Yesterday, Nvidia presented as expected the new generation of Geforce, RTX 30series based on the Ampere architecture. In addition to the fact that the new architecture means more efficient rendering, Nvidia has crammed in more cuda cores than ever before.
The “entry-level model” RTX 3070 also has significantly more cuda cores than the previous generation top model RTX 2080 Ti. This indicates that the entire 30-series will offer really good performance in both games and professional programs.
As you can see in the table below, all three new cards have greatly upgraded specifications. They have, for example, hdmi 2.1 which means support for significantly higher resolutions, hdcp 2.3 for the latest copy protection and several of the cards have more memory than before.
A big difference is also how much power the cards draw. That Nvidia has invested a lot in cooling and raising the maximum power was clear even before the launch, and if we look at the sheer numbers, we see that the company is not kidding. RTX 3070 draws 220 watts, about as much as RTX 2080. RTX 3080, which replaces that card, draws 320 watts – 95 watts (42%) more than its predecessor. RTX 3090 draws even a little more with 350 watts.
To get rid of all the heat all this electricity generates, the cards have also become larger. All three are longer but 3090 stands out. It goes from 2080 Ti’s 26.7 to a full 31.2 centimeters. In other words, the card requires a really large chassis.
More electricity also means a higher recommended minimum power on the power supply. The new cards require at least 650 (3070) and 750 (3080, 3090) watts, respectively.
Nvidia has also published a graph comparing the performance of the RTX 2070S (the already slightly hot-headed Super version) and the 2080S against the 3070 and 3080 in five games.
Judging by the graph, the difference between the different cards in the 30-series is greater than before, and both of these cards are real upgrades. 3070 is faster than 2080S in all games, with a pretty good margin.