Ask HN: Whatever happened to dedicated sound cards?
During the '90s and the early '00s, dedicated soundcards were in-demand components in much the same way GPUs are today. From what I know, Creative won, on-board sound became good enough sometime between Windows XP and Windows 7, and the audio enthusiasts moved on to external DACs and $2000 headphones. Today Creative still sells soundcards, but none of them appear to be substantial improvements over previous models.
So what other reasons could have caused the decline in interest? Was there nothing that could be improved upon? Were there improvements on the software side that made hardware redundant and/or useless? Is there any other company besides Creative, however large or small, still holding the torch for innovating in this space?
238 comments
[ 346 ms ] story [ 1396 ms ] threadI think another factor is MP3 players and phone audio; people stopped using their computer as the (interface to) media source when other things took that function over for them.
If you don't like the DAC in the headphones, you can also find a high-quality USB DAC and use the audio cable from there.
This has not been my experience at all. For my 5" powered studio monitors, the _only_ way to get a interference-free signal from my desktop computer was with an optical cable to an external DAC.
I 'worked around' the issue for the longest time by leaving covid based processing going on folding@home, which would force my CPU into turbo mode.
Eventually I found out that if I disabled cstates, it mostly went away.
The fiber is standard and you don't need any fancy Monster fiber costing $100 a foot.
For anyone curious, check out Schiit Audio. I have the Magnius/Modius paired with Sennheiser HD660s headphones and I couldn't be happier. JDS Labs Atom is also a good choice.
For example if you produce music, you probably are good with an external USB audio interface like a Focusrite Scarlett.
Been through three different external audio devices, and they've all sucked. I'll try yours.
I think in practice all of them are good enough but if you’re into numbers, dB, Hz, and measurements, check out Julian Krause audio interface reviews on YouTube.
https://youtube.com/playlist?list=PLv875tu-z7M4EyBeuofJ1Tehq...
The drivers are really important, and often audio interfaces aimed at professionals will use a driver that a DAW can utilise with control over latency. macOS users tend not to need to worry about this so much because as long as they are good it works, and on DAWS on Windows I use ASIO because MME, WASPI and WDM have had poor behaviour and/or latency in the past. But for general music listening those other driver types should be ok if the manufacturer supports them well - but this is where the problem might be.
I am guessing it might be the driver that sucks in your experience?
Have you tried S/PDIF output to an external convertor?
They have been for years as long as you don't need high end 3D or compute capabilities. Intel's iGPUs are plenty for most normal desktop use cases and AMD's are good enough to run most mainstream games as long as you're willing to turn the detail down a bit. Not to mention what Apple and the ARM world in general have done.
Someone who's not looking to play the latest games with the settings cranked can easily go without.
The top end will always want as much horsepower as they can get, so the dedicated video card is never going away unless we figure out some new sort of computational load that takes over the accelerator market in the same sort of way that GPGPU killed standalone audio and physics acceleration.
For games it's easy to find useful ways to use 10x the existing GPU power, not even counting an increase in display resolution.
For massively parallel computing, the sky is the limit.
Now every computer has a little chip that plays back at least CD quality audio from an infinite pool of storage and RAM. Nobody wants to hear MIDI in their games anymore. I'm not even sure what a better sound card could even do for me - reduce line noise, drive high impedance headphones or something. Boring!
For games today, I use the audio interface that came with the computer. For dealing with my synthesizers, I use a Scarlett by Focusrite [0].
[0] - https://focusrite.com/en/scarlett
Sometimes they would ask to open the box to see the chips on the board.
I was all about 3D acceleration and didn’t have the money or interest to do anything with audio so I never learned much more about it.
even a $15 USB-to-dual-3.5mm adapter sounded significantly better on other machines
I experienced this annoying line noise on ASUS, MSI and Dell motherboards
Even desktop publishing took a turn towards Windows in the mid-late 90s, whether it stayed 'dominant' is a fuzzy question but it was clearly losing market share.
But you won't find a professional music studio without a Mac, this has been true since the late 80s. This is not to say it can't be done with other equipment, just that as a matter of practice it isn't.
I have a Sound Blaster something-or-other in my PC but it's not connected to anything these days. I use a tiny Apple headphone dongle that was ~$10, has 0 noise and sounds great.
When DVDs and HDMI were becoming popular, and Windows Vista was launched, a lot of restrictions were put on drivers, I saw many people defending them claiming it was for better stability, avoiding blue screens and so on.
But a major thing the restrictions did, was restrain several of the sound cards features, most notably their 3D audio calculations that were then just starting to take off, people were making 3D audio APIs that intentionally mirrored 3D graphics API with the idea you would have both a GPU and a 3D audio processor, and you would have games where the audio was calculated with reflections, refractions and diffractions...
After that, the only use of sound cards became what the drivers still allowed you to do, that was mostly play sampled audio, so sound cards became kinda pointless.
Gone are the days of 3D audio chips, or having sound cards full of synthethizers that could create new audio on the fly.
Yamaha still manufactures sound card chips, and their current ones have way less features than the ones that they made during the sound card era.
EDIT: also forgot to point out the same restrictions kinda killed analog video too, for example before the restrictions nothing prevented people from sending arbitrary data to analog monitors, so you could have monitors with non-standard resolutions, non-square pixels, unusual bit depths (for example SGI made some monitors that happily accepted 48 bits of color) or not even having pixels at all (think vectrex) and so on. All this died and in a sense also affected video development, some features that video cards were getting at the time were removed and hardware design moved to a narrower path, more compatible with MS rules.
As for what the restrictions have to do with DRM: the point was not allow people to intercept audio and video using analog signals with perfect quality, since this would be an easy way to go around the DRM built-in on HDMI cables.
To be fair, realtime synthesis just became obsolete for most purposes once CD quality digitized audio became cheap enough to store (and later, to stream). And for musicians, once CPUs became fast enough, SW synthesis with its limitless possibilities took over from HW synthesis.
I think full 3d audio is a different problem though because it, at least, requires a version of the rendering problem (for waves). It's harder in some ways than light rendering (because phase/coherence matters sometimes, the wave equation is harder to solve), but easier in others (no need as much detail as light, wavelength is large), or just plain weird (nonlinear effects from rattling and such).
Or there’s still some “generic” box (I mean that you can program yourself) like the Symbolic Sound Kyma Capybara. They’re quite niche thought, like modular synth.
There are always slightly harder, and better ways of doing something which the accelerators are better at. Audio acceleration I guess peaked too early or they just couldn't get the tech demos as impressive as graphics.
I remember reading about a demo, which I believe was from Matrox. They managed to get an audio 3D environment working over a pair of stereo headphones which was good enough that you could play Doom headless. Just close your eyes and you could tell where people were.
A lot of what I read about here is that prerecorded samples is good enough, in the same way that raster lighting is good enough and raytracing is a waste of time.
It is cheap enough that it is bundled with everything. But that doesn't make it better than a discrete GPU/Audio chip.
Intel GPUs are barely good enough to run an OS desktop yet they hold the majority of the market.
Whether Dolby Atmos has improved since then or other games have implemented it better I don't know. I feel like we probably need an open source implementation middleware for object sound to headphone/surround speakers to really fix the situation.
It’s a shame because I would love more audio sources to support HRTF (head related transfer function) and “ray traced” audio.
The game deliberately includes many sound sources to facilitate this such as stepping on various surfaces, glass shards on the ground, and wildlife making noise based on player proximity.
This works amazingly well on regular, on-board PC sound chips, though headphones are quite mandatory.
(Disclaimer: not affiliated, just a fan).
Higher order ambisonics strikes a pretty good balance in terms of spatial resolution while still being a mixed signal. You can then pair it with objects for specific highlights. Atmos is a 7.1.4 static bed plus dynamic objects, so similar idea. In either case, most of these 3D sound systems support very few dynamic objects. For example, Windows Sonic only supports 15 dynamic objects on Xbox: https://docs.microsoft.com/en-us/windows/win32/coreaudio/spa...
Modern CPUs can ether do or emulate this, probably using less power than a sound card.
Very, very, few people have their PCs connected to an AV receiver or multichannel speakers, but positional audio is still widely supported in Windows applications using Xaudio2.
The reasons sound cards went away is the use cases went away:
1. People who want high quality recording shifted to firewire and later high-speed USB external audio interfaces. No matter how hard you try an external metal box with multiple inputs and outputs will always be better than a PCI/PCIe card inside a PC for recording. Rare use case in the recording world for sound cards.
2. Gamers who want 3d/positional audio either use headphones, find the 5.1 integrated outputs to be adequate, or like me, run a digital audio cable to a surround sound receiver. Rare use case in the gaming world for sound cards.
Dolby Atmos is awesome for positional audio in games but there are multiple less expensive and more accessible methods for surround audio nowadays. Decent positional audio can be experienced using a laptop and headphones-- no sound card required.
https://www.pcgamingwiki.com/wiki/Glossary:Surround_sound
Back in the sound card days you had to squint on the back of the box and ask "is this creative 3d? aureal?" nowadays you just plug in 5.1 to your PC's onboard audio, tell windows you have 5.1, and it works (mostly).
Then there's Korg's Oasys PCI, which was so powerful that for a long time people kept using Windows 98 after Korg stopped making drivers: https://www.soundonsound.com/reviews/korg-oasys-pci
USB can't offer as low latency as a piece of well-designed hardware plugged directly into your PCI bus, at least in my own limited experience. This comes into play when doing music keyboard recording.
eg. I found it difficult to find a USB MIDI adapter that didn't introduce unacceptable latency (when trying to record new tracks synced in real time to existing ones). Edirol was recommended to me but even after tweaking settings for hours it fell short. I wound up buying a second-hand Creative X-Fi Elite Pro PCIe card and love it.
I don't think it's a USB protocol problem but rather a driver/manufacturer problem.
This may be true but I've never had latency issues with USB soundcards. Right now I have a Line6 Helix Floor unit that I use to play guitar with. I can route the audio through the Helix effects, into Logic and back to Helix for more post-processing and have no latency problems.
I have had other brands and models and none introduced perceivable latency.
I don't use MIDI but I doubt it requires less latency than live guitar playing.
I had a PCIe soundcard a few years ago that made it almost impossible to get rid of ground hum though.
Last time I did audio, installing ASIO4ALL was essential.
Or get an external MIDI clock like the E-RM Multiclock and never bother about MIDI latency issues. Audio is completely fine, 30-50μs of latency won't ever be perceptible to humans, the latency of MIDI will.
In particular, the first readily available set of benchmark results I was able to find[1] suggest the difference in audio (not MIDI) latency between the lowest-latency PCIe cards and the listed USB devices (RME Fireface UFX+ USB3 and RME Babyface Pro [USB2]) is under a millisecond.
While I couldn't find similar results for MIDI I/O, it seems unlikely that either of these USB devices' MIDI interfaces would introduce an order of magnitude more latency than their audio counterparts.
[1] https://gearspace.com/board/showpost.php?p=15796206&postcoun...
https://www.bleepingcomputer.com/news/security/heres-a-list-...
So USB is pretty good, and for most sound card, USB2 is enough. Otherwise you can go Thunderbolt, which offer on par experience with PCIe.
What a consumer sound card offer now a day is better dac compare to the one of your motherboard or better output for headset.
Ok, I just checked, and a 192kHz 24-bit WAV file is only 0.56 MB per sec. Nice.
...in part because there's no way to do that, and if you do it by using the headphone jack, in addition to low quality you're also going to get all system sounds
I don't think most people know what spidf toslink or RCA even are.
And beyond that, this is the first I'm seeing a claim that "doing 3d audio calculations" was restricted and that this had anything to do with intercepting pre-encoded multi-channel DVD/digital media streams. They seem completely separate from each other as far as technical pipelines go.
Sound cards as a general consumer product were dead long before Vista. The last hurrah I remember was the SBLive!/Turtle Beach Santa Cruz era 1998-2001 stuff, Vista didn't come out until 2007 (Longhorn was famously botched, etc...).
CPUs just got fast enough that all of that, including 3d calcs, could be done better on common CPU by the mid-2000s. Do it on a sound card, you have to buy a new sound card to get improvements. Do it directly in the OS or in-game, and you can benefit from improvements from the OS or library or game devs immediately.
In particular, EAX (environmental audio extensions), which was a feature of the X-Fi cards, were definitely EOL'd due to Vista's changes around the DirectSound3D APIs.
https://en.wikipedia.org/wiki/Environmental_Audio_Extensions
https://www.tomshardware.com/reviews/future-3d-graphics,2560...
(It’s not a great chart but shows a general trend.)
A dedicated chip is often better than a general purpose CPU (hello GPUs?). Game audio made a huge step back with Vista and beyond. Since audio cards could no longer do what they used to because of the limited driver model and developers/studios were more focused on graphical fidelity and physics calculations, nobody was going to waste precious CPU cycles on audio, at least not any more than the bare minimum.
drivers live in the kernel, and prior to Windows Vista, they had the same rights as the kernel. pre-Windows Vista, a driver could easily be malicious and exfiltrate anything it wanted to anywhere it wanted. the driver architecture change fixed this gaping security hole, while still allowing drivers to exist.
drivers needed to be rewritten to accommodate LARGE changes to how they needed to do their work, and the result of that is that drivers which interacted with hardware directly no longer could, they had to ask the kernel to do stuff, and the kernel could say “no.” imagine being a driver maintainer and needing to react to this change.
this change often required a complete rewrite of the driver. this is why drivers of the era were so feature-limited.
this architecture change allowed kernel-level DRM drivers to become a thing, but DRM would have happened with or without any changes to driver arch, i assure you.
everyone suddenly needing to rewrite their drivers is what caused drivers to appear limited in the new paradigm. it simply took time to reimplement everything that existed in the old driver model, and people wanted working drivers before everything was implemented in the new drivers.
This is a thing too, but the main problem was what a kernel level drivers not only could, but WOULD crash the kernel (ie cause a BSOD) if something goes wrong. And things gone wrong very, very often.
People like to bitch about changes in NT6+, but I have seen waaay less BSODs (not related to the botched hardware) after that.
If you never seen a system BSODing from the sound drivers - I'm glad for you. I've seen enough sound card drivers crashes to tell you what that WAS a problem. Along with network cards, video cards, TV-tuner cards and almost anything what needed a driver.
> After that, the only use of sound cards became what the drivers still allowed you to do, that was mostly play sampled audio, so sound cards became kinda pointless.
Discrete sound cards became pointless because by 2001 almost every consumer motherboard had an AC'97 compatible audio coded on board.
So if you didn't need super extra fidelity 5.1235435 sound system AND didn't want to shell out additional ~$100 (SB Live! in 1999) or $2-300 (SB Audigy 2 in various variants, 2003) - you just could use the onboard one.
> having sound cards full of synthethizers that could create new audio on the fly.
NO THANKS: https://youtu.be/3AZI07_qts8?t=9
And this is a Creative card! I had my share of good synthesized music (because computers couldn't yet do a proper digitized sounds yet), but the tech should had die and it did.
Games ended up congealing around the Sound Blaster standard, which put Creative in the centre of the sound universe. Everyone else was always just "SB compatible", which meant they were playing for the "$10 placeholder sound card" market in the Pentium days. By the time we were getting onboard audio (I think my first board with it was Socket A), it was all hidden behind DirectX, and the market became the "90 cent placeholder chip soldered onto the mobo" market, and then they're all undercut by Realtek.
Unfortunately, Creative is a mediocre steward of the premium-sound landscape. Their product matrix is complicated, support is all over the place, and the drivers are sketchy. I have an Audigy RX that I pulled out of circulation because it could crash two different boards (B550 and X570) and the general consensus was just "they're not that compatible."
I suppose that market technically didn't crush everyone else, there was still the pro-audio market, but that had entirely different needs than a typical home enthusiast. If you're building a studio PC to run specific studio software, you can put up with a narrow compatibility list and finickiness.
But it feels like there's a reasonable niche there for the "eager to throw money around" audiophile crowd. Cards full of high-markup capacitors and filters so you can claim to offer cleaner power and a lower noise floor seem tailor-made to unlocking those wallets. Where is that card? Although I suspect for that audience, they just pipe out stuff via optical to an external DAC because the inside of the PC must be full of RFI.
For the masses, the high end today is mostly encompassed by a USB headphone DAC. Headphones get you high quality in a small form factor, and a headphone DAC doesn't need a lot of power or I/O. Once you go bigger, again, physical experience takes hold. People want their vinyl collections and so forth in their listening room, and thus where there's demand for digital, it's usually outside of the classic PC form factor too - it could be an iPhone and a Bluetooth speaker, or a dedicated receiver for the home theater setup. Going this route means it can(if built carefully) avoid crashes and updates interrupting the experience.
[0] e.g. early versions of Pro Tools https://www.pro-tools-expert.com/home-page/2018/3/27/a-brief...
Creative drivers were a double digit % of all Windows BSODs. Microsoft gave Creative plenty of time to fix their drivers, creative never did, so sound drivers got booted from the kernel.
3D sound and other processing got baked into middleware for games because it became trivial to do all of the processing in software - and the processing became more advanced than anything that the sound card vendors were offering (and they didn't move quickly enough anyway).
Pro audio vastly progressed past anything that is possible to provide in fixed silicon. For input, dedicated USB (and ethernet) audio interfaces progressed to the point where it would be ridiculous to provide such functionality on a general "sound card".
It's just evolution - there just isn't a compelling enough niche for a dedicated sound card any more.
This gets annoying.
I USED to have two powerful and rather higher quality speakers attached to a creative card back in the day when I did all that with the PC though.
3D audio on the PC was deliberately killed by Creative.
They sued Aureal into bankruptcy, bought it in the court auction, and the day the sale closed they nuked the support website and took the drivers offline.
They used similar scummy tactics to decapitate any other competitors. Then they considered their reverb based spatial audio solution sufficient, and promptly sat on their heels doing zero innovation while collecting a rent.
And them as chip technology improved, a basic "Soundblaster 64" chip became so cheap that motherboard manufactures started bundling it in as a selling point (which made a ton of sense for non gaming PC users btw). Additionally MS stepped in and provided some software spatial functionality within DirectX, as processors had improved to the point where dedicated hardware for it wasn't necessary.
Back then I worked in gamedev, and I briefly considered going into competition with MILES et all with a 3D audio library after the Aureal fiasco, after I stumbled on some interesting papers doing Fresnel Zone Tracing variations as low overhead spatial audio, but ultimately wasn't serious about it vs other options at the time.
Ie, Steam Audio: https://valvesoftware.github.io/steam-audio/
Which papers were they, if you recall?
Don't forget about bluetooth speakers, etc where the source is your phone or laptop.
As for people who just listen to music, IMO, built-in audio interfaces or just digital Bluetooth headphones have been good enough for a very long-time in digital-audio conversion.
- During the MS-DOS era, there wasn't really a standard API for sound, so using a cheap, off-brand sound chip (including anything that might be integrated) often meant compatibility problems. Even though it might not necessarily have offered the highest quality sound, Creative's Sound Blaster line was the gold standard for compatibility during this time. Standardized sound APIs have largely eliminated this issue.
-Throughout the '90s, music for games (and a number of other applications) was distributed as MIDI (or MIDI-like) instructions to be generated by a synthesizer, and the quality of the music was very much dependent on the synthesizer used. The Roland Sound Canvas series was the gold standard at the time (in part due to its quality, and in part because that's what the composers themselves used), but it was very expensive and out of reach to the mass market. Software synthesizers were either too slow, or the quality sucked. That gave an opportunity for sound card manufacturers like Creative to offer higher-quality hardware synthesizers on their sound cards than what cheap/integrated cards could do. These days, most audio is PCM, and hardware is perfectly capable of high-quality software sound synthesis, so hardware synthesis has become a non-issue and modern consumer sound hardware doesn't even have hardware synthesis capabilities anymore.
- During the '00s, sound cards began to offer accelerated environmental and positional audio (e.g., Aureal3D, Creative EAX), which games quickly adopted to improve the sense of immersion. However, changes in the Windows audio architecture introduced with Windows Vista broke this functionality without a replacement. Advances in CPU hardware have since allowed this type of processing to be done on the CPU (e.g., XAudio 3D, OpenAL Soft) with acceptable performance.
In the current era, we do have dedicated soundcards, although not in the form of PCIe add-in boards. External DACs (either dedicated USB, or integrated into a display or AV receiver) are popular, as are the DACs used by wireless/USB headphones. Also, there has been some work done to utilize the computational capability of GPUs for real-time audio ray tracing.
I have a Scarlett attached to the underside of my desk with a pair of Sennheiser HD600 headphones and a Monoprice Stage Right condenser mic on a desk stand.
I'm guessing the inner-PC card space has a bit too much of a EM noise problem for people who care about quality higher than integrated sound can provide (which is pretty good these days anyway) and external devices have room for the inputs people actually want.
I was going to use a different headphone amp with the hd600s and a modmic for general listening and calls though. Sounds like I might be able to drop that 2nd amp (I was thinking about the soundblaster x4)
I guess I'll get the Scarlett first and listen.
You should be fine dropping the additional headphone amp, which I think beyond the Scarlett or similar audio interface will perhaps be different depending on the amp, but not clearly making up for a deficiency in interface power.
There's this trusty Creative Audigy Rx that I nabbed from a closing down sale at an electronics retailer. Poetically, both are facing the same fate.
I built a Ryzen 7 machine and installed Windows 10. Whilst installing the CD drivers for the soundcard Windows 10 BSOD'd and rebooted. Not to be deterred, I tried the latest downloadable drivers (marked as Windows 8 compatible) but it is all WDM so surely OK? Not so. Another BSOD.
There was no help to be found online so I very reluctantly gave up. Now it lives in the retail packaging somewhere in my house, the bright and elaborate box promising an audible experience that exists only in my mind.
Windows _never_ worked well with it. A reboot may or may not have my sound card undetected due to some bug, I think it was related to their PlugNPlay, but I don't recall.
Until they completely dropped support for it, so then I just flat couldn't use it. And Linux support for it never really came around (maybe now? no idea).
So my policy now is fuck Creative. I purchased good money for a top-of-the-line sound card that never worked well and rather than fixing it they dropped support for the very next version of Windows.
I just use my motherboard audio because I don't have to deal with that shit and if I do ever purchase another separate soundcard it sure as shit won't be Creative.
I gave up on official drivers years ago.
1) Most people are happy with good enough. To most people's ears, speaker quality makes a bigger difference than audio output, and people already settle there. Furthermore, when iTunes was a big deal it turned out people got accustomed to low bit rates and mediocre equipment and thought it sounded better than the good stuff, because it's how they expected their music to sound.
2) With most computing moving to laptops and then to mobile, people generally don't have a choice about the audio processing technology inside their computer.
Well, the internal ones still exist [0]. However, with higher bus speeds, external interfaces are more practical: you can connect more devices to them, and you can move them - a lot of music today is done on laptops.
[0] E.g.: https://www.esi-audio.com/products/maya44ex/
In the early days of PC's most EVERY peripheral was provided by an IO expansion card save for the keyboard. My 386 had a 16-bit multi-IO ISA board that provided the essential coms ports: ATA, floppy, serial, parallel. You purchased a VGA card and then a sound card. You had at least 2 or three ISA cards because your motherboard was taken up by all the CPU, FPU, RAM, and essential control logic chips. My second 486, a DX2 66MHz, had the ATA, serial, parallel and floppy ports on-board which amazed me as it eliminated a whole ISA board. (Now everything fits on a single silicon die...)
Early on-board audio was usually a sound card soldered the the motherboard. Then Intel developed AC'97 which integrated standard audio into the south-bridge. Coincidentally that made Microsoft's life easier as all the PC's they were running on would use this standard meaning all they had to do was provide AC'97 drivers and everyone with an Intel machine had sound. No more competing 3rd party audio api's from Creative, et al. it was all Wintel. PC builders could now provide multimedia PC's for cheaper prices with audio by default. Also, USB happened which allowed people to plug in things without opening cases and fiddling with circuit boards which is alien to many.
And as CPU's became faster, the need for dedicated DSP processors to handle audio processing or synthesis is eliminated. You can now run a whole DAW complete with synth, sample playback, effects and mixing in real-time on a cheap general purpose off the shelf computer with no special hardware.
In the Mac/Amiga/non-DOS world sound was good enough very early. As soon as we said "we need sound," it was there, as long as you weren't on DOS.
The Amiga Video Blaster card was a thing, too. Your Android/iOS/PalmOS device surpassed that long ago.
It's still fun today, 25-30 years later, to crack open in a 90's MIDI in a DAW and route the channel outs through virtual instruments to see how they sound.
Add-on sound devices still exist, but they are simple because they don't include extensive hardware acceleration anything like what a GPU has. In fact, if you want hardware acceleration for audio processing algorithms today, like really fancy 3D sound propagation or something, GPUs would actually be great at that, and they support digital audio output too.
At roughly the same time, there were more peripheral buses like USB and Firewire being introduced, which meant that an add-on peripheral did not need to be an internal ISA/PCI card in order to have sufficient bandwidth for rich audio streams. These external devices could also be built with lower noise/interference compared to the boards inside a computer.
And of course, silicon integration always increased so that the bundled onboard IO chip became good enough for many users. So, add-on peripherals had to move up market or into niche settings. That is a bit like how the iGPU in Intel CPUs got rid of the market for basic VGA/XGA/etc. graphics cards for office machines.
These days basic "audio correctness" is readily available from onboard audio, though. Motherboards have gotten much, much better at noise isolating the audio area, and dacs & amps have generally improved.
For those that care about noise, you don't want the analog audio anywhere inside the case since it's a horribly noisy place. So you get an external DAC.