Date: Sunday , January 13, 2019
Lately we have produced a full series of reviews covering NVIDIA RTX Ray Tracing in Battlefield V. A lot of the BFV performance content you have seen at other sites has likely not covered real in-game 64-player performance. We opted to use actual 64-player gameplay, like most of you will likely be playing, in the same BFV map that NVIDIA uses to show off its ray tracing effects; Rotterdam. We did three reviews each covering the performance of a specific RTX card. We showed multiplayer performance of the RTX 2070, RTX 2080, and the RTX 2080 Ti.
Immediately after our first RTX BFV review, we did get criticism from our readers that our 5GHz overclocked i7-7700K (4C/8T) was holding our performance back since it did not meet the recommended specifications from EA of an i7-8700 (6C/12T) processor, and quite frankly that is a fair concern. We have spent a lot of time over the past 20 years testing game performance and did not think our CPU was an issue, but we did want to address those concerns. We could have been wrong, and we wanted to know if our testing was "off." We then purchased an i7-9700K (8C/8T) and tested again, and gave a review of performance with that CPU. Long story short, the 7700K at 5GHz was "faster" than the 9700K, and we did not see BFV multiplayer performance being harmed by our 7700K in any way.
After our 9700K testing, we got a bit more criticism, telling us that since the 9700K is still "only" an 8-core / 8-thread processor, and not the 12-threads that is recommended by EA, that our results were being hindered, again, by the CPU. While it is a fair criticism, we did not think this was an issue, but again, we wanted to know if we were being CPU-limited in our testing. At that point we shipped Brent our i9-9900K (8C/16T) for testing. Since the 9900K is a 16-thread processor, we should finally be able to put all the scuttlebutt to bed.
We will be using the latest version of Battlefield V including the DXR Performance Patch and the latest December 11th patch. All of these framerates were taken from real 5 minute 64-player gameplay so of course there is variance in runs. We have played through no less than five times, and then took the average of the three runs that we think best shows actual performance while gaming.
The i9-9900K is running at 4.7GHz with a 3600MHz memory clock under BFV multiplayer across all cores. Our 7700K is clocked at 5GHz across all cores. We are using the ASUS ROG STRIX RTX 2080 Ti OC in order to give us as little GPU limitation as possible.
We did more testing than what is shown here, but we wanted to focus on 1080p resolution since that is a more CPU limited resolution. As resolution scales, so does GPU workload, and we wanted to stay as "CPU limited" as possible for the purposes of this review.
We have seen some big swings in performance in BFV between DX11 and DX12, so we wanted to give that a look first.
We do see the 9900K perform faster in DX11 than the 7700K however, "faster" may be a bit of a stretch, as our results are inside the margin of error, but overall, we did see this as being repeatable. DX11 is not the API used for NVIDIA Ray Tracing, so that data point is of not much importance to us, but we did think it was worth sharing as just the API is part of the problem with BFV ray tracing gameplay.
We see a DX12 result in which the 7700K is faster than the 9900K, however once again the delta is small, but it is certainly outside our margin of error. The decrease in framerate is ~7% exhibited by the 9900K.
Worth noting is that DX11 is still a good bit faster than DX12 in BFV as we have seen through all of our testing.
Low DXR is the least taxing NVIDIA Ray Tracing setting that you can play with. Keep in mind that our 7700K is running at 5GHz across all cores, while the 9900K is running at 4.7GHz across all cores.
Our slower, yet much more widely threaded, 9900K gives us a ~10% decrease in average framerate.
Moving to the most resource intensive level of DXR in BFV, Ultra, we see our scaling become almost non-existent, as you might expect. Our delta is less than 1 frame per second. This is of course showing us that we are GPU limited, even at 1080p with the NVIDIA Ray Tracing features turned on. If high levels of NVIDIA Ray Tracing required more cores and threads, we would see that here assuredly.