In the second half of this year, Nvidia and AMD will release a new generation of graphics cards, although the graphics card market in the past two years can be said to be very unpeaceful, so that many consumers have lost the desire to buy graphics cards, but the arrival of a new generation of graphics cards may allow some old players to rekindle the desire to buy new graphics cards.
Of course, there is still a lot of time before the official release of the new graphics card, and during this time, it is the time for people to show it.
According to the recent news released by the extranet whistleblower @Greymon55, the Ada Lovelace architecture AD102 GPU of the NVIDIA RTX 4090 graphics card is expected to provide more than 100 TFLOPS of single-precision floating-point arithmetic performance, which is 2.5 times that of the RTX 3090 Ti.
Of course, AMD is not to be outdone, and the single-precision floating-point performance of the RX 7900 XT will also come to the level of 90+ TFLOPS.
Another blogger, @Kepler_L2, broke the news that according to the information mined from the AMD GPU driver information, the RDNA3 architecture will still retain the traditional CU computing unit structure system without canceling it.
In addition to the flagship graphics card, the sub-flagship RTX 4080 and 4070, which are slightly lower positioned, also have some news.
Whistleblower @kopite7kimi said in a recent article that the RTX 4080 is based on the AD103 core, with 16GB GDDR6X video memory, video memory speed of 21Gbps, full card power consumption of 500W, FE public version with 16pin external power supply.
The RTX 4070 is based on the AD104 core, equipped with 12GB GDDR6 video memory, the entire card power consumption is 300W, and the FE public version is equipped with 16pin external power supply.
Only look at the video memory capacity of this aspect of the RTX 40 series is more generous than the RTX 30 series, the previous generation of RTX 3080 only 10GB video memory, RTX 3070 is only 8GB video memory, the truth and performance some do not match.
Of course, the power consumption of this generation is of course higher than that of the RTX 30 series, and the RTX 4070 is fine, only 80W power consumption is higher than the previous generation's 220W, but the 500W of the RTX 4080 is much higher than the 320W of the 3080.
In addition@kopite7kimi also broke the news that NVIDIA is testing a full-spec AD102 beta version, which can be described as luxurious, with 48GB 24Gbps GDDR6X video memory, 2 * 16pin interface, and its power consumption can reach up to 900W. Whistleblowers said they did not know whether the beta version would have a final finished product, while the outside world speculated that if there was a finished product, it might be prepared for the future RTX 4090 Ti.