I find the "a preamp makes it sound so much better" a little confusing. My only conclusion is that many people try this without doing any homework.
First - there's a huge question: how does the DAC perform volume control? Volume control, in general, is the single largest source of sonic impairment in a preamp (IMNSHO, but i have done a lot of actual research on this). So if you eliminate the preamp you are shifting the major sonic coloration from the preamp to an ancillary function of the DAC, which often is **awful** but occasionally is outstanding.
As an over-simplification: the only DACs to truly do volume control well are based on 32-bit, sigma-delta DAC chips, that allow for volume attenuation in the digital domain but, do to their 32 bit architecture, do not suffer this (significant) problems that most digital volume control does. If, for example, you use your PC or streamer to do this, digitally, before a regular DAC, forget it. The distortions are huge and built right into the math. While i explained the generic result i think the only common chip like this is the ESS Sabre 9038. Being 32 bit you can do the digital volume math, have all the truncations and still be left with more than 24 bits of resolution.
Other methods of volume control, that are cheap, and suck, include vary-int he feedback loop of an opamp, or using a buffer and a IC based resistive network. None of these are designed to be awesome - just convenient.
I use one DAC with the ESS chip directly (sometimes) and it sounds great. Its only failing is that its analog buffers don't drive my cables as well, and i think i can hear the lack of weight -- but also the tiny little increase in transparency that comes from eliminating one stage with no other changes (except the aforementioned output drive capability and stability).
So - to answer this you really MUST do some homework not just on the chip and volume method but on the analog stage. Or guess. Or listen to random people's thoughts which is basically guessing.