Sometimes. I’ve owned a computer with a sound chip soldered to the motherboard and also a plug in sound card with exactly the same chip on it (same drivers, etc). The one on the motherboard had faint but audible noise (like squeaking sounds when the hard drive motor was busy). The one on the card had a noise floor not noticeably higher than the quantisation noise floor.
The electrical power for a chip on a motherboard tends to be very noisy. The power the chip on the card was was being fed via the PCI pins was much cleaner. :)
A good motherboard will feed clean power to its sound chip. IME mediocre ones don’t and you’ll get better sound out of a $5 USB sound card since it gets the advantage of cleaner power.
I think this is even more noticeable with USB mics sounding nicer because mic input is even more sensitive than speaker output?
One big win from a lot of the early cards was having a big MIDI sample bank. This became pointless first when CPU speeds and RAM were fast enough to handle the MIDI synthesis and then more so when RAM got big enough that games used recorded (or even procedurally generated) music. Beyond that, they offered buffering and low-latency mixing. This was important because a CPU would often be interrupted at a time and leave a buffer empty (or, since they were rings, leave the card replaying the same sample instead of moving to the next), being able to DMA from a few ring buffers in memory and mix meant that the CPU could have a lot of jitter without causing audio artefacts. By the time CPUs were 100 MHz or so, this was less important. Then they got DSPs for dynamic effects and positional audio. By the time CPUs were a few hundred MHz, you could do most of this on a CPU.
These days, the only thing that the audio interface is doing that’s really useful is the digital to audio conversion. Higher-quality DACs make a noticeable difference (or, more accurately, low quality ones do. Modern DACs have such ridiculously high tolerances that they can be way above what a human can perceive, but that doesn’t stop people putting absolutely terrible ones in some systems). More commonly, decent electrical isolation makes a difference. After the DAC, you have analogue signals flowing to an amplifier / speaker and these will easily pick up interference. This is why a lot of setups have an optical connection between the digital source and the DAC / amplifier. Using an external USB audio device can give better quality for a similar reason: moving the DAC away from the case can make a big difference.
There’s also some complexity in synchronisation. Audio is often carried over HDMI now so that latency from HDMI encoding and decoding is the same for the audio and video signals. Doing this moves the DAC out of the computer, though not all displays have decent electrical isolation and so this can make the signal very bad.
I still have my CMS Gameblaster. No real reason, other than lovely nostalgia.
Does anyone know if sound cards still make sense today? From what I understand, a good motherboard will be good enough.
Sometimes. I’ve owned a computer with a sound chip soldered to the motherboard and also a plug in sound card with exactly the same chip on it (same drivers, etc). The one on the motherboard had faint but audible noise (like squeaking sounds when the hard drive motor was busy). The one on the card had a noise floor not noticeably higher than the quantisation noise floor.
The electrical power for a chip on a motherboard tends to be very noisy. The power the chip on the card was was being fed via the PCI pins was much cleaner. :)
A good motherboard will feed clean power to its sound chip. IME mediocre ones don’t and you’ll get better sound out of a $5 USB sound card since it gets the advantage of cleaner power.
I think this is even more noticeable with USB mics sounding nicer because mic input is even more sensitive than speaker output?
One big win from a lot of the early cards was having a big MIDI sample bank. This became pointless first when CPU speeds and RAM were fast enough to handle the MIDI synthesis and then more so when RAM got big enough that games used recorded (or even procedurally generated) music. Beyond that, they offered buffering and low-latency mixing. This was important because a CPU would often be interrupted at a time and leave a buffer empty (or, since they were rings, leave the card replaying the same sample instead of moving to the next), being able to DMA from a few ring buffers in memory and mix meant that the CPU could have a lot of jitter without causing audio artefacts. By the time CPUs were 100 MHz or so, this was less important. Then they got DSPs for dynamic effects and positional audio. By the time CPUs were a few hundred MHz, you could do most of this on a CPU.
These days, the only thing that the audio interface is doing that’s really useful is the digital to audio conversion. Higher-quality DACs make a noticeable difference (or, more accurately, low quality ones do. Modern DACs have such ridiculously high tolerances that they can be way above what a human can perceive, but that doesn’t stop people putting absolutely terrible ones in some systems). More commonly, decent electrical isolation makes a difference. After the DAC, you have analogue signals flowing to an amplifier / speaker and these will easily pick up interference. This is why a lot of setups have an optical connection between the digital source and the DAC / amplifier. Using an external USB audio device can give better quality for a similar reason: moving the DAC away from the case can make a big difference.
There’s also some complexity in synchronisation. Audio is often carried over HDMI now so that latency from HDMI encoding and decoding is the same for the audio and video signals. Doing this moves the DAC out of the computer, though not all displays have decent electrical isolation and so this can make the signal very bad.