Turing GTX 1160 / RTX 2060 gaming performance estimation vs GTX 1060

The next generation of Nvidia and AMD GPUs is upon us. The new NV Turing architecture is here in the form of the expensive desktop graphics card RTX 2080 and 2080 Ti and niitial tests show significant, but puzzling, performance improvements compared to top GPUs of previous generation –  The GTX 1080 and 1080 Ti, based on Pascal architecture. The puzzlement comes when considering the costs and power consumption, the performance / cost ratio has not improved significantly if at all in some benchmarks.

In this short post, I want to try and draw a rough estimation of the gaming performance of the future successor to the popular GTX 1060 (laptop version), the GTX 1160 / RTX 2060 (not really sure). We’ll do it by calculating the performance / power consumption ratio, according to tests and benchmarks of desktop versions from big sites.  The estimation based on the assumption that the RTX 2060 will employ the basic characteristics of the bigger parts RTX 2080/2080 Ti, meaning – considerably amount of Tensor functionality.

It is important to say that this assumption may show to be untrue. It just might be that lower performance “Turing” based parts will have lower ratio of Tensor cores, making them mostly an improved “Pascal” GPUs for gaming. But, it’s probably not the case as Nvidia tries to push their new tech everywhere, before the competition and it can be seen in the form of future games support.

OK, short and quick specs overview for the GTX 1080, 1080 Ti, RTX 2080:

Gaming performance per watt - Pascal vs Turing DLSS, FFXV@4K GTX 1080 RTX 2080
Gaming performance per watt : Pascal vs Turing (DLSS) at FFXV@4K at max graphics. GTX 1080 vs 1080 Ti vs RTX 2080

GTX 1080 GTX 1080 Ti RTX 2080 FE RTX 2080 FE (DLSS)
Cores count256035842944 2944
Clocks/Boost [GHZ]1.67/1.7331.48/1.5821.515/1.8 1.515/1.8
Tflops (boost)8.911.310.6 10.6
ArchitecturePascalPascalTuring Turing
TDP180250225 225
Relative performance (avg), FFXVx1X1.3X1.31 X1.78
Relative performance (99th), FFXVx1X1.35X1.42
Relative performance (avg), AoSx1X1.3X1.3
Perf/W (avg)x1X0.94X1.05x 1.424

These specs and stats are based on other sites benchmarks (1, 2, 3, 4). I used the official TDP numbers because they are pretty accurate in this case. This is a very rough test and I used Ashes of Singularity TomsHardware numbers and Anandtech and Guru3D FFXV numbers. Unfortunately, all sites had some kind of test missing – either the 99th percentile, or the FFXV DLSS benchmark.

We see two interesting points: (a) With DLSS activated, FFXV sees big jump in performance – around +36% in average FPS. (b) If we consider the 99th percentile, the RTX 2080 take a small lid over the GTX 1080 Ti where before it didn’t have it.

Now, with DLSS activated, the RTX 2080 was ~x1.42 as power efficient as the GTX 1080. This measurement type is important to estimate what can be achieved in a more TDP limited system. If we trust these numbers (and we shouldn’t really), then with the future RTX 2060 one could get 40-50% higher gaming performance compared to the GTX 1060 6GB.

Mind you, these numbers are lower than what we see in Nvidia marketing slides (below). Nvidia tells us that FFXV runs around 45% faster without DLSS and more than 100% faster with DLSS on the same hardware (GTX 1080 vs RTX 2080). It’s not clear what they are measuring (minimal FPS? future projection of optimized games?). That way or another, the promise of much higher gaming performance remain.


If prices remain the same (which makes sense), the performance level of today’s $1000 gaming laptop could jump to the performance levels of a full GTX 1070 (laptop version) as the difference between GTX 1060 and 1070 is currently indeed around 45-55% more or less. This is close to the jump in performance between the Maxwell II GTX 970M (a $1000-$1200 king) to the Pascal GTX 1060 (same). It could be even higher if Nvidia slides promises will be realized, but that’s all marketing and speculations for now.

Important to remember that these numbers are the best case scenario based on current FFXV benchmarks versions of other sites, given that a game does use the DLSS feature well enough. It could be higher and it could be lower, depending on the game, CPU, graphics settings, different implementations of laptop version GPUs and more. But, then again, it seems that the list of supporting games is getting longer by the month.

One question remains – where is AMD in that story? they are surely not going to be left behind just because of this DLSS.

Finally, check this list to keep in touch with what kind of gaming laptop you can get for up to $1000.

0 0 vote
Article Rating
Notify of

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Oldest Most Voted
Inline Feedbacks
View all comments
2 years ago

I hope your’re right with your estimates, DLSS indeed looks promising.

2 years ago
Reply to  junky

> after 2 years with the same GPUs, it’s quite plausible to see 50% performance jump for the same price, no?

Unfortunately no, the way NVIDIA sees it is 50% more performance for 50% (or ever more) price increase. In my country RTX 2080 costs more than GTX 1080 Ti.

> drivers and games aren’t optimized yet

As I said before in other comment on your blog, Turing is basically the same architecture as Pascal, so drivers optimizations won’t increase performance a lot. Moreover, game developers are lazy, they rarely optimize their games.