Nvidia released a warm-up message, and the "computational graph of the Bernoulli digital engine" appeared, which was published by Ada Lovelace in 1843, thus indirectly confirming The RTX 40 will use the Ada architecture.
In addition, a post-it note appeared in Nvidia's warm-up information, and the numbers shown above may refer to the specifications of the AD102 GPU - 629 mm² chip size, 75.38 billion transistor count, 208 may represent its performance is 2.08 times that of the previous generation.
Nvidia has now announced that it will hold a special live broadcast of GeForce Beyond on September 20, which is expected to introduce information about the new generation of GPUs.
The following is the latest information on the RTX 40 graphics card:
- RTX 4090: Equipped with AD102-300 GPU, 16384 CUDA cores enabled, GPU frequency up to 2520 MHz, 24GB 21 Gbps GDDR6X memory, 1 TB/s bandwidth, 450W default TGP, and the maximum configurable TGP is 660W.
- RTX 4080 16GB: Equipped with AD103-300 GPU, enabling 9728 CUDA cores, GPU frequency up to 2505 MHz, equipped with 16GB 23 Gbps GDDR6X memory, TGP is 340W, can be configured to 516W.
- RTX 4080 12GB: Equipped with AD104-400 GPU, enabled 7680 CUDA cores, GPU frequency up to 2610 MHz, equipped with 12GB 21 Gbps GDDR6X memory, TGP is 285W, can be configured to 366W.
0 Comments