GeForce RTX 4000: the latest rumors on the specs of the high end
The new ones are expected in the last part of the year, probably starting from September / October GeForce RTX 4000 are for the umpteenth time the subject of indiscretions. NVIDIA should introduce the models in the current year GeForce RTX 4090, 4080 e 4070 while at CES 2023 in January we should see the announcement of the RTX 4060, in view of a debut in the following weeks.
The numerous leaks in recent months, also thanks to the hacker attack by Lapsus $, have roughly revealed the technical specifications of the GPUs that will make up the NVIDIA offer. As always, however, it is difficult for a GPU to be activated in all its parts, especially at the beginning of the generation, so leakers have long been looking for the latest information that is missing on theactual configuration of the boards.
Continue to update.
RTX 4090, AD102-300-A1, 16384FP32, 384bit 21Gbps 24G GDDR6X, 450W, base 2235 boost 2520 actual max >2750;
RTX 4080, AD103-300-A1, 10240FP32, 256bit 21Gbps 16G GDDR6X, 420W,
RTX 4070, AD104-275-Kx(x is a number)-A1, 7168FP32, 160bit 18Gbps 10G GDDR6, 300W.
— kopite7kimi (@kopite7kimi) July 4, 2022
According to the latest rumors released by Kopite7kimi, the GeForce RTX 4090 should show up with an AD102 GPU equipped with 16384 CUDA coreflanked by 24 GB of GDDR6X memory a 21 Gbps su bus a 384 bit. There is talk of a TDP of 450Wsimilar to the current GeForce RTX 3090 Ti.
According to the leaker, the card should show up with a base frequency of 2235 MHz and a boost clock of 2520 MHzvalues that during gameplay – based on the operating parameters – will turn into an effective clock greater than 2750 MHz. An important increase is therefore expected for the frequencies (+ 50/60%), considering that the RTX 3090 has base and boost frequencies well below 2 GHz, a value reached only during gameplay in some cases. The RTX 4090 may be able to offer FP32 computational power of 90 TFLOPs, roughly double the 3090.
The GeForce RTX 4080 it should instead adopt a different GPU, AD103, with 10240 CUDA active cores and 16 GB of GDDR6X memory at 21 Gbps on a 256-bit bus. For this video card we speak of a TDP of 420W, that is 100W more than the 10 GB RTX 3080 but with much higher specifications, albeit with a lower bus. The frequencies are not known, but they should be higher than those indicated by the 4090. Finally, the RTX 4070 should adopt an AD104 GPU with 7168 CUDA core active, accompanied by 10 GB of 18 Gbps GDDR6 memory on a 160-bit bus. There is also talk of a TDP of 300W.