while VT1616 did not. However, we should keep in mind the fact that other components of the analog section also affect the overall result. Of course, the mainboard maker is the one responsible for proper integration of an AC’97 codec, which implies that they have to choose proper electronic components, design wiring layout, make sure they meet the codec maker’s recommendations and so on. Thus, the same codec may sound differently on two different mainboards: in test programs and by ear. I often came across mainboards where the working hard disks or even mouse movements produced noise in the integrated audio. So, the results you have just seen are only typical of these two mainboards.
As our testing confirms, hardware implementation of certain audio subsystem functions helps to reduce the CPU workload. As for sound quality, many add-on sound cards are still better than integrated audio solutions. The reason is simple: mainboard makers try to reduce the cost at any rate. They choose cheap codecs, and don’t care about proper screening of audio components. That’s why a user may consider purchasing a sound card to be a better decision than using the integrated audio.
Today, everything is being integrated: graphics and audio subsystems, network, FireWire, Serial ATA. This approach has its advantages: you don’t need to install a separate driver for each of the system components. A single, “all-in-one”, unified driver will be enough. Integrated solutions are good for many users, but not for all of them. The integrated audio lacks many functions and often offers not the best sound quality. Will you put up with it?