My computer has a built-in optical output capable of 24/96. And my DAC is likewise 24/96 capable, but will also upsample lesser bitrates internally. I have some hi-res music, but most of it is just 16/44.
 
If I understand correctly, the main benefit of upsampling on the DAC is reduced jitter. So it seems to me that it's better to output 16/44 from the computer in this case to take advantage of jitter reduction via DAC upsampling. And upsampling on the computer side defeats that, and produces even more jitter because of the increased data being sent over the cable. Is that correct?
 
And how about this: (Low jitter + 16/44 lossless) > (High jitter + 24/96 lossless) ?

well beside the jitter issue I think the DAC will do a much better job upsampling the signal than the driver of your onboard sound

If you can send bit perfect signals from your soundcard to your DAC and let the DAC do any upsampling that is the way I'd go. The problem is that the default mode for computer sound output is a single sample rate and bit-depth so mixing files of different rates and bit-depths isn't easily done without one of them getting messed with by the computer. I don't know how your computer playback is set up, but if you're in doubt search on here for bit perfect, (many folks here use ASIO and Foobar which you might also want to include in the search) and you'll find dozens of threads about it and how to set it up.
 
I wouldn't worry too much about jitter, unless you are thinking of adding a new soundcard to reduce it. If the same clock is used for both 44.1 and 96kHz sample rates then the absolute jitter will be the same, the relative jitter will be less on the 44.1kHz sample. Don't worry about jitter too much from the computer, most modern DACs are pretty good at dealing with it. Resampling by the computer will be more noticeable.

Thanks for the responses! I shall stick to 16/44 then. I use iTunes on a Mac, with 100% volume and no EQ, so it should be bit perfect from what I've read.

[size=medium]I wonder if upsampling really reduces jitter.[/size]
[size=medium]Running at a higher sample rate pushes the aliases out of the audible range and therefore be beneficial for the filtering[/size]
[size=medium][size=medium]ASRC (Asynchronous SRC) is a effective method  to reduce input jitter but I wonder if your DAC supports this.[/size][/size]
[size=medium]http://thewelltemperedcomputer.com/KB/Sampling.htm[/size]
 
[size=medium]There is a simple answer to your question.[/size]
[size=medium]You have 2 SRC’s, one implemented in software (iTunes) the other implemented in hardware.[/size]
[size=medium]Set iTunes to 96 and the software will do the job[/size]
[size=medium]Set iTunes to 44 and the DAC will do the job[/size]
[size=medium]Compare both settings in a listening test and you have the answer[/size]

My DAC does indeed support ASRC upsampling. It is an AMB γ2 DAC, built by MisterX.
To quote one of the tech highlights:

1. "An [...] asynchronous sample rate converter (ASRC) chip (Texas Instruments/Burr-Brown SRC4192 or Analog Devices AD1896) up-samples all input digital streams to 24-bit 96KHz. The ASRC and DAC are both fed with an I²S master clock generated by an onboard ultra-low jitter CMOS oscillator. Together with careful circuit board layout, the result is an extraordinarily low overall jitter."

 
Quote:

roseval said:

[size=medium]There is a simple answer to your question.[/size]
[size=medium]You have 2 SRC’s, one implemented in software (iTunes) the other implemented in hardware.[/size]
[size=medium]Set iTunes to 96 and the software will do the job[/size]
[size=medium]Set iTunes to 44 and the DAC will do the job[/size]
[size=medium]Compare both settings in a listening test and you have the answer[/size]

Click to expand...

Ha. Yeah, that seems so obvious now that you mention it. I guess I was just too caught with the theory of it all.

Okay, just a tiny update on some of my findings based on casual listening.
 

1. 24/96 original ==> 16/44 on iTunes ==> 24/96 on DAC

Definite loss of clarity here. Quite a big difference to me, actually.
 

1. 16/44 original ==> 16/44 on iTunes ==> 24/96 on DAC
2. 16/44 original ==> 24/96 on iTunes ==> 24/96 on DAC

I couldn't tell the difference here.
 

1. 24/96 original ==> 24/44 on iTunes ==> 24/96 on DAC
2. 24/96 original ==> 24/96 on iTunes ==> 24/96 on DAC

These two sounded equally good. So it appears that rescaling the bit-depth has a much greater effect that rescaling the sampling rate. Since most of my music is 16/44, with a few 24/96 albums, I think using 24/44 is the way to go for me.
 
Note that I am not claiming that 44.1 and 96 makes no difference. Only that my ears and/or equipment are perhaps not sensitive enough to perceive it casually. But I like to think that the Gamma 2 is so awesome that even jittery Toslink can't faze it   I don't really care enough to conduct more thorough testing.
 
 
Equipment used:
iTunes ==> iMac ==> TOSLink ==> Gamma 2 DAC ==> Aune Mk2 SE (with AD4627-1) ==> Beyer DT-880 Pro
 
The Gamma 2 is powered by a TREAD, and the Aune by a PowerVolt linear regulated supply.