Morning. I have an iMac Pro with 8 core Xeon processor, 32 Gb RAM. I stream music (mainly Tidal) to a NAD M32 digital amplifier connected to my Fibre Optics router using an Ethernet CAT 7 cable.
My question is whether is should bother to use the Roon Sample Rate Conversion at all in the DSP Engine or just leave it disabled. The M32
amp upsamples to 48 KHz, 96KHz or 192 KHz (user choice) at the final digital to analog phase so would using the Roon feature be a waste of time?
Have a listen. See if you can hear the difference.
Iāve recently tried it with my Kef LS50W and I prefer the sound up-sampled. (I use max PCM, Smooth, minimum phase) and my perception is a slightly less edgy sound, slightly more āanalogueā like. Things like jangly guitars, snare drum hits, some female vocals, for example, sound nicer to listen to for long periods.
I tried it with a Sonos system and canā t really hear a difference.
Try it. If you prefer the sound, then go for it! 
Thanks Dan - I suppose thereās no real substitute for spending time trying things and listening 
Yep. Thereās no technical reason why you canāt try up-sampling. Your core is definitely powerful enough. Your network is hard-wired so should be able to cope with the increased data rate.
At the end of the day, itās down to whether your D-A sounds better with max resolution (192KHz in your case) delivered to it, or if it sounds better if it does the up-conversion itself.
I thought my Kefās would do a great job upconverting themselves as theyāre a reasonably expensive bit of kit and I thought theyād have a great D-A.
I did a bit of an A/B test when I first got them about 15 months ago and didnāt really hear a difference.
However, I recently stumbled across a post on here about up-converting MQAās and them sounding better, so I tried it, and, to my surprise, my own 44.1/16 flacs sounded better as well.
Maybe now Iām really used to the sound of my system, and the speakers are run-in I can hear the difference. I donāt know. Itās a really subtle difference, but itās definitely there. A bit like a āmetallic sheenā has been removed and there is more air around things and less harshness.
I see the M32 looks to be a great bit of kit. You say youāre using Ethernet, are you using the BlueOS module with it?
I find Roon upsampling sounds slightly worse in the context of my system (Mac Mini -> Linn Klimax Katalyst), especially at high volume. I guess it is because of the bit depth conversion - up to 64 bit and then dither down to 24 bit, thus losing precision. It would have been better if the conversion is from 64 bit to 32 bit, but my streamer does not support 32 bit depth.
Hi Dan. Yes Iām using the BluOS module with the M32 and Roon integrates perfectly with it.
I do get a bit confused by some of the processing that happens - for example with an MQA file it does a ā24 bit to 64 bit floatā conversion but then almost immediately does a reverse of ā64 bit float to 24 bit with MQA signalingā conversion.
Iām getting great sound but just wonder why it bothers to do the first conversion - unless once again thereās a specific setting Iām using thatās causes it to do this then reverse itself. 
Are you using volume levelling or any filters in Roon DSP?
Yes Iāve just tried pushing low bass frequencies by 0.5 dB - from 30 Hz to 75 Hz and use the headroom switch. This is just to hear what they do and may switch off the filters completely
Thatās why the 24bit to 64bit float conversion is happening.
Itās much more accurate to do anything mathematical, such as EQ, to a sound data stream if the process is done at 64bit floating point, then converted back again.
It reduces ārounding errorsā.
Thinking in Decimal, Itās like, if you wanted to divide, say 11 by 2, but you werenāt able to use decimal points. The answer would be 6 (rounded up), which is inaccurate.
In Binary, describing a 24bit signal with 64bit float, is like increasing the number of ādecimal placesā if you see what I meanā¦ I know Iām confusing the two, but itās the best way I can think of describing it!
That makes total sense! Thanks again
Depends a lot on your DAC. About 1.5 years ago I tried the T+A DAC 8 DSD. I upsampled to DSD and it was spectacular - I can only imagine the great sound at DSD 512 (which is what the reviews indicated was its strength in soundstage). There was great 3-dimensionality to the music, although a bit less saturated (more airy, less thick). Probably made a mistake in returning it (at that time I had a lot of volume restrictions and I had to turn up the volume to appreciate the soundstage at its best).
I now have the Oppo UDP 205, and overall itās a great PCM streamer. Back then I had B&W CM 9 S2 speakers. Now I have the B&W 802 D3 speakers and wish I had the T+A DAC 8 DSD. Oppo 205 is very solid with PCM files. But with DSD there is almost no perceptible difference and it really doesnāt convert / render DSD the way T+A DAC 8 DSD did.
I also have the Chord Hugo 2 and it doesnāt benefit from sample rate conversion. It performs best when the native file resolution is sent to it for its proprietary FPGA analysis and conversion Itās really a better headphone amp. As a stand alone DAD for the larger system it lacks the 3-D sound stage (more flat front to back, but overall quite solid sound quality).