MD has finally revealed the specifications, prices and details of the performance of the graphic cards of Radeon RX 9070 and Radeon RX 9070, with full criticisms planned in the next few days. While we expect this, we have already discussed the possible effects on performance, the removal of FSR 4, and now we would like to offer an additional context to improve the quality of the coding of the new Radeon, an often ignored appearance of the new GPU (including that of us).
The AMD GPU code has been criticized for a long time for a poor quality of video coding when popular formats and binary speed for transmission games and Nvidia are used as a clear option for all those who wish to use video coding as a clear option.
With DNA 4, the coding of the AMD coding, the quality of the quality improves considerably and the examples it presented were certainly very aware. AMD emphasizes in particular 1080p H.264 and Hevc at 6 megabit per second, one of the most used configurations, which shows a significant increase in visual quality.
It remains to be seen if this is applied in a variety of scenarios, but historically, they discussed discussions on the quality of the coding of new formats such as Av1. With DNA 4, AMD focuses on concrete improvements on real applications, indicating that they are much safer in the quality of its encoder.
AMD estimates a 25% gain in the quality of the H.264 codes with low latency, an 11% improvement in the Hevc, a better AV1 coding with frame B support and an increase in the coding surrender from 30% to 720p. These figures are likely to refer to VMAF scores.
In addition to the coding, there are many other significant improvements. The nucleus monitoring shelves now offer two motors cut instead of duplication for the boxing of rays and the triangles of the rays. A new blocking of the package has been introduced, which unloaded some aspects of the surveillance of online radiation to the RT nucleus.
The BVH (hierarchy of the limit volume) is now double and the implementation of the shelves has brought many other improvements, a reason why the results of the DNA 4 radiation exceed their rasorization improvements.
Calculate, memory and indicate improvements
The computer also contains different optimizations, as well as a 5.0 x16 PCIe interface and a 256 -bit memory bus with GDDDR6. AMD declares that the improvement of the compression of memory and GPUs is equipped with 16 GB of VRM, which should be sufficient for most modern games.
However, the screen engine is a mixture. Although it is compatible with Displayport 2.1, its functions remain unchanged by RDNA 3, with a maximum UHBR 13.5 band instead of the complete UHBR, which is now used for exhibitions of about 4K 240Hz and is compatible with the Blackwell architecture of Nvidia.
HDMI 2.1B is also included. It should be noted that AMD can now be downloaded from the GPU, a lower inactive consumption for more monitor configurations and video controls of video controls.
The 48 navigation system is 357 mm² in the 4 Nm TSMC silicone with 53.9 billion transistors. This is 5% lower than Nvidia’s Blackwell GB203, which is used in RTX 5080 and 5070 ti, but contains 18% more transistors, which means that design is well wrapped.
However, Nvidia always has an advantage over the area and efficiency of the transistor, since RTX 5080 should be about 15% and perhaps faster in surveillance based on AMDS RX 9070 -Text -Text -Lox, while less transistors and smaller deterioration are used. However, the TDP RTX 5080 is 360 W compared to 304 W for 9070 XT, a power extract of 18% higher, although real energy consumption can vary in games. The 9070 XT must be closer to RTX 5070 TI with its 300 -W -tdp.
