When the new NVIDIA graphics cards were introduced, the company showed that the RTX 5090 outperformed the RTX 4090 in terms of performance. Experts observed that various data compression techniques were used in the tests, which might have distorted the outcomes.

The RTX 4090 was tested using FP8 (8-bit) compression, while the RTX 5090 used FP4 (4-bit) compression. The difference in compression has significant performance implications, as FP4 requires half as much video memory, which explains the RTX 5090's better results. Despite this, NVIDIA presented comparison charts where the RTX 5090's performance gains over the RTX 4090 were nearly twice as high.

Reddit forum users are concerned about the results of the tests. Many believe that such a comparison is unfair, as it does not take into account significant differences in compression methods and it can mislead consumers. In addition, FP4 reduces compute quality compared to FP8, which was also not adequately addressed in the tests.

Despite the criticism, NVIDIA continues to promote new graphics cards, and their marketing materials are increasingly raising questions about the veracity of the data presented. The company did not give a clear explanation regarding the use of FP4 for testing the RTX 5090.

It is possible that in the future NVIDIA will provide more precise information about the testing methods of graphics cards in order to avoid dissatisfaction from users and experts.

Main image: reddit.com

Comment on this