During this year's GDC, Nvidia announced that a driver update would give GTX graphics cards basic support for ray tracing. A few weeks later, this is the case for cards of the GTX 1060 6 GB, the recently launched GTX 1660 GPUs as well as the Titan X, XP and V.
Previously, you needed an Nvidia Turing RTX graphics card to take advantage of the visual benefits of the ray tracing function. For various reasons, which we will discuss later, you can now access the function with many other GPUs. To put this test together, we took the most powerful Pascal GPU we had on hand – the Nvidia Titan X – and compared it to Nvidia's RTX lineup in the three games that previously supported ray tracing.
Now we all know that ray tracing on Pascal GPUs sucks. So please don't be surprised. If you are using a GPU that does not have specific acceleration built in for a very powerful graphics effect, this will happen. However, there are some interesting questions about ray tracing on Pascal that we wanted to answer.
The first question is to what extent Pascal will embarrass himself. Is a card like the Titan X faster than the slowest RTX GPU, the RTX 2060? Nvidia has shown some benchmarks that give us some pointers, but we would like to check this based on our own data, and in particular consider not only average frame rates, but also 1% lows and general performance fluctuations within a game session.
The other question is whether you can still get acceptable performance with a Pascal card in ray tracing, although you know for sure that the performance will generally not be great. For example, can this card run a game with over 30 FPS at 1080p and acceptable ray tracing quality? This would allow some players to really try ray tracing without being forced to watch a slideshow.
For these benchmarks, we used our Core i9-9900K test bench in conjunction with 16 GB DDR4 memory. All data was collected using the latest versions of the games with the latest drivers installed. Games like Battlefield V continue to optimize their ray tracing performance. Therefore, using these benchmark data is critical in these tests. You will see the areas that we tested in each game as soon as we reach them.
Let's start with Shadow of the Tomb Raider … the latest game that includes ray tracing through the shape of shadows. When we tested this game, we found that only one of the three ray tracing modes makes sense, and that's Ultra mode. In our opinion, the high mode produces a poorer visual quality than the deactivation of the ray tracing function, while medium only offers a very limited scope for ray tracing shadows. So for today's tests, we stuck to Ultra mode.
As expected, the Titan X fell between RTX 2070 and RTX 2080 when the ray tracing function was disabled and all other settings were set to their maximum value. However, if you enable ultra-ray tracking, the Titan X immediately drops to a level below the RTX 2060. It is 10% slower at average frame rates, but 34% slower at 1% lows.
And this shows one of the main problems with ray tracing on Pascal: The experience is very inconsistent. This is because there is such a big difference between a card's capabilities like the Titan X without ray tracing and with ray tracing. So if you move in an environment with different numbers of beams, interactions and degree of beam tracking, the performance of the Titan X fluctuates massively. Performance is decent in areas with little ray tracing, but if you are in an area with lots of shadows, your frame rate will absolutely decrease.
As you can see in the table, a 1% performance of 21 FPS is not playable, and that at only 1080p. But if you've only achieved average performance, 47 FPS sounds fine, it sounds decent. However, the actual experience of playing the game is far from it. Of course, you will also get a fluctuating frame rate with RTX GPUs, including RTX 2060, but the problem is less pronounced – the GPU is not as fast in areas without ray tracing and can keep up better when ray tracing is activated. The 1% low frame rate for the RTX 2060 was a touch above 30 FPS, which is not surprising, but a far more passable experience.
Switching to 1440p will make things worse for the Titan X. We have an average of 30 FPS and only a 1% low of 14 FPS, which is not playable. The margins between the RTX 2060 and the Titan X are somewhat lower here, since the 2060 also has to struggle with 1440p. However, since the Titan clocks in the most intensive areas by more than 30 percent slower, Pascal simply cannot keep up.
Let's take a look at a more positive game for Pascal, and that's Battlefield V. Here we recommend using low reflections that don't have as many effects as the High or Ultra modes, but it's a good place to start and nowadays the performance Hit is nowhere near as bad as it used to be.
At 1080p we see a somewhat significant drop in performance for the Titan X, which goes from well over 120 FPS with the ray tracing function deactivated to an average of only 70 FPS with the ray tracing function. However, 1% lows are hammered again, which are only halved here by low-ray tracing.
Perhaps the most interesting is that, unlike Shadow of the Tomb Raider in Battlefield V, Pascal is more competitive with the RTX 2060. On average it is a bit faster and a little slower in the most intensive areas. We're not nearly as powerful as the RTX 2080, which comes closest to the Titan X when the ray tracing feature is disabled. Due to the relatively easy implementation of the ray tracing function in low mode, the Titan X is not so overwhelmed and works fine here.
In fact, the game is playable with a 1% low of 46 FPS. Performance still fluctuates widely, but at least it doesn't regularly drop to slideshow levels. At 1080p with low ray tracing and a card as powerful as a Titan X or 1080 Ti, you may be able to play the game without pulling your hair out. Would anyone sacrifice over 100 FPS for this performance at this resolution? Probably not, but at least it is possible to try it out.
At 1440p, the situation is not as promising as 1% for the Titan X is closer to 30 FPS, which is not acceptable in a multiplayer shooter. We also see a divergence between the Titan X and the RTX 2060: the Titan is faster on average, but is much slower in the most intense areas of our benchmark run. This makes sense since more beams have to be thrown with higher resolutions and an increase in beam tracking penalizes Pascal more than Turing RTX.
The last game we see here is Metro Exodus, which uses ray traced global lighting. When we tested the game for the first time, we used the benchmark tool. Since then, however, we've switched to an in-game run and focused on high mode, which we think is best used in the game based on our previous explorations.
We didn't bother to test 1440p in this title as we're already at an unplayable level with the Titan X at 1080p. An average of 30 FPS with a 1% drop of 23 FPS is not good. And in this game there is no ray tracing level below Highl, so you can completely exclude Pascal. The Titan X is more than 30% behind the RTX 2060, which makes sense since the global lighting is an intense ray tracing effect and acceleration is very useful.
The comparison of the Titan X with the RTX 2080 shows that the RT cores offer double to three times the power when the effect is activated. We can expect a similar difference between Pascal and accelerated Turing in games that use multiple ray tracing effects: the more effects added, the more useful the RT cores become.
What does this study tell us overall?
- It was not necessary to test a GTX GPU slower than the Titan X. Two of the three games cannot be played with the Titan at 1080p.
- The only test condition that could be used remotely was Battlefield V at 1080p with low ray tracing, but even then I would expect a GTX 1080 to barely reach 30 FPS in the most intense areas and the performance would drop further from there.
- Unless you have one of the top-end GTX cards and the ray tracing effects aren't too intense, Pascal isn't fast enough to offer gamers a usable ray tracing experience. As more and more raytraced games are released, Pascal will only fall behind.
- At best, the Titan X corresponds to the RTX 2060 in terms of ray tracing functions, but often falls back by 30% or more, especially when important data is viewed with 1% lower data. The Titan also offers lower frame rate consistency. Normally (no DXR) the Titan X is at least 25% faster than the RTX 2060 and higher, for example in the area of the RTX 2080.
To what extent do the RT cores and other bits and bobs in the Turing architecture help to speed up beam tracking compared to Pascal? This is more of a matter of interest, but we now have some rough data that can give some insight.
The main comparison would have to be between the Titan X and the RTX 2080, which are appropriately coordinated outside of ray tracing. The RTX 2080 is faster, but not much. However, taking into account the ray tracing function, the RTX 2080 is between 26% faster with low reflections in Battlefield V, more than 50% faster in Shadow of the Tomb Raider and more than twice as fast in Metro Exodus.
Considering the TU104, the GPU used in the RTX 2080, which contains only 13 percent more transistors than the Titan X and the GTX 1080 Ti, we would say that this acceleration is actually quite impressive and at least justifies the additional RT cores for it Games and effects. Nvidia was unable to perform brute forced ray tracing by crowding in more CUDA cores. The data show that ray tracing with RT core acceleration is more efficient. We also don't think the level of acceleration is disappointing. More than a twofold improvement with heavy use of ray tracing is a good start for a first generation design.
Of course, in most games it doesn't help to use the space for special cores for ray tracing, but we've already talked endlessly about the value proposition of ray tracing and RTX cards …
Finally, we have to ask ourselves why Nvidia bothered to test and activate ray tracing on Pascal. It doesn't do well even on high-end GPUs, is unlikely to improve with future games, and it just seems to be something people wouldn't use, even if it were available.
We have two theories as to why: The first is for developers. Let's say you have a development studio that has invested heavily in Pascal and has a lot of cards like this Titan X in their development machines. Rather than forcing developers to upgrade to Turing to develop ray tracing games, the introduction of ray tracing to games could be improved, albeit slowly, by Pascal cards. Developers don't need 60 or even 30 FPS to test ray tracing in their games. This could be useful for you. This has a number of advantages for Nvidia as they are interested in improving acceptance to sell more RTX cards.
Second, simple old marketing to incentivize RTX GPU upgrades. In our opinion, ray tracing alone is not enough to justify an upgrade, but this works for a number of Pascal players regardless.
It's generally a good thing to unlock features for more GPU owners, and across the board there could be a ray tracing game that actually does well on Pascal. But like most of our previous raytracing articles, we're still at a very early stage in technology, and it only becomes a significant factor in a few generations.