We always expect a significant performance jump between different generations of hardware to justify the switch. How about about 70% more in core numbers? That's what one leaker suggests by comparing the current top-of-the-line NVIDIA GeForce RTX 3090 Ti with the future (and likely) RTX 4090, the strongest GPU in the manufacturer's next-generation Ada Lovelace.

The information, most likely, comes from the hacker invasion that stole source codes for NVIDIA's technologies, such as the GPUs' DLSS and LHR limiter, among other information. There was more than 1TB of data stolen by the hacker group. So it would not be surprising to see more and more information about new NVIDIA technologies popping up all over the internet.

The Twitter user named @davideneco25320, published what would be the SM count of the entire lineup of the next-generation NVIDIA RTX 40 Ada Lovelace. He compares it to the last two generations, Ampere and Turing, and the numbers are surprising, especially among the stronger chips. While the GA102 chip, top of the line in the Ampere generation, has 84 SMs (10,752 cores in the RTX 3090 Ti and 10,496 in the RTX 3090), the supposed AD102 chip should have 144 SMs, which gives 18,432 CUDA colors.

The AD102 chip should equip everything from the RTX 4080 to a likely RTX 4090 (Ti), as with the GA102 chip in the current Ampere generation. So this increase by 71% in numbers of CUDA colors, could be between the RTX 3090 Ti and an RTX 4090. The chip below the top of the line, AD103, on the other hand, with 84 SMs, would have the same core count as an RTX 3090 Ti (10,752) and we could be talking about a possible RTX 4070.

The performance difference narrows among the weaker chips. Among Ada Lovelace GPUs based on AD107, AD106 and AD104 chips, the performance difference, compared to the Ampere generation, should be between 20% and 25%. It is worth remembering that the more performance, the more consumption. If the RTX 3090 Ti has a TDP of 480W, what about an RTX 4090 then?

Written by: Joshua Wilson