We start the week talking about the specifications of NVIDIA GeForce graphics cards RTX-4090GeForce RTX4080 and GeForce RTX4070although the only ones that expand the information refer to the top model, whose frequencies it will arrive at are indicated.
According to a very reliable source like @kopite7kimithe NVIDIA GeForce RTX 4090 would come to market using an AD102-300-A1 graphics chip consisting of 16348 CUDA coreswhich would be able to reach a Base/Turbo frequency of 2235/2520 MHz. This graphics chip is accompanied by 24 GB of memory GDDR6X @ 21 Gbps with 384-bit memory interface, while TDP will be 450W.
The NVIDIA GeForce RTX 4080 would arrive with the AD103-300-A1 graphics chip with 10240 CUDA cores (+18% vs. RTX 3080) with 16 GB of GDDR6 memory @ 21 Gbps (vs. 18 Gbps of the last rumours), with a TDP to be fixed at 420W. Finally, the NVIDIA GeForce RTX 4070 would happen with 7168 CUDA cores (+21% vs RTX 3070), but that would come with significant cuts, like being accompanied by 10 GB of memory GDDR6 instead of 12 GB, with a memory interface of 160 bit instead of 192 bits, and its TDP would be 300W.
As always, the most important detail, the startthere is no news, recalling that initially these GPUs were delayed for a month to allow more space to clean up the current stock of GPUs, to the point that given the low demand that the sector will experience in during the second half of the year, that these graphs can be published by 2023 instead of the end of 2022.
For now, we expect that GeForce RTX 4090 move from arrival in August to September, while the GeForce RTX 4080 sees its launch shifted from September to October and the GeForce RTX 4070 from October to November. That is to say that all would arrive a month later than expecteduntil they finally get their release pushed back to 2023.