The State of Motherboard Audio
Dave Salvator
There was a time when the very thought of motherboard-down audio would give any self-respecting techie the willies.
High CPU overhead, low signal quality and poor game support topped a long list of shortcomings that kept motherboard-based audio solutions on the cheapest of desktop PCs. But in the mobile space, motherboard-down has been the standard on laptops for as long as there have been laptops.
Today, CPUs have plenty of cycles looking for something useful to do, and thanks to improvements in both codec designs and motherboard layouts, motherboard audio can be implemented in a way that even enthusiasts might consider using. Unfortunately, despite strong performance and good signal quality, a poor motherboard layout can doom an audio processor to mediocrity, or worse.
How far has motherboard-down audio come? We decided to find out. We tested five different solutions, and, came away mostly disappointed. Bad implementations, we found, can ruin even good solutions. It wasn’t all bad news, though. We found some pleasant surprises along the way, and some very encouraging trends. Read on to find out more.
The list of hardware makers with integrated audio solutions includes several obscure companies. The short list includes:
Analog Devices SoundMAX
ATI
C-Media
Crystal
ESS
nVidia
Realtek
SigmaTel
Via/IC Ensemble
Among the companies in this list, two (ESS and Crystal) deal almost exclusively in the area of laptop audio. ATI is a newcomer with its Radeon 9100 IGP chipset for the Pentium 4. Both Analog Devices and SigmaTel make codecs that are used on Intel motherboards. nVidia has what is perhaps the most robust motherboard-down audio solution on the market in its MCP-T audio sub-system, which includes three DSP cores to handle audio processing. Realtek, Via and C-Media make motherboard-down audio codecs that appear on a variety of Taiwanese motherboards.
For our story, we looked at the following audio solutions:
Company
Audio Solution
Motherboard
Analog Devices
SoundMAX
Intel 865GBF
ATI
Radeon 9100 IGP
ATI Reference Board
C-Media
8738 MX
Via P4PB Ultra
nVidia
nForce2
MSI MS-6750G
Realtek/AOpen
RTL2100BL
AOpen AX4GE Virtuoso Tube motherboard
Via/IC Ensemble
Envy24 PT
Albatron Intel 86 SEPro II
We focused our testing in three areas:
CPU usage
Audio signal quality
Game audio performance
The challenge was trying to keep the tests balanced between AMD and Intel based systems, since AMD uses a PR rating rather than actual clock rates for its CPUs. Further complicating matters was the fact that some AMD chipsets don’t yet support the 400MHz frontside bus (FSB).
To evaluate CPU usage, we used Ziff-Davis’ Audio WinBench, focusing on the tests that play 32 CD-quality (44KHz/16-bit/stereo) sounds simultaneously. These tests use both static and streaming DirectSound buffers. We run these tests first on DirectSound, and then on DirectSound3D.
For audio signal quality, we used our standard SoundForge noise-floor test, at resolutions of both 44KHz/16-bit/stereo and 96KHz/24-bit/stereo.
We used the Line-In as the recording source, using each audio processor’s mixer panel. Unfortunately, Windows’ standard mixer panels are notoriously vague, and don’t allow any real discernible gain adjustment by a set amount (linear percentage or dB), so we’re left to approximate as best we can.
As it turns out, there isn’t a firm definition by Microsoft as to how Windows mixer gain settings are supposed to behave, which accounts for the inconsistency you see from one sound card to the next. And appallingly few sound cards actually present you with dB markers underneath the mixer faders (like you have on a physical mixing board), or give any indicator when they’re at 0dB. All the motherboard audio processors we tested also lack these dB markers, although Analog Devices SoundMAX gives you a linear percentage in its mixer app.
To run this test, we removed all connections from the sound device, both inbound (signal generator, other audio devices) and outbound (speakers) to get a reading of the card’s noise floor. We then recorded ten seconds of silence, and then brought up SoundForge’s Statistics window to read the RMS Power Value, which is the average signal energy value present throughout the ten-second recording.
This test measures “what’s there when nothing’s there,” or the noise the device itself generates when sitting still.
The theoretical maximum signal-to-noise ratio (SNR) of a 16-bit audio device is about 96dB, since each additional bit used for sampling buys you about 6dB. Audio CD players have this same theoretical limit, since each sample contains 16 bits. So a 20-bit device has a theoretical SNR of 120dB, and a 24-bit device, the new high-resolution sound quality found in DVD-A players would be 144dB. The key word to ponder when mulling over those numbers is theoretical, since audio devices often don’t get to theoretical, and those that do usually perform some kind of oversampling/downsampling algorithm (a kind of antialiasing actually) to get there.
We also ran the RightMark Audio Analyzer 4.3 on all of these solutions to gauge audio signal quality.
And finally we ran three games, Comanche 4, Jedi Knight II and Unreal Tournament 2003, at 640x480x16. We did a baseline test run with audio disabled, and then re-ran the same test with audio enabled. The idea was to measure the percentage drop in frame rate that was due to audio being processed and see if any of the audio processors fared better than others.
We also tried to run 3DMark03’s sound card test, but nearly all of these motherboard-down audio solutions report zero hardware audio buffers to DirectSound, and 3DMark’s test won’t run unless hardware audio buffers are present. This is an unfortunate limitation, as we’ll undoubtedly see more host-based audio solutions in the future that will robustly support multichannel audio in games.
For comparative purposes, we also put a Sound Blaster Audigy 2 through the same paces.
The test systems for each audio solution were as follows:
Motherboard
CPU Used
System Chipset
Audio Solution
ATI R9100 IGP Reference
3.06 GHz P4
R9100 IGP
ATI Digital Audio
Intel 865GBF
3 GHz P4
865G
AD SoundMAX
MSi MS-6750G
AthlonXP 2700
Nforce2
Nvidia MCP-T
Albatron Intel 86 SEPro II
3 GHz P4
865G
Via/IC Ensemble Envy24-HT
AOpen TubeSound 533
3 GHz P4
Intel 845GE
Realtek 8201BL
Via P4PB
3 GHz P4
Via P4X400
C-Media 8738
Audigy2
3 GHz P4
865
Audigy2
Constants across all systems were:
Memory
512MB PC3200 DDR SDRAM
Check Prices
Hard Drive
Western Digital WEB1200 120 GB ATA-100 EIDE
Check Prices
DVD-ROM
Toshiba DVD-ROM
Check Prices
Operating System
Fresh install of Windows XP Pro Service Pack 1 with latest system updates as of 6/20/03, and DirectX 9.0a
Check Prices
PAGE
What does a motherboard-down audio processor need to do? Simple, everything a consumer-focused discrete PCI sound card like an Audigy 2 or Fortissimo III can do. Here are the chores any PC audio solution has to be able to handle:
Must-Haves
Wave/MME playback and recording in Windows
Multichannel (4.1/5.1) analog output
S/PDIF output (coax or optical, though having both is optimal)
DirectSound support
DirectSound3D support
A3D 1.0 support (for legacy games)
EAX 1.0 support (mostly for games)
Cool Extras
EAX 2.0 support (for gaming)
OpenAL support (mostly for games)
Customized mixer application
ASIO/ASIO2 support (a low-latency driver standard used for studio recording applications and developed by Steinberg for its MIDI sequencing /hard-disk recording and apps.)
Many of the audio solutions we tested can support all of these features, but a motherboard maker may omit some of them to lower the total cost. As a result, some of an audio processor’s more interesting features are never available to the end-user. But many PC and motherboard makers now realize that the road to the living room requires solid multichannel audio, and they realize that this feature is a key to bridging the PC and CE worlds.
What Winds up on the Motherboard
Essentially, any audio solution consists of two major parts: the DSP (digital signal processor) and the codec (A/D and D/A converter). A “traditional” PCI sound card puts both of these items on the card, and these two processors then handle all audio-related tasks. Some sound cards offload some DSP tasks to the CPU, delivering only part of a full hardware.
Most motherboard-down audio solutions are essentially codecs, with the rest of the audio processing implemented in software and executed by the CPU. Interestingly, PCI transaction overhead can cause a discrete sound card with a poorly written driver to introduce more CPU overhead than some motherboard-based audio implementations. However, as we’re about to see, that isn’t the case with Creative’s Sound Blaster Audigy 2.
Of the solutions we evaluated here, only the nVidia nForce2’s MCP-T audio solution has an actual DSP in hardware. As a matter of fact, it has three. All the other solutions we looked at rely on software and the CPU for their DSP work. So for motherboard-down audio, the software takes on a key role, since it is virtualizing essential audio services and running them on the CPU.
Two of the audio solutions do a very good job minimizing CPU usage – the exceptions are Analog Devices’ SoundMAX, the AOpen TubeSound and the C-Media 8738. These three all caused lots of CPU overhead during wave playback tests. In the case of SoundMax, we reran the tests, and disabled some of its features, like microphone noise reduction, to verify that no extraneous audio services were impacting the test results — but the results were the same.
This was the only major knock on the SoundMAX solution, as it fared well on the rest of our tests. The same cannot be said however for the C-Media 8378, whose overall performance ranged from fair to poor in other testing, and its high CPU overhead is just another strike against it.
Creative’s Audigy 2, our comparison audio device for this round of testing, does very well here, using very little of the CPU on the DirectSound3D tests, and none of the CPU on the DirectSound tests. To Via’s credit, its solution performs admirably here as well, never poking its CPU usage above 1.5%.
nVidia does well here, although we would have expected it to be in the sub-1% range as well, given its plentiful built-in DSP horsepower. That said, as long as CPU overhead remains below 5%, we believe the potential performance hit in a gaming situation is imperceptible. And to be fair, nVidia was tested on an AthlonXP 2700+ whose actual clock rate is 2.16GHz, well below the 3GHz that the Pentium 4s were running. When all is said and done, nForce2’s audio sub-system delivers low enough CPU overhead, and its alluring feature set also makes it a chipset to consider.
We used RightMark’s Audio Analyzer (RMAA) to try and dial in the cleanest gain structure we could for all the products tested. We primarily looked at the real-time Fast Fourier Transform (FFT) that RMAA shows when adjusting I/O levels. What we tried to achieve was an FFT plot that looked like this:
As opposed to something like this:
Note the spikes appearing above the 1KHz test tone (called the fundamental) that are occurring at odd- and even-order harmonics. The goal is to “flatten” as much of the frequency band (above 1KHz, up to 22KHz), removing spikes. Establishing a clean signal path is something of an arbitrary process in Windows, unfortunately, since sound cards and motherboard audio implementations vary quite a bit.
We used the line-in input as the recording source for all the systems, and muted all other inputs, except for the wave and master faders. We began with the RMAA gain structure settings, and then adjusted them to achieve the necessary signal levels to run the tests, along with trying to produce the cleanest-looking FFT we could.
The Sound Blaster Audigy2 sets the pace here, turning in a fairly solid -84.5dBFS noise floor.
Among the contestants, SoundMAX is the standout, and our close listening tests align with this test result – we could hear no hash, blitter noise, or “zipper” effect when we moved mixer faders up and down. The nForce2 ran into trouble here, with one channel in the 50dB range, while the other was up around 38dB. The rest of the herd doesn’t do especially well here, and the Via Envy24 is especially noisy when idle, which was audible during close listening tests.
We sometimes see improvement when running at the higher 96KHz/24-bit resolution, however here most audio solutions didn’t see much appreciable gain, and several actually lost some signal quality at the higher resolution. Audigy2 was very clean on this test, coming in at –105dB, whereas the next nearest part, the SoundMAX was in the –70’s, similar to its noise floor at the 44KHz/16-bit resolution. Again, the rest of the field doesn’t do well here, and we found fairly high at-idle noise floors on these other parts.
All told, none of the integrated solutions are anywhere new the Audigy2 for providing quiet signal environment with the audio sub-system at idle. However, as we’ll see next, the integrated solutions do fare better on RightMark Audio Analyzer dynamic range and noise tests. What this says to us is that noise floor may only be present when the system is at idle, but this noise-floor drops significantly when signal is present.
We test each audio solution using the line-in input as the recording source, and muted all other inputs, except for the wave and master faders.
These tests contradict the at-idle noise floor test results we just saw, and what we believe is happening is that because the audio codecs are on the motherboard, they are much more susceptible to noise from other motherboard components.
However when signal is present, these motherboard-down solutions are able to achieve considerably better dynamic range performance. In fact, the Via Envy24 scored better at both resolutions than the Sound Blaster Audigy 2. The main take-away here is that all of the solutions provided at least fair performance on this test, although the only one that stands out is the Via Envy24.
Here are the rest of the data points, along with which audio solution turned in the best performance. Notice that in some instances, the winner was ahead only by a single dB.
44KHz/16-bit
Aopen TubeSound
AD SoundMAX
Nforce2
Via Envy24
C-Media
Audigy2
Best Performer
Noise level, dB (A):
-78.2
-79.1
-73
-87.6
-74.8
-84.9
Via
Dynamic range, dB (A):
75.5
77
69.6
84.8
69.8
82.9
Via
THD, %:
0.244
0.0096
0.011
0.02
0.37
0.0035
Audigy2
IMD, %:
1.251
0.065
0.11
0.075
17.29
9.384
SoundMAX
Stereo Crosstalk, dB:
-0.9
-69.7
-60.5
-0.3
-69
-75.8
Audigy2
96KHz/24-bit
Aopen TubeSound
AD SoundMAX
Nforce2
Via Envy24
C-Media
Audigy2
Best Performer
Noise level, dB (A):
-77.4
-73
-75.6
-87.8
-75.1
-83.4
Via Envy24
Dynamic range, dB (A):
75
69.6
73.2
84.8
69.4
82.2
Via Envy24
THD, %:
0.282
0.011
0.011
0.02
1.43
0.0045
Audigy2
IMD, %:
32.371
0.11
0.091
0.094
39.57
0.061
Audigy2
Stereo Crosstalk, dB:
-0.9
-60.5
-65.1
-0.3
-66.8
-67
Audigy2/C-Media
Noise Level
Interestingly, the Via Envy24 bests the Audigy2 here by nearly 3dB. That a motherboard audio part can hit close to -90dB on a test like this is impressive. Actually, all of the parts do at least fairly well on this test, with the nForce2 being the noisiest at -73dB.
We see a repeat performance here at 96KHz/24-bit, with Via atop the pile, and the other integrated solutions in the -70’s, a level of performance that’s not terrible, but not especially impressive either. SoundMAX does well at 44KHz/16-bit, but takes a pretty considerable hit at this higher testing resolution dropping 6dB, meaning its performance was only half as good at 96KHz/24-bit.
Dynamic Range
Again, Envy24 is at the head of the pack, and ahead of Audigy2 at both test resolutions. And again, Via Envy24 is within spitting distance of -90dB, an impressive feat. SoundMAX exhibits a similar behavior as seen in the Noise Level test, doing pretty well at 44KHz/16-bit, but then taking a pretty severe hit at 96KHz/24-bit. It’s worth noting that the AOpen TubeSound performs decently on these tests, although tube-based devices often don’t bench especially well against transistor-based systems.
THD
Here it’s all Audigy2. The only integrated part that’s even at the same order of magnitude as Audigy2 is the SoundMAX, and that’s only at 44KHz/16-bit. The C-Media part does especially poorly here, and goes from really bad to even worse at 96KHz/24-bit. nForce2 does well here at both resolutions, as does SoundMAX, although it fares better at 44KHz/16-bit. As one would expect, the AOpen TubeSound doesn’t fare well in this test, which is generally the case for tube-based audio circuitry on THD tests.
Intermodulation Distortion (IMD)
An audio measurement designed to quantify the distortion products produced by nonlinearities in the unit under test that cause complex waves to produce beat frequencies, i.e., sum and difference products not harmonically related to the fundamentals.
Source: Rane Pro Audio Reference
Audigy2 runs into trouble here, however we believe this result is inaccurate. If the Audigy2’s IMD level was really this high, it would be plainly audible when playing anything back on the card and that clearly is not the case. SoundMAX gets the win here, although the Via Envy24 is hot on its heels. C-Media goes from bad to downright awful on this test, and does so at both test resolutions. AOpen also does poorly here, but again, because of the tube circuit section, we expected sub-standard performance on this test.
Stereo Crosstalk
Any phenomenon by which a signal transmitted on one circuit or channel of a transmission system creates an undesired effect in another circuit or channel.
Source: Rane Pro Audio Reference
Interestingly, C-Media goes from zero to hero on this test, tying the SoundMAX for best performance among the integrated parts. However, both trail the Audigy2 at 44KHz/16-bit. At 96KHz/24-bit however, C-Media catches up to Audigy2, tying it for the top spot. nForce2 fares pretty well here at both test resolutions. Envy24 and AOpen fare very poorly here, and these scores indicate that there is no channel separation whatsoever. The result is that these parts will be unable to produce a correct stereo image, since each channel bleeds so heavily into the other.
We applied the test methodology to three games from our own 3D GameGauge suite of tests. We used Comanche 4, Jedi Knight II and UT2003, first running each without audio, then re-running the same test with audio enabled to look at the frame-rate drop. We ran all of these tests at 640x480x16 with a Radeon 9800 Pro installed to get fill-rate out of the equation. We did however rerun the Jedi Knight II test at 1024x768x32 to measure the hit induced by audio when running at a more realistic gaming resolution.
The two parts that showed the greatest frame-rate slowdown (percentage-wise) are the C-Media and nForce2 solutions. The latter’s result has to be taken with a grain of salt however, owing to it being an AMD platform, and one with a clock rate that’s about 33% slower than the Intel-based competition. Of the other parts, the Envy24 actually incurred less of a performance hit than the Audigy2, although not by much.
All products tested here take fairly sizeable frame-rate hits in terms of percentage, but bear in mind that in all instances, frame rates are over 100fps, this difference is for all intents and purposes academic. The field is more compressed here than the Comanche 4 test, with C-Media and nVidia again taking the biggest percentage hits.
To sanity-check our results, we dialed the resolution up to 1024x768x32 to see how this change would affect the percentage drop in frame rate. There were only small changes in the percentage drops between the two resolutions, with the likely cause being the fact that the 10×7 resolution barely slowed down the Radeon 9800 Pro, either with audio enabled or disabled.
C-Media takes the biggest hit here, followed by SoundMAX, although SoundMAX is less than one percentage point above the group average. Audigy2 suffered the smallest frame-rate hit, followed closely by the AOpen TubeSound/Realtek audio solution. When running at higher pixel resolutions, some of this performance hit will be masked by the game being fill-rate bound, but it does indicate that the additional CPU overhead that C-Media introduces during game-play could cost you too many frames-per-second to be part of a useful gaming audio solution.
In conducting our close listening tests, we used a set of Shure’s new EC-2 earphones, which are consumer offshoots of Shure’s line of in-ear monitoring systems for musicians. These listening tests in some cases picked up artifacts that might not be audible in more typical situations, but still, it’s good to know what’s there. We were listening primarily for three things while the system was idle:
Blitter Noise: Caused by the graphics card’s blitter, and can be heard when windows are moved around on the desktop, and when windows are opened or closed.
Mixer Zipper: Occurs when changing mixer fader settings, and you can hear each degree of change (remember this is digital), and it usually occurs on systems that also allowed blitter noise to enter into the audio system.
Hash: Hissing noise emitted by an audio system when sitting idle. None of the motherboard audio solutions we looked at exhibited this problem, even with all faders wide open.
Blitter Noise
Mixer Zipper
Hash
EQ?
Comments
ATI RS300
No
AD SoundMAX
None
None
None
No
Very clean overall
Nforce2
Some
Yes
None
No
Harsh when channels open, especially with Mic. With channels closed, no hash
Via Envy24-HT
Small amount
Small amount
None
No
Has booth coax and optical SPDIF
C-Media 8738
None
Small amount in master
None
No
AOpen Tube/Realtek
Very noticeable, even mouse cursor movement created audible noise
Very noticeable
Small amount
Yes
Despite noise presence, musical material did sound very good.
Audigy2
None
None
Small amount
Yes
Hash present only when master mixer fader is full-open. At 75%, not audible
In our close listening tests, SoundMAX fared very well, as did C-Media oddly enough. Despite some of the terrible numbers we saw in our RMAA testing, C-Media was pretty clean during these subjective tests. nForce doesn’t do especially well here, but by far the noisiest part in these listening tests was the AOpen TubeSound. Appreciate that with these tests we’re doing close listening at all volume ranges, including some fairly extreme ones that are rarely used. However, the goal is to ferret out whatever noise may be lurking in the sonic shadows, since the goal here is to have as clean a system as possible.
Recently motherboard audio hardware makers (and motherboard makers themselves) have been differentiating themselves in a crowded field by adding audio features. Most of these features take the form of interesting bundled software, or specialized audio apps, but one motherboard maker, AOpen, has taken a very different approach.
AOpen Details: Let’s Get Tubular
By far the most exotic motherboard audio solution in this roundup belongs to Aopen, with its TubeSound Virtuoso Platform tube audio section. Despite over 50 years of solid state transistor technology, audiophiles remain dedicated to tube amplifiers, and high-end tube amps can easily climb into the tens of thousands of dollars. Most cite the “warmth” and “body” of the tube pre-amps and amps as the reason they prefer the old-school technology. Empirical signal quality measurements generally show that tube amps exhibit higher amounts of Total Harmonic Distortion (THD+N), but some audiophiles cite this as what gives tube amps their character.
What Aopen delivers for around $160 is an Intel 845-based motherboard that includes a Realtek 8100BL codec coupled to a tube pre-amp section that uses a Sovtek tube. The audio sub-system also includes a second Realtek codec that outputs straight to an 1/8″ stereo jack that’s typically found on most motherboard audio implementations. Also in the box was a back-panel containing S/PDIF in and out in both optical and coax flavors.
Also included in the box is a separate manual just for the TubeSound sub-system, including a list of other compatible tubes that can be used on this motherboard. As with the other audio offerings in this roundup, we did close listening tests using earphones, and the TubeSound exhibited large amounts of blitter noise, mixer fader zipper effect, and even noise leakage when the mouse was moved around on-screen. But despite all that, once we began playing musical material on it, the tube pre-amp’s “warmth” took over, and the TubeSound impressed us.
Building a tube section onto a motherboard that consists of relatively low-voltage components is an impressive engineering feat all by itself. The fact that it works at all is pretty impressive.
The TubeSound also includes an odd-duck of a media player (pictured above). This odd-skinned media player handles most typical audio playback chores, but missing from the mix is any kind of mixer. Instead, Realtek relies upon the plain old Windows mixer to cover these duties, which is unfortunate, given the standard mixer’s many shortcomings.
The TubeSound, like the tube pre-amp/amp heritage it comes from, is something of a paradox. By most empirical measures, transistor-based amps clearly outperform the older tube technology, and yet despite whatever objective data you can show to this effect, it has to be said that tube amps do have a pleasing sound to human ears. The TubeSound is an interesting and unique approach to differentiating AOpen in the crowded motherboard market, but despite its warm tones it comes with some pretty sizeable warts, too.
Analog Devices (AD) has been working for quite some to develop its SoundMAX motherboard audio solutions. An outgrowth of AD’s acquisition of Staccato Systems, SoundMAX is designed to deliver a host-based audio solution that meshes well with AD’s motherboard-down audio codecs.
The end result is a well-appointed audio solution that includes support for EAX 1.0 and DirectSound3D, along with noise cancellation technology from Andrea Technologies and stereo microphone array support. These last two features improve noise cancellation and speech recognition performance. In addition, AD’s latest codecs include jack-sensing, which helps users wire all the audio I/O correctly.
Here’s SoundMAX’s mixer app, which is designed to look like a home audio component. All mixer faders use linear percentages, but as you move the fader, you get an update as to the fader’s level. There are no “Vu meters” however, so there’s no visual cue to tell you when you’re approaching full-scale, which would be a useful addition.
Additional panels include the ability to set an audio environment, which are essentially reverb presets. Some of these are useful, others are just downright silly. Presets like Stadium, Canyon and Padded Cell are fun the first time or two, but ultimately provide little real benefit.
Some of the different Hall presets, however, add just enough reverb to be felt, without putting you in an Echo Chamber from Hell. However, there are no end-user panels where you can design your own presets, and tweak parameter “knobs and levers” until you create one tailored to your ears and listening environment. Also missing is an EQ, which is useful for helping customize sound quality when running through different-sized speakers.
SoundMAX did exhibit high CPU usage in our Audio WinBench testing, but its frame-rate drop was about mid-pack in our game tests, and its signal quality tests were for the most part good, though not great. SoundMAX is the most consumer-friendly audio solution we looked at in this roundup, and is a nice fit for users who want a good feature set, but don’t want to climb a cryptic learning curve.
nVidia’s nForce2 chipset features the MCP-T audio processor, the only solution we’ve seen with a DSP core (it actually has three). This audio processor powers the Xbox, and has the unique ability to encode any audio stream into Dolby Digital 5.1 in real time – an enticing feature all by itself. The baseline MCP chip supports only basic two-channel AC97 audio services, whereas the MCP-T comes with all the bells and whistles.
ThenForce2 chipset currently only supports the Athlon / Athlon XP line of processors. There are two separate versions. The IGP includes an integrated graphics controller, while the SPP does not. Both support AC97 (host-based) audio functionality. For more details, see our in-depth review of six nForce2 motherboards.
nVidia has added features to nForce2 that it now calls SoundStorm, which add multichannel speaker support, S/PDIF output, an enhanced set of audio utilities and the ability to encode Dolby Digital in real time. The entry-level nForce2 MCP audio processor lacks these features.
Details: The Rest of the Herd
Neither the Via Envy24HT or the C-Media 8738 impressed us with any especially interesting features or software apps. Their overall performance was a mixed bag, and C-Media in particular left us cold. Although its noise floor test results were decent, especially compared to the competition, the rest of its performance story is unremarkable. And anecdotally, my home server’s motherboard uses this very same audio processor, and I wound up having to swap in a Sound Blaster Live because the line-in capability was so poor that recorded TV audio was barely discernible.
Via’s Envy24 codec powers some impressive sound cards, both in the consumer space (M-Audio’s Revolution 7.1) and in the pro audio market (Roland DA-2496). However, the Albatron design appears poorly implemented, as evidenced by its poor performance on our noise floor test. However, these results are offset by its solid performance on RMAA, where it actually bested the Sound Blaster Audigy2. This indicates that whatever noise floor is present when the Envy24 is idle goes away when live audio signal is traveling through it.
When all is said and done, we can’t unconditionally recommend any of these motherboard-based audio solutions. Each one comes with its own warning label:
Upside
Downsides
AD SoundMAX
Great software-based feature set including noise cancellation for speech recognition and voice apps; jack-sensing technology is a nice touch
High CPU overhead
nForce2
Real-time Dolby Digital encoding, three DSP cores to power through audio chores. Low overall CPU usage
AMD-only for now, limiting CPU choice. MSI implantation could be cleaner
Via Envy24
Low CPU overhead; good dynamic range performance
Implementation we tested had no extra bells and whistles, and no extra software
AOpen Tube
That old-school sound
Noisy in close listening tests when idle, but sounded fine playing audio test material. Expensive compared to other solutions.
C-Media
Decent performance in dynamic range tests
High CPU overhead shown in all tests. So-so signal quality
Of the five, there are three that are worth a serious look: nForce2, SoundMAX and the AOpen TubeSound. This last choice is one where you’ll have to make a conscious decision to live with the rosebud and thorns of tube-based audio. For hobbyists and tube devotees, though, there is no other choice.
nForce2 is interesting for its processing horsepower, and its feature set. However, the MSI implementation leaves a lot to be desired, and we’d like see a very well done implementation of nForce2’s venerable MCP-T that matches its cool feature set to a worthy implementation to let this solution be all it can be.
SoundMAX is probably the most well-rounded solution, balancing features and decent performance, despite its appetite for CPU cycles. Despite its high score in Audio WinBench, its frame-rate drop was about on average with the group tested here.
Motherboard audio has progressed quite a way, but it’s not ready to kick discrete audio out of every performance PC. That day may come, and with solutions like the beefy nForce2 available, companies like Creative need to continue innovating to differentiate its products.
For pro audio applications and home studio weekend warriors, discrete is still the way to go, since none of these solutions offer support for ASIO/ASIO2. However, given the pro-audio cards powered by Via’s Envy24, it wouldn’t be too hard to bring those high-end capabilities to a motherboard-down offering.
We keep hoping that nVidia’s nForce2/MCP-T will support the Pentium 4, although such an offering doesn’t appear to be on the near-term horizon. This is a topic we’ll continue to follow and bring to you, as it may represent the future direction of much of the audio in PCs.
Copyright © 2003 Ziff Davis Media Inc. All Rights Reserved. Originally appearing in ExtremeTech.
