Just pulled the trigger on a Radeon VII video board plus a 9900k i9 and what a combo. Cpu is air cooled with an NH-D15 and video card is stock air cooled too. Red Car comes in at 15 seconds compared to 25 seconds on my e5 Xeon paired with an old RX580. Curiously a NUC I built 6 mos ago beats it by 4 secs transcoding a 57 minute video from mxf to mp4 with only trims and splices but no fx... I'm guessing because the Nuc's 4gb Vega GPU is bonded into its otherwise unremarkable 8809g cpu; transcoding between cpu and gpu must be happening without even hitting the pcie bus. But the real shocker to me is the transcoding time I clocked for a 3 HD camera merge with color correction on each individual camera track. That's a step in my workflow after choosing camera angle crossfades on a multicam track and expanding it back out to 3 tracks and applying fx individually to match color shot on different cameras having different color depths. Almost 2 hours rendering time on the old system down to 28 minutes on the new one! Which is even quicker than the next step of transcoding trimmed and spliced mxf to mp4 deliverable which includes opening screens, titles, and rolling credits. This also resolves another mystery to me... my renders of a really long festival that was shot without pause from opening to intermission kept crashing at around the 90 minute point on my old system if I used gpu. Turning off gpu worked but rendering to mxf took over 7 hours. My more practical workaround was to break the 110 minute run into 2 adjacent 55 minute regions and render them separately w/gpu then glue them back together with grouping in the next step. I thought it was Vegas screwing up but it rendered flawlessly without splitting with the Radeon VII. So my new theory is that my RX580 with 8gb of internal memory ran out of table space which the VII with 2x more memory did not. Other observations... the Xeon still rules on floating point but Vegas must use only integer arithmetic. Also, increased cpu and ddr4 memory speed have less impact than I hoped. Same with the m.2 pcie ssd; I got identical run times on Red Car putting everything on an internal 2tb Samsung m.2 pcie 870 EVO as I did putting it on a 2 tb Samsung EVO 850 2.5-Inch SATA III drive connected to a sata3 port. And for the 57 minute color-correction render, putting everything onto and rendering to pcie m.2 only showed a 20 second improvement over the sata ssd.