Unreal Engine 4.26 – NVIDIA GeForce RTX 3070 Ti & 3080 Ti Performance

Introduction

NVIDIA has a tradition of releasing "Ti" variants of their video cards which serve as a more powerful variant of the base model and often fill gaps between product tiers. The launch of the RTX 3070 Ti and RTX 3080 Ti continues this trend. Compared to the base RTX 3080, the new RTX 3080 Ti has 2GB more VRAM (12GB versus 10GB) and 18% more CUDA cores, although the maximum boost clock is a bit lower. In fact, in terms of performance, the RTX 3080 Ti should be closer to the RTX 3090 than it is to the RTX 3080.

The RTX 3070 Ti takes a different approach. It only has a small bump in CUDA cores, just 4% more than the base RTX 3070, but has a faster clock speed paired with the faster GDDR6X VRAM. While there aren't many more cores, it should be able to utilize those cores more faster.

Professionals using Unreal Engine have a wide range of needs when it comes to the GPU. Some on the Game Development side don’t need much power because their target audience probably isn’t running the latest hardware. Meanwhile, some of those in the ArchViz or Virtual Production space may want every bit of graphical power they can get. For some, 12GB of VRAM may be more than plenty, while others need all 24GB available in the RTX 3090. In this article, we’ll look at performance in a variety of scenes to see how these new cards stack up to the existing GeForce line.

If you want to see the full specs for the latest NVIDIA GPUs we recommend checking out the NVIDIA GeForce RTX 30 Series product page. But at a glance, here are what we consider to be the most important specs:

As we noted earlier, the RTX 3080 Ti has 2GB more VRAM than the RTX 3080, and has close to the same CUDA core count as the RTX 3090. Price-wise, it is $500 more than the RTX 3080, or $300 less than the RTX 3090. On paper, it looks to be a great card for those that need high GPU performance, but not necessarily the large 24GB of VRAM that is found on the RTX 3090.

One concern we have is that this is a 350W card – just like the RTX 3090 – but uses a 2-slot cooler that is essentially the same as what is on the lower wattage RTX 3080. In our testing, this makes it both louder and hotter than any other RTX 3000 series card, which can be a significant concern if your system does not have adequate cooling.

Test Setup

Listed below is the specifications of the system we will be using for our testing:

*All the latest drivers, OS updates, BIOS, and firmware applied as of June 8th, 2021

To test each video card, we used Unreal Engine 4.26. These tests were run twice on each GPU, with the best result being included in the graphs below.

Raw Benchmark Results

While we are going to go through our analysis of the testing in the next section, we are providing the raw results for those that want to dig into the details. If there is a specific sort of project you work with, you can compare it to the scene that most closely matches your work.

ArchViz – High texture count and resolution, highly detailed meshes, baked lighting

Megascans Apartment – high-end game quality assets and lighting. Both with and without Ray tracing

Megascans Temple – Cinematic quality assets an textures. Numerous shadow casting lights. Ray tracing. High VRAM usage (15GB+)

Virtual Studio – Lower end Virtual production scene with composited videos.

Overall Performance Analysis

Overall, the raw performance of the RTX 3080 Ti is quite impressive and lines up about where you would expect given its specifications. In most cases, it nearly matches the RTX 3090. This is certainly an interesting card for many Unreal Engine users. It offers nearly identical performance to the RTX 3090 for a few hundred dollars less. On the other hand, it is considerably more expensive than the standard RTX 3080, and that extra cost may not be worth the modest 10-12% bump in performance.

The RTX 3070 Ti did not pull off as much of an improvement as the 3080 Ti did. We saw only a 7% increase in performance over the base RTX 3070. While every bit of performance helps, for only a little bit more money, the RTX 3080 is 14% faster and offers an additional 2GB of VRAM.

Averaging the performance over several projects and at multiple resolutions gives us a good overall representation of how a particular video card compares to other video cards. However, each industry has its own specific needs and requirements that can give us a more nuanced view of how useful that card is to those industries.

GameDev Specific Analysis

Game Development is usually less concerned about GPU performance than the people actually playing the games. Better FPS in your viewport doesn’t make much difference past a certain point, say 30-60 FPS. And if you can only manage 30 FPS in your game while using a 3090, not many potential consumers of your game will be able to play with acceptable frame rates. GameDev also doesn’t push the VRAM too far for the same reason. If your target consumer has 4-8GB of VRAM, then using 24GB isn’t going to help you. As we can see from averaging out the GameDev scenes from our tests, the RTX 3080 Ti is roughly 10% faster than the current RTX 3080, and almost tied with the RTX 3090.

The RTX 3070 Ti does outperform the base RTX 3070 for most game developers. Among this range, the question is less about bang for the buck, and more about how much can you get for a given dollar amount.

ArchViz Specific Analysis

ArchViz leans a bit more toward baked lighting than real-time for performance reasons. Either way, a hallmark of ArchViz is nightly detailed models and large textures. This causes a bigger need for VRAM than we see in GameDev. For this reason, we see the RTX 3080 Ti both gain in score over the RTX 3080 but lose ground to the RTX 3090. One of the two scenes used for this score saw the system crashing while running in 4k. This is likely due to a driver issue, so hopefully, that will be fixed in time. For ArchViz, we would still recommend the RTX 3090 because it has double the VRAM of the RTX 3080 Ti. Running out of VRAM will give significant performance penalties to your projects.

Again the RTX 3070 Ti outperforms the base RTX 3070, this time by 9%. However, VRAM is so important to ArchViz users that they RTX 3080 scores higher by a wide margin. Enough that its hard to pick anything lower than 10GB of VRAM.

Virtual Production Analysis

Virtual Production runs the gamut from simple scenes with minimal geometry and textures, to cinematic quality assets needing huge amounts of VRAM. On the simpler side of Virtual Production projects, we see the RTX 3080 Ti nearly tie the RTX 3090, just as we did with Game Development. On the upper end, with cinematic quality assets, we saw the RTX 3080 Ti gain ground over the 3080 due to the increased VRAM, but lose ground to the 3090 for the same reason. We also saw the stability issues here in one scene while running at 4k. The massive 24GB of VRAM available with the RTX 3090 is still the best choice for high-end virtual production workflows. See the next section for more details on high VRAM usage.

Once again, VRAM is very important to Virtual Production users. So while the RTX 3070 Ti does show an improvement, its 8GB of VRAM is limiting any other potential it would have.

A note on High VRAM usage

When the project needs a significant amount of VRAM, as is often the case with Virtual Production or ArchViz workflows where high resolutions and ray tracing are required, we see the RTX 3080 Ti gain ground over the RTX 3080, and then falter a bit. The Megascans Temple scene consumes quite a bit of VRAM. When running at 1440p, VRAM usage hovers between 12-15GB, while at 4K, it will spike just over 20GB. At 1080p, we see the RTX 3080 TI performing 40% faster than the standard 3080, and almost triple the speed at 1440p. That extra 2GB of VRAM is just enough to help.

However, at 4k things took a turn for the worse. The RTX 3080 Ti would cause Unreal Engine to crash. Usually in Unreal Engine, when the VRAM is used up it will overflow to system RAM with a significant performance hit. The exact same system with a standard RTX 3080, 3070 Ti, 3070, or even 3060 was able to complete the tests without crashing. This is an improvement over our initial tests where the whole system would bluescreen, so the latest Nvidia driver has helped somewhat. They continue to work on a better solution. Until then, we cannot recommend this card to anyone with very large scenes.

How well does the NVIDIA GeForce RTX 3080 Ti 12GB perform in Unreal Engine?

The RTX 3080 Ti offers nearly identical performance to the RTX 3090 for a lower price, as long as VRAM is not an issue. However, stability issues were seen in projects that require large amounts of VRAM, specifically those used for Virtual Production or ArchViz. Those that need less than 12GB of VRAM can get similar results to the existing RTX 3090 but at a lower cost. Hopefully, a new driver will fix the stability issues, but even so, users working with large scenes would be better off opting for the RTX 3090.

One concern we do have is that this card draws as much power as the 3090, but has a much smaller cooler. So far in our testing temperatures of a single card have been within the specs laid out by Nvidia. Currently there is no blower style card, so stacking 3 or 4 in a system isn't an option.

The 3070 Ti offers a modest 7% improvement on average. This isn't a much as would have been hoped for, especially when for a little bit more, the RTX 3080 is 14% faster with 2 extra gigs of VRAM. While it is an improvement over the base RTX 3070, it also is priced close enough to the RTX 3080 to make it appear even more interesting.

As always, keep in mind that these results are strictly for V-Ray. If you have performance concerns for other applications in your workflow, we highly recommend checking out our Hardware Articles (you can filter by "Video Card") for the latest information on how a range of applications perform with the latest NVIDIA and AMD GPUs, as well as with different CPUs and other hardware.