Allow me to be the slight contrarian here. But first, I would strongly recommend you go to the DMP beta review thread. This topic is discussed frequently and at some length.
In my experience, I thought that DMP playback had greater micro-detail, but a narrower soundstage when I first began testing. That soundstage eventually opened up, and several smart folks here wondered how this could be, inasmuch as the DMP is essentially an all-digital device. Ted and others had some speculation, but all I know is what I heard. When the DMP rolled out, the possible differences between DMP & Bridge were front and center for many of us. Paul noted that to his ears, the Bridge sounded slightly more veiled when compared to the DMP. This was the focus of my listening tests. My primary NAS is a Synology 2416RP+ with 6gb RAM and 64tb of storage. Roon runs on this server on the SSD drives. I have a second Synology running MinimServer for comparison, and for serving another DAC. The server and switches are optically isolated from the DS/Bridge. The entire audio chain is a on a separate subnet of the network, with STP and QoS optimized for audio.
It was the DMP that led me to some of these improvements because, indeed, with some recordings, the original disc seemed to have more clarity. I wanted to find out why, and to at least eliminate all possibilities other than “the Bridge and the DMP are two different transports with different computational capabilities.” Once I finished that process, network file playback through the DS, at least in my system, is audibly equivalent to the DS. And I have all the advantages of network playback and the Roon interface, with the noisy components isolated in a different location, and discs stored but accessible when I want to play them.
I still listen to the DMP, especially with a new disc, or when a friend visits, and when I simply want the experience of leaning back and ignoring the Roon environment. I was so impressed with the DMP that I decided, of course, to keep it, and to add a UMT to my MSB Select DAC which I use with my headphones.
Small clarification of something that @badbeef wrote: The primary technical reason is the very large part you left out – the SOURCE playing the file through the Bridge. And in the case of the USB input, USB is inferior to the direct I2S connection. There may be other reasons why the disc playback is best – it’s a different chain – but it’s likely losses in the delivery TO the DMP.
When I first read this, I thought badbeef was equating playing through the Bridge with playing through USB. I don't think that is what he meant, but I could be wrong. He is absolutely right about the source, and almost everything in between the source and the Bridge. Bits are indeed bits, but the computational treatment, timing [jitter], and noise associated with those bits matter. Audibly. Bridge playback, with a properly delivered source over a robust network is competitive with the DMP in my experience. In fact, I can no longer consistently discern the difference between the two. Cable differences in my speaker system produce far more variation.