Nvidia GeForce RTX 3090 SLI Review – It could have been that beautiful

What could be thicker than the fastest video card of the moment? Exactly: twice the fastest video card of the moment. Armed with two GeForce RTX 3090 Founders Editions and Nvidia’s new NVLink bridge, in short, almost EUR 3,200 in GPU hardware, we are investigating what performance this monster combination is capable of.

SLI in short

The term SLI nowadays stands for a scalable link interface and was first used by GPU manufacturer 3dfx in the late 1990s. Nvidia acquired the brand name through acquisition and breathed new life into it, for a technique to combine the computing power of multiple video cards. In recent years, this has usually been done on the basis of alternate frame rendering, in which the video cards render an image alternately. An inherent disadvantage of this is that the time between images (the frame time) can vary because not every frame is ready at the same time. You perceive that as ‘stuttering’.

With the arrival of DirectX 12, everything would be different – and above all better. This API contains a technique called explicit multi-adapter, which allows a game to control multiple video cards itself. It is even possible to mix completely different video cards from AMD and Nvidia. Most of the DX12 games, however, do not apply this technique, while Nvidia officially pulled the plug this summer from the ‘old’ way with SLI profiles in the driver.

The SLI bridge that you used to get with every nice motherboard has not been used for a long time. After a high bandwidth version of that bridge, Nvidia switched the RTX 2000 series to NVLink, a much faster interconnect that originated in the data center world. The same technology is used for the RTX 3000 series, but due to the changed connectors, you still need a new bridge.

The controlled death of SLI

In the past, you could put just about any video card in SLI. AMD’s counterpart CrossFire was even more flexible on that point: you could even mix different GPUs. If we look at the last four GeForce generations, it is striking that support for SLI has been phased out further and further. In the past, for example, you could still easily combine two GTX 960s with each other, with the 10-series you needed at least one GTX 1070 for this and with the previous generation it was only possible from the RTX 2080. In the new series, only the RTX 3090 offers support. for SLI.

RTX 3000 RTX 2000 GTX 1000 GTX 900
** 90 / Titan Yes Yes Yes Yes
** 80 No Yes Yes Yes
** 70 No No Yes Yes
** 60 No No No Yes

SLI with the RTX 3090

If you continue the line, the next generation will end the story for SLI. Life was not made easy for you as an SLI enthusiast: over the past three generations, functionality has become increasingly limited while the required hardware has increased in price. The NVLink bridge required for the RTX 3000 series uses different connectors from the previous generation for the third consecutive generation. Moreover, it costs no less than 85 euros.

For this test, we built a different test system than in our regular GPU reviews. Where PCI Express 4.0 adds little according to our tests , with the mainstream platforms from AMD and Intel, the GPU bandwidth is again cut in half. This is due to the limited number of PCIe lanes. To prevent this and to minimize the expected CPU bottleneck with so much GPU power, we performed these tests on a HEDT platform, about which more on the next page.

Test justification

In order to minimize the bottleneck of the rest of the test system for the SLI configuration of RTX 3090s, we put together an overclocked system based on Intel’s X299 platform. In this way, both video cards have sixteen lanes of PCI Express 3.0 bandwidth, while a possible CPU bottleneck will occur less quickly than with our standard 3900XT test system. Intel’s midrange platform was not an option. Due to the lack of sufficient lanes and PCIe 4.0, the bandwidth per video card would then be reduced by half.

As the processor, we installed an Intel Core i9 10980XE with eighteen cores, overclocked to 4.5GHz, on an ASUS ROG Rampage VI Apex motherboard. G.Skill delivered four 32GB modules from its Trident Z Neo range for a total of 128GB at a speed of 3200mt / s. We run the OS on a Samsung 970 Evo 250GB ssd, the games are on a Kioxia Exceria + 2TB. Furthermore, the system consists of a Corsair AX1600i power supply and an extensive water cooling set from Alphacool. The latest version of Windows 10 Pro 64bit is of course installed as the operating system.

Test system
Processor Intel Core i9 10980XE @ 4.5GHz
Motherboard ASUS ROG Rampage VI Apex
Random-access memory G.Skill Trident Z Neo 128GB DDR4-3200 CL16-16-16-36
SSD Samsung 970 Evo 250GB (OS) + Kioxia Exceria + 2TB (games)
Nutrition Corsair AX1600i
Cooling Alphacool Eisbaer Extreme Liquid Core 280 + Eisblock XPX Aurora
Operating system Windows 10 Pro 64bit May 2020 Update (2004)

Drivers and measurement method

For the game benchmarks in this review, we used the GeForce 456.55 driver. We performed the ‘creative’ benchmarks with version 456.38 of Nvidia’s Studio driver.

Using PresentMon, we measure the performance in each game tested, from which we calculate both the average frame rates (fps) and the frame times of the 99th and 99.9th percentiles and report the latter two in milliseconds. Please note that this is a software measurement. In an SLI setup, not all calculated frames are necessarily displayed on the monitor.

In the graphs on the following pages you will always initially find the average frame rates or the average number of images per second that a video card can calculate. The frame times do not give a picture of the average frame rate, but of the outliers in the negative sense that can ensure that a game does not feel smooth despite a good average.

The time it takes to render images within a 3d game and thus within our benchmark varies from frame to frame. Our frametime measurement stores the render times of all individual frames. Then we discard the 1 percent slowest frames. The highest render time of the remaining 99 percent of the frames, or the slowest frame, is the 99th percentile frame time.

At the request of some readers, we also added the 99.9th percentile values. So we only leave out the 0.1 percent slowest frames for this. In theory this is even more precise, but in practice incidental causes and measurement errors sometimes throw a spanner in the works. For now, we’ve listed them in the review, so keep that in mind when reviewing these results.

Games and creative workloads

Since not all modern games support SLI or multi-GPU in general, we have made a different selection for this test than in our regular GPU reviews. After all, testing SLI in a game that we know does not support SLI is not the most useful activity.

In addition, we took the opportunity to run some ‘creative’ benchmarks, in the field of video editing and rendering. Several users therefore asked in response to our RTX 3090 review. To give an idea of ​​the performance in video editing software, we render a project with GPU acceleration enabled in both Adobe Premiere Pro and DaVinci Resolve. We also calculate a Blender scene in two modes: standard GPU acceleration for which the Cuda cores are switched on and Optix, where the ray-tracing cores are also used. Finally, we render three scenes in OctaneBench, OctaneRender’s benchmark tool.

Power consumption

Given the limited time in which we had two RTX 3090 FEs, we did not perform extensive current measurements. Total system consumption during the heaviest workloads was 880 watts, but the peaks were probably much higher. Initially we had set up a system with a 1200W power supply, but it failed during the benchmarking of Shadow of the Tomb Raider.

Benchmarks: 3DMark and games

We kick off with 3DMark, in various flavors. The regular Fire Strike test, which runs in 1080p resolution, barely shows scaling, and in the Extreme and Ultra version (1440p and 4K) that doesn’t get much better. That is striking, because with 2080 Ti’s in SLI we still saw excellent scaling.

In Timespy, the DirectX 12 version of the benchmark, it’s a completely different story. The SLI set-up scores almost 60% better than a single RTX 3090. In the ray tracing test Port Royal, the difference rises to an excellent 92%.

It seems clear that Nvidia not only actively supports the ‘old’ SLI, in which the driver-controlled how multiple GPUs worked together in DirectX 11 applications but has also disabled existing implementations thereof. If you want to benefit from SLI in games, you can only do so in games that support the new form of multi-GPU.

Games

As discussed on the previous page, we have only tested games that support combining multiple GPUs. Two of them are also part of our standard test suite, namely Shadow of the Tomb Raider and Red Dead Redemption 2 (in the Vulkan API). We also ran the Quake 2 remaster with RTX, Battlefield 1 (in DirectX 12 mode) and Deus Ex: Mankind Divided. Battlefield V no longer supports SLI.

Shadow of the Tomb Raider

In 4K resolution, you can already achieve a comfortable 117fps in Shadow of the Tomb Raider in medium settings. Nevertheless, with a second map you make a nice jump to 167fps, a scale of 43%. In ultra settings, the second card adds 73% to the performance, so you can fully saturate even the 4K monitors with the highest refresh rate.

Red Dead Redemption 2

In Red Dead Redemption 2 you have to use the Vulkan API to use the computing power of two video cards. In 4K resolution, we see around 60 percent scaling on both medium and ultra. However, with a single RTX 3090 you can even go a long way in ultra settings, so SLI doesn’t really make the difference between playable and unplayable here.

Where 4K ultra was still feasible with one GPU, for this game in 8K, the frame rates are not yet optimal even in SLI. You can even forget a tight 60fps on medium. It is noticeable that the scaling seems to get worse, which may indicate that there is a different bottleneck in the background than pure GPU computing power.

Quake 2 RTX

Since Quake 2 RTX requires a lot of the GPU, we also benchmarked this game on the lower wqhd resolution. In medium settings we reach 57% scaling, on ultra, it rises to 63% for a neat 134fps.

At 4K resolution, an RTX 3090 already suffers quite a bit with medium settings. SLI conjures up a comfortable 72fps on the monitor. Even with high settings, you will on average still exceed 60fps in SLI.

The requirements set by Quake 2 RTX on 8K resolution are too heavy even for the SLI setup. Although we see excellent scaling in percentage terms, the result is unplayable on both medium and high.

Battlefield 1

The World War I shooter Battlefield 1 was the last installment in this series with multi-GPU support; the newer Battlefield V lacks that functionality. Since this game is a bit older, 4K resolution is no longer a challenge for the RTX 3090. We even see a slight negative scaling.

At 8K resolution, the extra GPU power of the second RTX 3090 is more than welcome. In both medium and ultra settings we see gains of almost 70%: on medium we go from 63 to 106fps, on ultra from almost 50 to more than 80fps. However, the frame times are not getting any better in our software measurement, although unfortunately this does not say everything in the specific case of multiple video cards.

Deus Ex: Mankind Divided

The last game we benchmarked is Deus Ex: Mankind Divided. In 4K resolution, the frame rates with one card are actually fine. The profits of a quarter on medium settings and a third on ultra are nice, but no more than that.

However, at 8K resolution, one card is short, so the second card makes the difference. In medium settings, the scaling is even almost perfect, so you can shoot from a measly 36fps to 71fps. In a relative sense, the scaling on ultra-settings is clearly worse, but even 100% scaling would not have yielded 60fps in that scenario.

Benchmarks: video editing and rendering

In Adobe Premiere Pro, it is first of all striking how much faster GPU rendering is than CPU rendering. The same project when we render on the overclocked 10980XE CPU takes almost eight and a half minutes, but a single RTX 3090 got the job done in 29 seconds (!). Adding a second GPU yield a time saving of six seconds, in a relative sense a nice 20%. In DaVinci Resolve, the added value of SLI is minimal, barely six percent.

Blender

Using Blender, we render the ‘Barcelona Pavilion’ scene, which you can download for free here . In the cuda test, only the normal GPU cores are working. The scaling is then almost perfect: the SLI system is ready in just under eight minutes, while a single RTX 3090 takes more than 15 minutes.

If the ray tracing scores are also allowed to contribute, the render times will go by more than half. Again the SLI configuration manages to complete this task almost twice as fast.

OctaneBench

Finally, we run OctaneBench, the benchmark of the fully GPU-based OctaneRender software. This renderer also uses the ray tracing scores of Nvidia RTX GPUs. The total score of the benchmark comes to 1343 points, which makes the scaling almost perfect with 97%.

The benchmark consists of four scenes, each of which yields three sub-scores for info channel performance, direct exposure and path tracing. All subtests show near perfect scaling without exception, always high in the 90 percent.

Conclusion

In games that support DirectX 12 multi-GPU, adding a second RTX 3090 can make the difference between playable and unplayable. The scaling we see is generally excellent, with peaks of more than 90%. In games like Shadow of the Tomb Raider and Deus Ex: Mankind Divided, the SLI setup makes it possible to play comfortably in 8K resolution. With one RTX 3090 you can forget that in those games.

It could have been that beautiful for many gamers. However, for the five games we tested in this article, we already had to dig a lot in our game archive. In practice, almost every game that ends up in stores in 2020 does not support any form of multi-GPU. Unless you happen to be a big fan of one of the games on Nvidia’s short list , the second card hangs most of the time for Jan with the short last name in your system.

In addition, Nvidia has discontinued support for the traditional driver-based SLI. Even in older games for which such a profile was available, a second RTX 3090 is of no use to you. Now these are mainly games that you can now easily run on one card, but still.

What remains are the so-called ‘creative’ workloads. Rendering software, in particular, usually scales almost linearly, which means that adding an extra map almost halves the time required. In video editing software, the influence is more limited: you have a lot of the potential gains with GPU acceleration with one card long and wide, you have to have a very thick configuration and very heavy projects to make a second RTX 3090 worthwhile to be.

Fanatic renderers are bound to be very satisfied with that news, but the fact is that there are far fewer of them than gamers. And for that target group, the added value of SLI in 2020 is nil, we have to conclude with a heavy heart. Plus, hopes for a sudden turnaround seem vain, both due to lack of interest from game developers and Nvidia’s careful deathhouse construction.

Loading...