At a GeForce special event today, Nvidia CEO Jen-Hsun Huang announced the long-awaited news: Nvidia has announced the next step in its RTX revolution, with nearly twice the performance per watt of the previous generation of Turing, and AI with acceleration promised.

The new card features a second-generation RT core, a third-generation Tensor core, and an enhanced streaming multiprocessor (SM), representing a complete overhaul of the GeForce technology present in the 20 series. Not only do we see a significant evolution in core size, improved PCBs, redesigned (now 12-pin) power connectors, and "new low-profile leaf spring" mounts, but also (necessary) improvements in cooling solutions. Fan placement and the software stack that controls them have also been revamped.

After an ominous countdown, the specs are finally confirmed, and as has been believed from a barrage of leaks and rumors, Samsung's new 8nm-based Ampere architecture packs quite a spec.

With twice the number of floating point units in each SM, the new card touts twice the rumored CUDA core count and looks to be a great card for ray tracing as well as standard rasterization.

We had expected the RTX 3070 specs to be kept under wraps until October, but in a reversal, they will be announced along with other specs; Huang claims that the 5,888-core GPU will have a 1,730 MHz boost clock, but if it is a previous Nvidia chips, it would actually be much higher than that.

The card is faster than the RTX 2080 Ti with 6GB GDDR6X memory and up to 384Gb/s bandwidth on a 256-bit bus. And it costs $499, which is quite eye-popping considering the Ti is a $1,200 GPU.

The 3080 will certainly pack a punch as the "flagship" of the Ampere RTX generation, with 8704 CUDA cores and an expected boost clock of up to 1710 MHz. As for memory, it has 10GB of GDDR6X memory and a whopping 760Gb/s bandwidth on a 320-bit bus.

The top-end BFGPU 3090 has a luxurious 10,496 CUDA cores, which Nvidia guarantees will hit a boost clock speed of 1700 MHz. Add to this 24GB of GDDR6X memory on a 380-bit bus, pushing (a whopping) 936GB/s of bandwidth, and it is said to be capable of up to 8K gaming experiences, but the real question is the price. At $1,499, the price is not cheap.

This makes the frankly massive RTX 3090 the most expensive Nvidia GeForce GPU ever released. However, Jen-Hsun has smartly assembled this as a way to create a Titan graphics card that is as widely available as a standard GeForce GPU, whereas Titan quantities were limited and only available from Nvidia itself, The advantage is that any add-on board manufacturer can create their own.

If only there were a cooling solution to rival the spectacular triple slot installed on the RTX 3090...