Does anyone know if the sound quality should be the same whether using the CPU's headphone jack vs a headphone jack attached to speakers, attached to the CPU's line out?
They can't always (in absolute terms) be identical since they're not the exact same circuit.
The question is whether they'll both sound the same to any and every listener's ears, since in some cases they can both be comparatively equally bad or equally good depending on what you're comparing one to, and in those cases the differences may not be audible to everybody even if the differences are enough to show up on measurements. In your particular case, you can hear them, and no, that does not always indicate a placebo effect.
The opposite can also be true in that two amplifier circuits can have the same measurements and sound different to a human listener playing actual music without this being due to placebo. Why? Because you change several variables, only one of which is susceptible to placebo (even if all of it goes through that, ie, the listener). Putting a dummy load on an amp at a given impedance is technically the worst scenario since that dummy load doesn't have the same behavior as all transducers out there, as impedance shifts away from the nominal impedance rating depending on what frequency is being reproduced. And then even if you use headphones and measure them with the same mic and head-simulation set-up you still need a sine sweep to see if there are obvious differences from one amp to another (it won't be perfectly flat if you do it this way, because transducers have varying response curves), so that has a purpose; however it still has a limitation in the sense that real music isn't one frequency at a time but many, many, many frequencies playing at different amplitude and coming from varying distance to the microphone that recorded it (and then remixed by the engineer). Not saying placebo effect isn't real, especially if you're not doing a blind test,
but the reality is that a lab test isn't the same as a real world test.
This isn't even exclusive to audio, it just behaves differently. Take imaging for example. You can take an ultra-wide angle lens and take a shot of a test chart with lines coming off a single point, and sharpness at the corners will be extremely obvious when taking a photo or those concentric lines in black over a white background. Take it out to the real world where the corner of that photograph can be foliage and even a longer focal length lens without corner issues now has to deal with wind moving the foliage, or surfaces that don't have extremely obvious contrast and detail, like a wall in the corners. Does this make the scientific test irrelevant? No, not when you can't test a lens on every location you plan to use it on before buying it, so you might as well see a guy compare on a test chart and decide on which one is better for the addiitonal cost (and surprise...in some cases you either spend a lot more to do better on that test chart, or spend more and still do a bit worse in the corners or some other part of the frame, like real world non-noticeable issues on a Korean 12mm lens which at least lets in more light than a 10mm to 24mm/18mm zoom lens from the camera body manufacturer that at best can do f4 aperture at 12mm).
I am also wondering if the audio quality is the same when routing my phones through the speakers as it would be running it directly through the headphones jack on the front of my CPU (Dell XPS). I tried doing some A/B testing listening to the same song on iTunes and the sound seems to be slightly more 'boomy' coming through the speaker jack, and slightly less detailed. Maybe I'm wrong?
Hard to be absolutely certain to corroborate that it is not placebo if I have not heard those two.
However since they are entirely different circuits with more compromises than your average dedicated amps (if within comparable price points), then it's entirely possible that that is not placebo.
It's kind of like why I check GSMArena to make sure I'm not getting a Samsung that measures badly other than power output (then again I don't need more than 5mW of clean power for a 117dB/1mW IEM), if at least back when they still have 3.5mm analogue outputs. They actually measure better than LG in terms of distortion and noise, the LG just ditches those for a module that can deliver way more power (that I don't need but will have to go with thanks to Samsung ditching the 3.5mm jack).
I just got some computer speakers, Logitech Z333. They come with a wired 'puck' for volume control that also has an input for headphones. I noticed that when plugging in my Beyer DT 770s (80 ohm), I have to put the puck's volume all the way to the max, or else sound will only come through the left side of the headphones.
That's because the Logitech has to use hardware analogue potentiometer that while having a more direct feel to the control all suffer from channel imbalance to some degree. The motherboard uses digital signals to tell the DAC chip to keep the signal voltage of what the output stage will eventually work on lower than its max ie what it was actually recorded at.
This is why for example you will notice that some amps will advertise that they have an Alps Blue potentiometer, which, short of a really high sensitivity IEM or even some headphones (depending on gain) will not have channel imbalance once you're a few degrees of rotation past the 0 mark (note this is not 0dB like on a digital pot, like on a home theater receiver). Or how Meier uses an analogue amp with an analogue knob but that knob has solid endpoints and its rotation is interpreted by an ADC chip (not an audio ADC, this is more like on on analogue gaming peripherals) to translate into the more accurate digital control (note: this is an amplifier, it does not work by reducing the digital audio voltage, just on how much the preamp circuit pre-amps the signal into a higher voltage) but ultimately it's to avoid using an analogue pot and have a medium gain and high gain instead of the ridiculous -10dB gain on older Meier amps (which were necessary for Grados and IEMs to have enough travel on the pot).
