DLSS is quietly saving Nvidia from a problem no GPU manufacturer can brute-force anymore
For 30 years, we've improved performance with more transistors, higher clock speeds, and more power, but in recent years, it feels like that era has finally come to an end. The RTX 5090's Blackwell chip is nearly 750 mm², which is extremely close to the 858 mm² maximum size a silicon wafer can physically handle. This means Nvidia can't just make chips bigger or hotter anymore. Instead, DLSS is a cheat code that uses AI to simulate the performance of a 1,000 W GPU using only 450 W of actual electricity.
For 30 years, we’ve improved performance with more transistors, higher clock speeds, and more power, but in recent years, it feels like that era has finally come to an end. The RTX 5090’s Blackwell chip is nearly 750 mm², which is extremely close to the 858 mm² maximum size a silicon wafer can physically handle. This means Nvidia can’t just make chips bigger or hotter anymore. Instead, DLSS is a cheat code that uses AI to simulate the performance of a 1,000 W GPU using only 450 W of actual electricity.
Felix Collins
Ireland
Ireland
Published by: aplhsindia.in
