Let’s start with 720p. Actually, this is supposed to measure CPU load and bottlenecks, but it is the amount of drawcalls that makes the bus interface interesting again. And so it is not surprising that the greatest differences can be measured here. With a massive 6.35% performance loss, the card at PCIe 3.0 @8 loses a lot of performance compared to the run through @16, despite a pronounced CPU limit. The difference in both PCIe 4.0 measurements with the Radeon RX 5700XT is, however, with 0.37% still clearly below the one percent that one can already allow oneself as a measurement tolerance. Since the RX 5700 XT is inherently much slower, a difference of only 4 FPS to the Quadro RTX 6000 @8 documents this disadvantage very clearly
You can also see the difference quite well on the FPS curves:
Now let’s take a look at the percentiles, because the 99th percentile is the most important one. Percentile isn’t everything. What you can see is the supposed break-in of the Radeon, no matter with which connection, when it’s over the 99. …goes out. But there are other interesting details!
The frame time shows that the supposedly faster Quadro RTX 6000 @8 gets the short end of the stick, because the Radeon’s course is much more even. Since superimposed curves only confuse, I will have the whole progressions later as individual graphics for each map.
The evaluated variances can then illustrate this in more detail. There the values of the Radeon are much more balanced!
Individual graphics for each run as picture gallery
Kommentieren