A man who is crazy about hardware and loves users, a man with a passion to make everything he thinks a reality,
This is QM Jilim, the chief operator of Quasar Zone.
This news is a simple translation of foreign media and does not include the subjective opinion of Quasar Zone.
Months have passed since NVIDIA launched the RTX 30 series graphics cards based on the Ampere architecture. Four products were released: RTX 3060Ti, RTX 3070, RTX 3080 and RTX 3090. More products will be added to the lineup, but the biggest concern for players with the new RTX30 series is when they will be available.
Players weren't able to purchase Ampere-based graphics cards, but information about the Ada Lovelace architecture used by NVIDIA's next-generation RTX 40 graphics card was gradually leaking.
The naming method for next-generation architectures retains the scientist's name NVIDIA has always used. Ada Lovelace is the only daughter of the famous English poet Byron Lord Byron. She published the algorithms used in the first analyzers claiming that computers aren't just counting numbers, and she is also recognized as the first person to recognize the full potential of computers.
The information currently exposed is mainly based on the AD102 core of the Ada Lovelace architecture. Following the old naming convention (GA102 is used by the RTX 3080 and RTX 3090), this core will also be used in the flagship class of the RTX40 series in the future, which is also the potentially highest performing architecture in the consumer market.
All products of the new architecture will upgrade the process technology to 5nm and will be equipped with a larger cache capacity than the previous second generation. The AD102 chip is said to have 12 image processor modules (12 * 6 architecture), 72 sets of texture processing units, 144 sets of stream processor units, up to 18432 FP32 units (CUDA cores) and up to 66 TFlops of computing power.
As a core specification of graphics card operation performance, the CUDA core provided by AD102 is more than twice as many as 8704 CUDAs currently adopted in the RTX 3080, and twice as much as the 10496 CUDA cores installed in the current RTX 3090. Closer, but looking at these specs, the theoretical performance of the Ada Lovelace architecture will be about 71% more powerful than Ampere.
The main ray tracing performance of the RTX series graphics card is the number of core indicators, the number of RT cores, has not yet leaked a lot of effective information, but the current upgrade speed will certainly not be weak.
If NVIDIA can keep up with these performance gains, it will continue to dominate the competition for graphics cards in the next-generation consumer market. For the average player right now, when to buy the RTX 30 series is key. From sales so far, the Ada Lovelace architecture will be announced in 2022.
Original source mydrivers
0 Comments