GeForce Beyond, where and when to see the presentation event of the new NVIDIA GPUs

Today September 20 the day of GeForce RTX 4000the new video cards from NVIDIA. After a lot of chatter and more or less apt anticipations, we will finally know the whole truth about launch lineup of cards based on Ada Lovelace architecture.

The Videocardz site has published the image of the RTX 4090 FE shown by Jensen during the event

At the event GeForce Beyondas part of the GTC 2022, the CEO Jensen Huang will show the Founders Edition cards for the first time, illustrating the main technical improvements and consequent performance improvements. We will also know the date of arrival on the market and, probably, also the prices.

The event will start at 17:00 Italian time and you can follow it are YouTubeabout are Twitch. After the event, NVIDIA recommends visiting GeForce.com for more information.

What to expect? The rumors before the launch

NVIDIA should present three models GeForce RTX 4090, RTX 4080 e RTX 4070. The top of the range should be the first to arrive on the market, during October, while the other proposals will land later but, according to rumors, always within the year.

Ada Lovelace architecture should represent an important leap forward compared to Ampere, for example we know that there will be a large increase in the L2 cache in order to keep more data close to the compute units. Furthermore, decisive steps forward are desirable in terms of ray tracing management thanks to renewed RT core and Tensor core.

According to the latest rumors, the GeForce RTX 4090 it should have an AD102 GPU with 16384 active CUDA cores. The memory, 24 GB of type GDDR6X, would be set at 21 Gbps for a bandwidth equal to about 1 TB / s by means of the 384-bit bus. As for the TGP, the default one would be 450W, like that of the RTX 3090 Ti.

The GeForce RTX 4080instead, it could come in two versions. The most powerful model would have an AD103 GPU with 9728 CUDA cores and 16 GB of GDDR6X memory at 23 Gbps on a 256-bit bus, for a bandwidth of 736 GB / s. The TGP, on the other hand, would be set at 340W.

The other GeForce RTX 4080 should instead have 12 GB of GDDR6X memory at 21 Gbps, with a 192-bit bus that would bring the bandwidth to 504 GB / s. In the past, this amount of memory had been charged to the RTX 4070 Ti or the RTX 4070, but apparently NVIDIA would have decided to offer a pair of RTX 4080s.

The two RTX 4080 models are expected to hit the market in the first two weeks of November. The GPU on board the second RTX 4080 should take the name of AD104 and offer 7680 CUDA cores. As for the TGP, the default value would be 285W.

If so, the alleged GeForce RTX 4070 it could still be based on an AD104 GPU but with only 7168 CUDA cores, flanked by 10 GB of GDDR6X memory at 21 Gbps on a 160-bit bus. The card could be accompanied by a 250W TGP and arrive between late November and early December. All these specifications, of course, are not to be considered official and are to be taken with a grain of salt, the truth we will find out from 17:00 onwards.