First I have both without, and with DLSS 2.0 benchmarked. Then I used the NVIDIA frame limiter to slow down the DLSS throughput so that in the end the average values were virtually the same at around 76 FPS each. The fact that the min-framerate turns out a bit worse when passing with the limiter is also a finding that was made over all tested games. However, I will limit myself to only one game here, because the results of all games with DLSS 2.0 are very similar. There’s no need to repeat yourself.
First we look at the pure FPS curve, where the almost straight curve with the limiter is directly visible. It still looks good, but the devil is in the details.
In this case, however, the curves can be replaced by bars and you can see much more clearly where the differences lie. There are more frames that are below 11.11 ms, but the limiter then slows down so that frames are also reduced to approx. 17 to 22 ms can be artificially slowed down. This is also the reason for the difference in the first bar chart.
How hard the limiter intervenes can be seen even more clearly from the variances, because it doesn’t really get any clearer than that. However, in honor of the limiter method, one must also say that, purely subjectively, one only notices it from time to time, but the scrolling is still quite harmonious overall.
Now let’s get to the new metrics and the beautiful new curves. While the power consumption of the GeForce RTX 2080 Ti with and without DLSS remains almost the same over the duration of the benchmark, the GPU’s curve with activated limiter fluctuates extremely, but is visibly far below the other two. This already indicates an extremely reduced load and thus also power consumption.
And that’s right: Compared to the averagely equally fast run without DLSS with DLSS and limiter in the sum of GPU and CPU, you save a whopping 80 watts! That’s over 24% less! With the GPU alone, that’s even 81 watts and thus an incredible 30%. However, we also see that the CPU load remains about the same. There are about One more watt. With DLSS and without limiter, the power consumption of the GPU is about the same. 1 watt difference and slightly less with DLSS enabled. However, the CPU increases by a good 5 watts, which is not surprising, since more drawcalls have to be processed. The performance increases by a whopping 25% at approx. 1.6% higher total power consumption, but this is at the expense of the CPU.
Summary and conclusion
Finally, you can only add the efficiency calculation in watts per rendered frame to this. It wins DLSS 2.0 with Limiter before DLSS 2.0 and the card without DLSS. The whole thing happens very clearly, so that one can draw quite remarkable conclusions. For one thing, the frame limiters of NVIDIA and DLSS 2.0 do not yet understand each other 100%, which is annoying but hardly noticeable in normal use. Nevertheless, NVIDIA should check here what the reason for this is and whether the individual frames are not regulated too granular.
On the other hand, both the performance gain with DLSS 2.0 without limiter and the efficiency gain with DLSS 2.0 with limiter activated are a real house number. Once again for the record: a full 80 watts with actually identical picture quality is a real hammer that I didn’t expect. In older games like Wolfenstein Youngblood, the difference with limiter is not so extreme, but DLSS without limiter can really show its strength. Also a realization.
As a conclusion, I’m taking here with me that DLSS and the use of local AI on the graphics card is actually pointing in the right direction. If you don’t need extreme FPS because you can’t afford an eagle-eye gaming monitor with hundreds of Hz anyway, you can at least save quite a bit of energy with DLSS and limiter. Which means that the whole thing even in the beginning proves to be ecologically valuable. Jensen shows his green thumb here, although certainly unintentionally. But it’s always good to know where you can save, if you’re looking for it.