Let Roon upsample vs let Lumin T2 do the work


I've reluctantly invested in Roon for my Lumin T2.  I read somewhere to turn off all the processing in the Lumin app and let Roon do all the upsampling.  What's the benefit of that?  I figure my Lumin is built specifically for making the best signal possible. 

Secondly, my Roon core is an Asus ROG desktop computer, is it doing any work?  It's certainly not audiophile grade.
dtximages
Isn't the ROG a gaming computer? It should have the power to run roon with upsampling unless someone is playing a game on it. 
Roon is a little unusual in that all the DSP is done by the core.  So, upsampling and EQ and room correction happen there. If you are using a few EQ bands, with PCM to PCM upsampling the CPU is very little. On the other hand, upsampling to DSD can be ferocious in terms of CPU usage.  I use about 6 EQ bands and double the incoming PCM rate and my 7 year old AMD A10 CPU barely hits 5% usage.

Which to use?  I don't know, but not all upsampling is equivalent. The algorithms can result in different results.
T2 owner here using Roon. Yes, you use the Lumin app to turn off upsampling and do upsampling function with Roon. I've used Roon to upsample to DSD512 in the Lumin without any dropouts. If you get dropouts that will mean your ASUS server doesn't have enough computing horsepower. You may or may not like the DSD 512 presentation.
Yes it's a gaming computer with 16gb of ram and i7 processor.  It has plenty of power to do whatever, but does it do it the same as the Lumin?
@tuberist thanks for the input.  Why use the computer/Roon instead of the lumin though?
Also, why would one not like the DSD512 presentation?  Isn't that just higher quality or truer to the original?  Isn't it just a higher accuracy basically?
Also, why would one not like the DSD512

Remember that upsampling is not real DSD. I mean, you can make more data, but it's fake.  It's not actually carrying any more information, but it will tinker with the DAC's mechanics, and change the noise profile.


Then how does one get real DSD?  Download each individual song from some online music service? 
DTX, Peter Lie, the head Lumin software engineer, specifically instructs to turn off upsample processing in the Lumin and utilize upsampling through Roon. "Real DSD" is simply a presentation of some music sources although I have yet to find any in my library or browsing. However, if "native DSD" is available the Lumin will process it. BTW, Erik was correct in his upsampling comment. I suggest you check the Roon website, "community" and peruse the Lumin hardware  section....lots of useful info there. 
Sorry. To the DAC, native DSD and upsampled DSD look the same, in that they follow the same format, and the DAC doesn’t know or care where it came from.
What I meant about "real DSD" was a recording released as DSD to begin with, or "native." Of course, from the data in the USB, fake (i.e. upsampled) DSD and native DSD look the same. I’m sure the Lumin handles them well.


But if you are upsampling, lets say from redbook (44.1kHz/16 bits) to 192/24 or DSD, you don’t get closer to the recording, you are fiddling with the DAC and output filters, and what is an audiophile except a knob fiddler? :)


But because of this "fiddling" there’s really no right answer about whether to upsample or not, and how. It’s very much in the range of personal taste.


I’ve tried DSD upsampling and with my setup (Roon / Mytek Bridge) the top end sounded harsh and unnatural, but 2x PCM sounds VERY nice indeed. Take a look at Figure 7 of this page, to get some idea of what’s’ changing:


https://www.stereophile.com/content/mytek-hifi-brooklyn-da-processor%C2%96headphone-amplifier-measur...


Of course, 44.1 kHz data cannot support a signal higher than 22 kHz, but in the 10kHz to 20kHz range you see some differences in how the data rate affects the filtering.
Looks like 192kb just plays higher with zero difference in the audible range.. I have no idea how to interpret that graph not being an engineer :)


Upsampling creates more steps and the analog conversion becomes much smoother. The real difference happens when there are any cliffs in the sound. In that case, more granular data due to upsampling provides better transitions and smoothness. 

 

Check this video - https://youtu.be/SdPU2TZylSs?si=uVxV-Lgu7yOjaEsx