After a less encouraging debut of real-time ray tracing in Battlefield V, Nvidia and DICE worked together to optimize the implementation of DXR in the game. The improvements come in the form of new graphics drivers and a game patch that Nvidia claims can improve performance by up to 50%.
As you all know by now, Raytracing in Battlefield V was a disaster after a long anticipation at launch. The implementation was plagued by technical problems such as poor reflection resolution, wrongly reflected things and a lot of noise. If you activate the ray tracing function even at the lowest settings, the frame rate of the game is halved or even reduced to a third, depending on the GPU and resolution. This made the RTX 2080 Ti the only card suitable for ray tracing, and it used the Low setting at 1080p.
With the update from Nvidia, however, RTX 2080 Ti owners can enjoy ray tracing with 1440p 60 FPS on Ultra, RTX 2080 owners should be able to run 1440p 60 FPS on medium and RTX 2070 owners are set to 1080p 60 FPS medium. Based on our results so far, this seems to be a ~ 50% improvement they claim, and of course we run our own tests for review.
However, the big question still remains: How did DICE and Nvidia improve performance and in particular downgraded visual quality to achieve this? Nvidia has documented some of them in a video that explains the changes (see above). I'm not going to list all of them here, but some of the biggest ones include optimizing the variable rate ray tracing feature, adjusting the denoiser, and fixing raytrace sheets.
Variable rate ray tracing is the most important change. This allows DICE to prioritize ray tracing in areas of the game world that benefit most from the effect, e.g. B. reflective or shiny surfaces. In essence, the game increases the quality of beam tracing on these surfaces at the expense of a lower quality effect in other areas.
Since many surfaces do not need a ray tracing function at all, this can speed everything up. The foliage system was also a major performance issue, as DICE claimed that one mistake caused too many rays of leaf elements to be thrown. This has now also been fixed.
Visual quality impressions
So we're going to do some visual quality comparisons first to see if anything has changed, starting with some of the areas we've recently tested. If we look at the Ultra DXR setting in the previous version compared to the latest patch, we see no visual difference at all. It seems that one of the areas DICE couldn't improve is the quality of the reflections on matte surfaces like the gun. The effect is still a bit pixelated there.
The same applies to the Low setting: no real difference. In this scene, we see no significant difference in visual quality between low and medium or between high and ultra, although you may notice that the two higher settings, like last time, produce beam-tracked reflections on more surfaces.
Again, in this large puddle of water scene, there isn't much of a difference between all four DXR modes, and some of the problems with incorrectly reflected godrays still seem to be present.
However, there is a downgrading of the visual quality that I have noticed in this area. This new patch introduced some artifacts in how to deal with raytrace reflections. Everything that should be reflected was previously reflected, so you get these beautiful, clean, and mostly extremely accurate reflections that blow away the reflections of the screen area for visual quality. With the latest patch, however, some reflection artifacts on the screen have sneaked back into beam-traced surfaces.
If you look closely at the surface of the water, an object sometimes moves across the reflected area like an AI character or a falling leaf, and for a brief moment you will see the classic reflection strip on the screen caused by the obstruction of that object the way of reflection. I suspect here that DICE has chosen to sort out rays from invisible objects more aggressively, which has significantly improved performance. However, this is at the expense of the occasional artifact, in which something that should still be in sight is incorrectly discarded. It makes the reflections a little uglier, but it is still a significant improvement to the basic reflections on the screen, where this problem is much more common.
The other change appears to be a widespread modification of the particle system that makes leaves less likely to hover or land on reflected surfaces. I wasn't 100% sure, but I looked in different scenes and found that the same number of particles are generally visible in the air and on matte surfaces, but leaves don't cover puddles of water as often as they used to, and this is the case Case whether DXR is switched on or off. I even accidentally spotted some leaves that landed on puddles but then quickly disappeared.
While this change affects the look of the game without raytracing enabled, I think it's overall positive because particles also caused a lot of artifacts to reflect on the screen area. It is therefore good to mitigate a portion of this effect. And I'm not sure you would have noticed if I hadn't pointed it out.
As for the noise, the DXR reflections are still quite strong depending on the surface. In particular, it seems that DICE could not solve the extreme noise on larger water surfaces. The rippling waves are just too much for the clearer and what is left is a lot of ugly noises.
However, the noise has been improved for cleaner surfaces such as windows. In her video, Gamers Nexus showed this tram window, which had a lot of noise in its reflection when you move. With the latest patch, you can still detect noise when moving, but the descrambler responds much faster, so the image contains less "noise ghosts". In other areas, you can only detect sounds for reflected trees, while static objects stay nice and sharp. Again, I suspect that the foliage noise problem is due to some of the changes they made to improve raytracing foliage performance.
In our previous benchmark function, we struggled to tell the difference between the four DXR modes and came to the conclusion that there were really only two modes. On closer inspection, however, there seem to be subtle differences between the four modes in terms of reflection resolution and drawing distance. The higher the setting, the cleaner the reflections look and the more distance you get. It's all very subtle, so Medium looks almost identical to Low for the most part, and High Ultra looks very similar.
Overall, we would say that the patch largely did not change the visual quality of DXR, although clear adjustments were made. Most of the work seems to be aimed at ensuring that ray tracing is not used unnecessarily. Therefore, the performance improvements are due to this, although there are still some notable areas where further work such as reflection noise is required.
So let's talk about performance. First, you need both the latest Battlefield V patch and the new GeForce 417.22 driver from Nvidia to improve DXR performance. The driver alone seems to improve ray tracing performance by ~ 6 to 8% with the pre-patch version of the game, so Nvidia has definitely done some driver-related work to improve DXR performance. We noticed this in a couple of preliminary tests, but then we applied the game patch that further improved performance, and this is the new performance measurement we're reporting. The new driver had no impact on the "RTX off" performance, which is identical to our first DXR test.
Again, our tests are much more comprehensive. We tested the RTX 2080 Ti, RTX 2080 and RTX 2070 at 1080p and 1440p on our Core i7-8700K test bench overclocked to 5.0 GHz. After the patch, we tested all four DXR presets so you can see how the differences in performance stack up. We also tested two areas instead of just one: the first is the Tirailleur area, which we used in our first article. This is a bit of a ray tracing stress test and has shown the worst performance in the game. But we also took into account the performance in the Nordlys area, a snow-covered map that has practically no ray tracing function and, frankly, looks very similar, regardless of whether you turn DXR on or off.
First, we'll look at the RTX 2080 Ti performance in the intense Tirailleur card at 1080p. We expect an average frame rate improvement of 57% for Ultra DXR mode and a 21% improvement for Low mode. Despite a slight reduction in frame rate for DXR off, this mode is still 75% faster than Ultra DXR and 53% faster than Low DXR. There used to be more than one 2x difference, but this penalty is still pretty brutal.
At 1440p, it's similar: a 51% improvement in ultra performance and a 29% improvement in low performance. "DXR off" is more than twice as fast as Ultra, while there is a 53% advantage over Low.
Unfortunately, the situation on the Nordlys map is worse. Yes, there is a performance improvement at 1080p, but Ultra is only 18% faster and Low is only 5% faster than before and after the patch. In addition, DXR off is still 56% faster than DXR Low and in our test area there is essentially no visual difference between the two modes. The game still has to throw rays, even if it has no influence on the graphics, which leads to a massive drop in performance.
At 1440p again a similar story: the performance improvement is a bit higher, but DXR off is still much faster.
The good news when looking at these RTX 2080 Ti results is that Nvidia's performance requirements are right: you can play in Ultra DXR mode at 1440p and expect 60 FPS. With DXR Low at 1080p you can also achieve 90 FPS. In both situations, turning DXR off delivers well over 100 FPS, so ray tracing is still a major performance issue.
Time to take a look at the RTX 2080. In Tirailleur at 1080p, the improvements for Ultra and Low compared to the game pre-patch were 64% and 30%, respectively. However, with DXR disabled, the game is still over 50% faster than DXR Low.
At 1440p, Nvidia rightly says that the RTX 2080 can deliver an experience of 60 FPS in medium DXR mode. Performance improvements are similar to those we've already discussed, but again, DXR Off performance is more than 50% faster.
In the less intensive Nordlys with 1080p, the increases are modest, as with the RTX 2080 Ti with less than 20% for both the ultra and low modes. If the theme is repeated, the performance will be improved again by more than 50% when the effect is deactivated. The situation is similar with 1440p on this card.
The RTX 2070 is a GPU that we described last time as unusable for ray tracing because it couldn't even reach 60 FPS in low DXR mode at 1080p. After the patch in the Tirailleur area, we expect growth of 79% in Ultra mode and 33% in Low mode. The low mode is suitable for 1080p 60 FPS, as Nvidia indicates. Yes, you guessed it, performance is ~ 50% higher when DXR is off than in low mode.
It is probably not worth talking to the RTX 2070 about 1440p DXR as the GPU with DXR on Low is hard to reach 60 FPS, although this is a big improvement over what we saw before. If the DXR is deactivated, the performance is 42% higher here.
And finally we have the Nordlys card. At 1080p we see 22% increases before and after the patch compared to the low mode, while Ultra gets a 25% boost.
Ray tracing: take two
Nvidia's performance claims are spot on: a 50% increase in ray tracing performance in patched Battlefield V is real, and what they say is possible with their various RTX GPUs, no less in the more intense areas of the game. We are now in a situation where the RTX 2080 Ti can use DXR Ultra with 1440p, the RTX 2080 is suitable for DXR Low with 1440p or Ultra with 1080p and the RTX 2070 can now process 1080p with DXR Low. This is particularly important for the RTX 2070, which was previously hopeless. But there are many reservations about these claims …
In Ultra mode, only a 50% increase in performance is achieved. Gains are less significant in low mode, closer to the 30% mark. And these improvements only exist in areas where there are many DXR reflection effects. In areas where no reflections are visible, you will notice more modest increases in performance of 25% or less.
… a 50% increase in ray tracing performance in patched Battlefield V is real (but …)
And there are still problems with the ray tracing function in Battlefield V. In the Nordlys area we tested and in other areas with few DXR effects, the patch does not improve the performance significantly. This is a big deal as we are still able to turn DXR off to get 50 +% more performance for virtually identical graphics. When DXR is enabled in these areas, only computing power is wasted, reducing frame rates from ~ 130 FPS to ~ 80 FPS for zero visual improvement.
In other game areas where ray tracing is visually appealing, the performance difference between turning DXR on and off is still huge. At best, you want to get 50% more performance with the DXR off. However, if you use the high or ultra mode for the same comparison, this gap increases more than twice, depending on the GPU and resolution.
… is now the discussion of whether ray tracing is worthwhile in other games.
In general, we still don't think Battlefield V is a good game for ray tracing. As a technology showcase, of course, but for the actual gameplay, this is a fast-paced shooter that you want to run at high frame rates. Why would someone with an RTX 2070, for example, choose to play at 1080p 60 FPS with low DXR compared to 1440p 80 FPS with DXR turned off? Why should an RTX 2080 Ti owner commit to 1440p 60 FPS if you could run the game at 4K at higher frame rates or at 120+ FPS at 1440p? It is not a sensible choice for this type of game.
And when you play in multiplayer, it's almost impossible to tell if DXR is on or off. We have shown a lot of static comparisons, nice slow close-ups, and the like here, but the reality is that when you play the game, you don't stop looking at the nice reflections, but focus on enemies and targets. For multiplayer modes, we not only recommend disabling DXR, but also lowering other quality settings for maximum performance.
With this in mind, it is now being discussed whether ray tracing is worthwhile in other games, such as single player titles like Shadow of the Tomb Raider, where you have more time to appreciate the graphics. While Nvidia and DICE have clearly been working hard to improve ray tracing performance in Battlefield V and we no longer have a half to a third of the performance as a base, the drop in performance in other games will still be brutal.
If at best you can increase your frame rate by 50% by disabling DXR reflections, I can't imagine how high performance will be if games incorporate other DXR effects. For example, Shadow of the Tomb Raider has been set to use DXR lighting and shadows, which will only further improve tank performance. What if a game doesn't use the tweaks DICE has achieved in Battlefield V? We could see more losses in performance again. And without taking into account the noise and other problems with beam tracking at the moment.
We sincerely hope that Nvidia can continue to work on and optimize the ray tracing feature to get more performance without sacrificing graphics, as the technology is not yet ready, not for games. We welcome their efforts and understand that this is a first generation GPU that integrates such a complex rendering technique. Regarding what you're going to buy tomorrow, ray tracing cannot be a major selling point for Nvidia's RTX cards, a luxury bonus, but not something to base your buying decision on.