RTX 5090 chip production cost revealed. The new graphics card is a gold mine for NVIDIA

News

Translated with AI from the Russian original.

The flagship Nvidia GeForce RTX 5090 graphics card is one of the most expensive on the market, which is due to the large GB202 GPU used in it. It is estimated that producing a single such graphics processor costs Nvidia approximately $290–$340, but the expenses do not end there.

The area of GB202 is 761.56 mm², making it one of the largest GPUs produced for consumer PCs. It is manufactured on a 4nm process (4NP) at TSMC fabs, where the cost of a 300mm wafer reaches up to $16,000. Given the chip dimensions, about 72 such dies can fit on a single wafer, but not all are flawless. According to calculations, about 47 dies per wafer are fully functional, and some can be used with disabled cores in lower-tier models, such as the GeForce RTX 5080.

Even if the die cost is $290–$340, those are not all the expenses. The figure does not include testing, packaging, memory, the cooling system, and other components. Clearly, Nvidia remains comfortably in the black thanks to the RTX 5090 retail price of $1999, while more expensive models based on GB202 (such as professional and server solutions) bring the company even greater profit.

NVIDIA effectively utilizes partially defective dies, selling them as cut-down versions. This allows minimizing losses and extracting maximum profit from each manufactured chip.

***

Cover author: NVIDIA. Cover source:

Never miss our newsAdd this site to your preferred sources to see us more oftenAdd on Google

Now on the front page