From PC World: On January 21, a photo of an Nvidia GPU with code name GB202-200-A1 (see header image above) surfaced on the infamous Chip Hell forum. The user who posted it claims that it’s a prototype of the most powerful GPU in the upcoming RTX 50 series.
The fully configured GPU allegedly utilizes the maximum possible number of computing units for which the chip was designed. We’re talking about an RTX 50-series chip that supports 192 streaming processors and an incredible 24,576 shader units. The current top model RTX 5090 has “only” 21,760 shader units. Mathematically, the GB202-200-A1 has almost 13 percent more raw computing power.
The second striking feature are the two 12V 2×6 power connections. These also appear to be necessary, as the Thermal Design Power (TDP) is said to be an extremely high 800 watts. By comparison, the RTX 5090 has a TDP of 575 watts. The power consumption of the GB202-200-A1 is therefore greater by 39 percent.
Nvidia appears to be investing that additional power requirement in higher clock frequencies. Here, too, the Chip Hell forum thread provides very specific figures: the base and boost clocks of the GB202-200-A1 are said to be 2,100MHz and 2,514MHz, respectively. For comparison, the RTX 5090 runs at 2,010MHz and 2,410MHz.
View: Full Article
