People saying that this is a bad comparison due to the game being "unoptimized" are not looking at this correctly. This is perfect case for comparison for a couple of reasons:
- This is a game that uses a high unlocked frame rate across all modes. This will truly show how far each system can go with this particular workload. Most future comparison will be a lot less interesting since the frame rate will be capped at 30 or 60fps and we will rarely see 120fps modes w/Ray Tracing enabled.
- The performance you get from simply porting your game and adding features says a lot about how accessible the power/performance is for the system which I would argue is more important than raw theoretical power. This says more about how efficient and/or bottlenecked the system may be. This is why many Xbox 360 games looked and ran much better than the PS3 counterpart despite the higher theoretical performance in the PS3.
And what did we learn from this first comparison? Not much new...it's just really cool to see the results in action:
- Performance on both system is virtually identical and within general margin of error (+/- 5%). The DF video was not a 1:1 frame mapping across both platforms so you can't just freeze a frame of that video and look at the frame rate to compare.
- Neither system appears to be CPU bound at any point (CPU performance looks pretty much identical)
- Yes, XSX had a very slight overall advantage in the default 4K and RT cases. But again, the differences is so small that it is virtually imperceptible during gameplay (especially at 100+ fps). But yes, we knew that all things being equal the XSX might have a slight GPU advantage (much smaller than the 18% theoretical would suggest)
- PS5 has faster performance at times (in gameplay)
But the key takeaway here: the large drops in performance on the XSX that the PS5 does not have point to a bottleneck somewhere on the XSX that is not present on the PS5. This is the point that Mark Cerny and numerous developers have hinted at for months now: The key principle for the PS5's design was to remove as many bottlenecks as possible and design the most
efficient system possible. Whether it's the CPU, GPU, RAM, SSD, I/O, or the SDK it is all designed to be balanced, consistent, and accessible.
Rich from DF dismissed the XSX performance drops to not being a hardware issue but probably being an issue with the API/SDK. This is possible, but I wouldn't dismiss the hardware completely, particularly the memory bandwidth. Remember, the XSX RAM is not an even split and it's possible that the way DMCV uses the RAM, there are some performance sensitive data that may have had to extend beyond the 10GB "fast RAM" pool when running at that high of a framerate. Without a local profiling tool, it'd be difficult to say for sure but I'd suggest it's possible. If it is an API issue, then the drops could be fixed by a future update. I guess we'll see in time.
In the meantime, we're seeing pretty much identical performance when both are pushing their framerates limits. XSX is slightly ahead in most cases but PS5 is noticeable ahead in others with the XSX showing a potential bottleneck with high framerates that is not present on PS5. So PS5 "punching above it's weight" based on the theoretical specs and showing better consistency in performance. Pretty much exactly what was suggested