Tuesday, September 20, 2022

 

NVIDIA


NVIDIA apresentou a GeForce RTX 4090 e duas GeForce RTX 4080

Company officially introduced a new generation of GeForce RTX 4000 graphics accelerators

 As part of today's presentation, your boss Jensen Huang announced the first models of the new series - GeForce RTX 4090, GeForce RTX 4080 with 16GB of memory and GeForce RTX 4080 with 12GB of memory. In the reference version, the new items largely repeat the design of previous models.

GeForce RTX 4000 graphics cards use GPUs based on the new Ada Lovelace architecture. The chips have up to 76 billion transistors and are manufactured using TSMC's 4nm N4 process. Key features of the new architecture are 3rd generation RT cores and 4th generation tensor cores.

The company said that the new Ada Lovelace architecture graphics cards support DLSS 3.0 image scaling technology, which significantly improves game performance.

At the heart of the flagship GeForce RTX 4090 graphics card is the AD102-300 GPU with 16,384 CUDA cores. The declared frequency of its operation is up to 2520 MHz. The theoretical performance of the graphics card in single-precision operations is 82.6 Tflops, which is 2.3 times that of its predecessor based on Ampere architecture (GeForce RTX 3090). The new graphics card received 24 GB of GDDR6X memory with support for 384-bit bus, effective frequency of 21 GHz and peak bandwidth of 1 TB/s. The declared power consumption of the new product is 450W, which is 100W higher than the GeForce RTX 3090. According to Jensen Huang, the GeForce RTX 4090 is up to 4 times more powerful than the GeForce RTX 3090 Ti.

The GeForce RTX 4090 will go on sale on October 12th. Its recommended price is $1599.

Along with the new main accelerator, Huang introduced two cards with the same name but different technical characteristics. The first is the GeForce RTX 4080 16GB. It is based on a halved AD103-300 processor with 9728 CUDA cores, clocked up to 2505 MHz. The card received 16 GB of GDDR6X memory with a 256-bit bus width, a speed of 22.5 Gb/s per contact (effective frequency of 22.5 GHz) and a bandwidth of 720 GB/s. The declared power consumption of the novelty is 320 watts. This model will go on sale in November. Its recommended price is $1199.

What would become the GeForce RTX 4070 model ended up being called the GeForce RTX 4080 12GB. This card uses the AD104-400 GPU with 7680 CUDA cores running at up to 2610MHz. The accelerator received 12 GB of GDDR6X memory with a 192-bit bus width, a speed of 21 Gb/s per contact and a transfer rate of 504 GB/s. The declared power consumption level of new items is 285 watts. NVIDIA has priced this model at $899. It will also go on sale in November. The manufacturer claims that the performance level of GeForce RTX 4080 is 2-4 times that of GeForce RTX 3080 Ti.

Source: Nvidia

No comments:

Post a Comment

  TECH Galaxy A16 to come with 6 years of updates for €209 The Samsung Galaxy A16 is said to be launching soon in two variants, with prices ...