Nvidia wants to increase its market share with RTX 5090 after RTX 4090, which pushed the limits. Expected to debut in 2025, important details about the next-generation graphics cards codenamed “Blackwell” have leaked. According to this, there will be a huge increase in the core count of RTX 5090.

Nvidia RTX 5090 CUDA core, memory interface and clock speed leaked

X (Twitter) has revealed just how powerful the RTX 5090 will be. The graphics card, codenamed GB202, is rumored to feature more than 24,000 CUDA cores. This would represent a significant upgrade from the 18,432 cores found on the RTX 4090’s AD102 chip. Note that not all of these CUDA cores were used in the RTX 4090.

Other leaked specifications for RTX 5090 include a 512-bit wide memory interface and GDDR7 video memory clocked at close to 3 GHz. The combination of the massive core count and very high bandwidth memory will provide a huge leap in graphics power compared to the RTX 4000 series.

Clock speeds are expected to increase by 15 percent, bringing the maximum clock speed to 2.9 GHz. With Nvidia’s GPU Boost technology, that number will reach 3 GHz in games. This would make the RTX 5090 the first stock graphics card to cross the Rubicon without overclocking.

Early reviews indicate that the RTX 5090 will be 70 percent more performant than the RTX 4090. If the 24,000 core count is accurate, the RTX 5090 looks set to claim the GPU performance title.

Nvidia is also rumored to triple the L2 cache addition in the Blackwell series. This will help increase memory bandwidth, representing a significant boost in games.

What do you think about this? Please don’t forget to share your thoughts with us in the comments.