Does a USB DAC use cpu cycles in games?
Nov 20, 2008 at 3:51 PM Post #2 of 12
Yes, however I don't know if "eaten up" is the right term, however your CPU will be responsible for the processing of your sound.
 
Nov 20, 2008 at 8:43 PM Post #4 of 12
Guys are you sure about that?
I mean... the only thing a DAC does is to convert digital signal to analog. The processor can't do this conversion, it is responsibility of the DAC.
Also these DACs do not use any EAX, or these stuff...
I just can't see what the processor has to do at all in this process.
confused.gif


I'm not telling you're wrong, I just don't understand what the processor has to do.

Of course, the processor has to do the calculations of the sound effects in game, to create the sense of direction, distortion effects, etc... but it's no more than a soundcard like an Audigy SE would do anyway: leave the effects processing to the processor.

Besides that, it's supposed to be just feeding the DAC with the digital signal, which should take an imperceptible usage of CPU since it's just throwing the raw sounds to the DAC, and the rest is up to the DAC/amp.
 
Nov 20, 2008 at 9:11 PM Post #5 of 12
Quote:

Originally Posted by Vitor Machado /img/forum/go_quote.gif
Guys are you sure about that?
I mean... the only thing a DAC does is to convert digital signal to analog. The processor can't do this conversion, it is responsibility of the DAC.
Also these DACs do not use any EAX, or these stuff...
I just can't see what the processor has to do at all in this process.
confused.gif


I'm not telling you're wrong, I just don't understand what the processor has to do.

Of course, the processor has to do the calculations of the sound effects in game, to create the sense of direction, distortion effects, etc... but it's no more than a soundcard like an Audigy SE would do anyway: leave the effects processing to the processor.

Besides that, it's supposed to be just feeding the DAC with the digital signal, which should take an imperceptible usage of CPU since it's just throwing the raw sounds to the DAC, and the rest is up to the DAC/amp.




it has to generate the digital source for the DAC, that doesn't just come magically out of nowhere or from some other processor
even with something like x-fi, the primary CPU is responsible for a given element (it even creates things to offload to your GPU)

so its going to basically "set up" and "compute" all of the audio, and send whatever format digital stream to your D/A, which is simply taking digital and making analog, the CPU (for reference) could actually do this math, however it isn't neccisarily wired for this kind of output (its a general purpose microprocessor, it can do more or less anything, assuming time is no object)

so yes, the CPU is going to be working regardless of what solution you have for audio, the question is more, how much is going to be working, with a USB D/A its going to do a lot more than with X-Fi or Xonar, however it will never "not touch" anything being computed in the machine

::edit
just re-read your "raw sounds" thing, honestly, you probably will disagree with me based on that kind of understanding, simply put, the CPU does everything, specialized add-ons merely serve to enhance or speed-up the general ability of the system, in truth, the entire game could be run via the CPU, again, if time is no object, GPUs, audio processors, network controllers, etc, they all serve to help the processing time or provide features that are infeesable to implement on a conventional CPU (for example digital signal processing, which is very repetitive and benefits from highly parallel pipeline design, which would be wasted on something less predictable)
 
Nov 20, 2008 at 10:35 PM Post #6 of 12
Yeah, USB always uses CPU cycles when its in use.
For example, anytime you transfer a file to a pendrive via USB, CPU is being used (thats one of the pros of Firewire against USB, for example).
Another thing is that for a normal nowadays CPU like a Core Duo, that load is quite ridiculous.

If the DAC needs to generate the source, like very correctly obobskivich said, audio-data needs to be generated(which is CPU), and later moved from computer to the DAC, via USB, and that has load too of CPU.


But the best are the tries. According to impact of soundcards on the FPS of the games, which i things that matters you, u can check this link (TechReport is a reference for games):
Creative's Sound Blaster X-Fi audio processor - The Tech Report - Page 6

But hey, don't worry much, as u see there are not big differences there, even with the famous X-Fi XRAM :wink:
Just in nowadays games, the 80% of the FPS is the graphic card
wink.gif
(i'm a computer science engineer and gamer
atsmile.gif
)
 
Nov 21, 2008 at 1:09 AM Post #7 of 12
Quote:

Originally Posted by obobskivich /img/forum/go_quote.gif
it has to generate the digital source for the DAC, that doesn't just come magically out of nowhere or from some other processor
even with something like x-fi, the primary CPU is responsible for a given element (it even creates things to offload to your GPU)

so its going to basically "set up" and "compute" all of the audio, and send whatever format digital stream to your D/A



The question is, how much is that CPU usage? Is this even worth considering?
I'm not talking about effects, and sound processing in the games (which some soundcards, like I said, will still leave this to the processor also. and some like the Xonar or X-Fi will help with their own built-in processors), I'm only talking about getting and sending the digital signal to the DAC. The digital signal itself, is the raw content of sound files, it doesn't require, in itself, any processing.

Imagine for example, you're playing a song in a player like foobar2000. fb2k will decode the sound file (and the processing here will depend only of the compression used to store the sound), and the result of this decoding is the digital signal. There's no real big processing to achieve the digital signal in itself.

Not sure if I explained my idea in the best way...

Quote:

Originally Posted by obobskivich /img/forum/go_quote.gif
...taking digital and making analog, the CPU (for reference) could actually do this math, however it isn't neccisarily wired for this kind of output (its a general purpose microprocessor, it can do more or less anything, assuming time is no object)


No, hold on a sec., the processor cannot create an analog signal. Processors work with digital data, it can only work in the digital realm.
It won't be able to create an analog signal ready to the speakers by itself.

Quote:

Originally Posted by obobskivich /img/forum/go_quote.gif
so yes, the CPU is going to be working regardless of what solution you have for audio, the question is more, how much is going to be working, with a USB D/A its going to do a lot more than with X-Fi or Xonar, however it will never "not touch" anything being computed in the machine


Yeah these soundcards have built-in processing to help in the sound effects in games, but I'm comparing to simpler soundcards like the Audigy SE, since these still leave this processing for the CPU.
My point, is there will be no perceptible performance difference from something like an Audigy SE to an USB DAC for example. They will be doing the same things:
-Processing of the game's sound effects are done by the processor.
-And then the digital signal is simply sent to the DAC, and the rest is up to it.

The difference comes from soundcards with built-in processors, these will help with the sound effects processing and reduce CPU usage there, and only there.
 
Nov 28, 2008 at 7:12 AM Post #8 of 12
Sound processing does take a fair chunk of processing power.

You have to understand that its not just one sound source at any given time but possibly up to hundreds, and each soundsource is given a co-ordinate relative to the player and hence reverb in relation to other objects and sound sources.

Which is significanly HARDER than just playing one sound-source from an encoded file through foo-bar. And up-sampling takes a fair chunk of resources (if you do go down that path).

What is to be noted is that most USB sound-cards, would do the sound processing, more often than not, in a emulated software environment (sorta like how they used to render 3D). Since the CPU isnt made to deal specifically with sound, it must use significantly more resources to create the same sound environment vs a Audigy/Soundblaster/etc since they have a dedicated audio processor i.e. meant to deal with that stuff.

So using the same analogy as the GPU (video card dealing with 3D), since the CPU wasnt created to deal with certain vector calculations or multiple parallel processes for dealing with smoke, lighting etc. it takes a lot of software programming and emulation to create the same effects hence lots more processing power required from the CPU.

This is similar to Soundcards and it having to deal with reverb, and I would say yes that the USB dac will require more processing power than a dedicated soundcard to work as the USB sound in games is done through direct sound (or openAL if you're vista) and most of the processing is done on the CPU through software where as for a dedicated soundcard, these processes are dumped off onto the soundcards processor.

So if theres a lot of sounds and perhaps lots of effects and reverb, then it may take a fair slice of CPU time. However it is to be said that most CPU's now-a-days are pretty gosh darned grunty, and on a Core2 or AMD equivalent, that your framerate increase when changing to a dedicated soundcard, now-a-days is relatively negliable, however certain sounds may sound a bit better cause the soundcard can do post-processing effects on the in-game sounds.

Another example of this is the AC97 on-board sound, most of its sound processing is done through the CPU. Initially it took a fair amount of resources, sorta like 10-20% back in Quake3 times, but now-a-days, i dont think its very much.

So in conclusion, the USB-dac doesnt make that much of a performance impact, but certain post-processing effects may increase this performance impact or perhaps not even be able to be run by the USB-dac.

Cheers

gav
 
May 18, 2015 at 5:00 PM Post #9 of 12
I think the only time you need to be concerned about things like this is when every microsecond counts like when recording and mixing audio processed with real-time effects. The higher bit-rate, sample rate does mean a reduced roundtrip for latency however at the cost of CPU usage. In the age of multicore processors this is less of an issue however not every application will automatically do this (Earlier versions of Apple Logic for example). Even modern DSP solutions, VPU (Video Processing Units) and so forth will use the CPU alas SSE extensions and other parts of the help free up resources.
 
 
Audigy/Soundblaster and some other gaming soundcards are all rounders and in some cases will colour sound. So you know that pair of Grado SR325e you bought, they aren't meant to sound different from another system. Neutrality is a founding basis for any form of audio entertainment.
 
Dec 10, 2015 at 1:35 PM Post #12 of 12
I think the only time you need to be concerned about things like this is when every microsecond counts like when recording and mixing audio processed with real-time effects. The higher bit-rate, sample rate does mean a reduced roundtrip for latency however at the cost of CPU usage. In the age of multicore processors this is less of an issue however not every application will automatically do this (Earlier versions of Apple Logic for example). Even modern DSP solutions, VPU (Video Processing Units) and so forth will use the CPU alas SSE extensions and other parts of the help free up resources.


Audigy/Soundblaster and some other gaming soundcards are all rounders and in some cases will colour sound. So you know that pair of Grado SR325e you bought, they aren't meant to sound different from another system. Neutrality is a founding basis for any form of audio entertainment.


The USB interface itself also incurs pretty heavy latency penalties due to the topological model; there's a reason FireWire was so popular for external interfaces for so many years, and why a lot of prosumer and proline gear uses PCI or PCIe add-in cards (e.g. RME, Avid). USB is generally fine for basic audio output, but if you're after real-time multi-track recording you probably want to consider another interface.

You mean that a PCI sound card like STX II eats less CPU than lets say Sound blaster X7?


Ehh, that's tougher to say, specifically because you chose the X7 as an example, for two reasons:

1) I've never seen conclusive documentation as to exactly what parts of the audio chain SoundCore is doing in hardware and what it passes to software, and I've never seen conclusive documentation as to variations between SoundCore on different products.

2) The STX further complicates things because its based on a C-Media chipset that does no hardware processing, but is just an I/O router. So there's no "offload" there; it's all handled in the drivers.

Ultimately I'm going to venture that the differences between them are negligible, but its plausible the X7 is actually more efficient, depending on exactly how much work SoundCore is really doing. This largely doesn't matter in a modern context, as everything on the computer's side is likely done in software from the beginning (e.g. it isn't even trying to hook hardware, regardless of whether or not the hardware is available or not (e.g. have your Audigy all you want, it isn't doing any DSP mojo)). And modern CPUs are more than powerful enough to handle this kind of workload without missing a beat - when this thread was first started in 2008 (some 7+ years ago) that was mostly true, but not as universally true as it is today.
 

Users who are viewing this thread

Back
Top