The dates for the lifting of the embargo on publishing reviews of the newest graphics cards in the RTX 50 series have been formally announced by NVIDIA. The results of the initial testing of the flagship GeForce RTX 5090 model will be released to the public on January 24. Reviews of the RTX 5080 will be accessible on January 30.

The decision is unusual. Initially it was expected that RTX 5080 would be tested first. But NVIDIA changed the plan to give more time to finalize the BIOS. This way the company's partners will be able to prepare full support for the new product by the time of its release.

An interesting feature will be the simultaneous release of reviews of NVIDIA reference models and third-party solutions. This will allow potential buyers to get a full idea of the GeForce RTX 50 lineup on the day of its premiere.

The company demonstrated that the RTX 5090 performed better than the RTX 4090 when the new NVIDIA graphics cards were released. Experts noted that the tests employed a variety of data compression techniques, which may have skewed the results. FP8 (8-bit) compression was used to test the RTX 4090, whereas FP4 (4-bit) compression was used to test the RTX 5090. 

Because FP4 uses half as much video memory, the difference in compression has a big impact on performance, which explains why the RTX 5090 performs better. In spite of this, NVIDIA displayed comparison charts that showed the RTX 5090 outperforming the RTX 4090 by almost twice as much.

Users of Reddit forums are worried about the test results. Many people think that this kind of comparison is unfair since it can mislead customers and ignores important variations in compression techniques. Furthermore, the tests did not sufficiently address the fact that FP4 lowers compute quality in comparison to FP8.

Main image: reddit.com

Comment on this