Huang Renxun has predicted that NVIDIA will usher in a large-scale wave of product updates in the second half of the year, including but not limited to GPU products. Obviously, this means that RTX 40 Series graphics cards are on the line. A new disclosure combs the parameters of the first four RTX 40 Series graphics cards, including RTX 4090, RTX 4080, RTX 4070 and RTX 4060.
Access:
In terms of CUDA scale , RTX 4090 is 53% more than RTX 3090, RTX 4080 is 23% more than 3080, RTX 4070 is 25% more than 3070, and RTX 4060 is 26% more than 3060**
Based on this conservative estimate, the growth rate of CUDA should be the same as that of the overall performance of the graphics card. Of course, it can be expected that the RTX 40 series will double or even increase by 2 ~ 3 times in game scenes such as light pursuit and DLSS.
There are two slightly strange things about this disclosure. One is that the RTX 40 Series graphics card has changed to TSMC 6nm instead of 5nm. Second, the video memory specification of RTX 4060 is 4GB less than that of 3060. It seems that after replacing g6x, the cost and heat dissipation cannot be suppressed, and TGP has increased by 30W.
On the same day, kopite7kimi, another good disclosure player, also gave the news that the AD104 core has been tested and is expected to be used on the RTX 4070.
It is widely speculated that NVIDIA will release RTX 40 Series graphics cards as soon as July and go on sale in September.
