Back in summer 2016, when I started using Resolve after many years of producing my HD, multicamera BDs/DVDs with classical music - just like so many others I couldn't accept the fact that an extra video device (in case of UHD/4K@50p, only a then USD 1,000 Decklink 4K Extreme 12G with mezzanine HDMI) was essential for OS color-independent, 10 bit monitoring. Andrew, Mark and many other gurus on this great Forum (sorry - I most certainly omitted somebody, who patiently participated in my thread on the subject at that time) must have a good laugh trying to convince me of the simple fact that the 10-bit Quadro card I outfitted my first Resolve machine was not enough to preserve the full 10-bit path from Resolve internals to my grading monitor
But they were successful; after providing me with endless explanations, test procedure proposals and tools - I finally accepted that I simply must live with the fact that a BMD video capture & output device is necessary to take the full potential of Resolve's great playback engine, if I want to trust my preview especially from the Color page...
But you know what? Even though - as I mentioned - only one of the most expensive Decklink models would let me work with my demanding video formats, I never regretted the investment for a single second! I'll tell you more: even though Andrew (if I remember correctly) was advocating strongly for the use of OpenGL 10-bit capabilities of nVidia Quadro GPUs in Scratch - comparing grades I'm able to achieve using SDI monitor on Deckling with those on the same monitor, but using the Quadro as the system's secondary monitor, I will always like Decklink more! Exactly the same impressions in Nuke Studio...
But the quality of color in my grades (including PQ ones, where possible) is not the only aspect where my Decklink shines over Windows 2nd monitor - for reasons I cannot explain, all 3 systems (Resolve, Scratch and Nuke) playback fps performance is much more solid using a dedicated preview device (Decklink) to drive my hero display, and GPU left for acceleration jobs solely.
And now - the best part of it: again comparing those 3 systems, definitely the fastest (only calling for caching when really heavy effect are used - like TNR or some OFXes) is Resolve; no question about it... So perhaps we should stop alluding to making HW money as the sole reason for BMD sticking with the "old-fashioned" system architecture of Resolve? I admit 2.5 years ago when my experience was none, and I was made to come to terms with yet another expenditure in the form of my Decklink, but after those 2.5 years of gaining experience I'm taking all those suppositions back.
Of course, YMMV and I'm not pretending the above is always as I described it - or for that matter, that it will not change in time (who knows - perhaps it will, and quite soon, too). With the current state of affairs though, I find the BMD solution - as implemented in the ever-growing DR system - definitely the best to work with.
Piotr
AMD TR 2990WX CPU | MSI X399 CARBON AC | 64GB RAM@XMP3200 | 2x RTX 2080Ti GPU | 4x 3TB WD Black RAID0 media drive | 3x 1TB NVMe RAID0 cache drive | SSD SATA system drive | AX1600i PSU | Decklink 12G Extreme | Samsung UHD reference monitor (calibrated)