Nvidia’s latest flagship RTX 4090 is said to have 100 TFlops of processing power. That’s 2.5 times more than the current flagship RTX 3090 Ti.
Flops is short for Floating Point Operations Per Second or floating point calculations per second. It is used to indicate the computing power of processors. To achieve 100 TFlops of processing power, the RTX 4090’s AD102 GPU would be equipped with 18,432 cuda cores clocked at 2.7 GHz. The GPU is expected to be partially disabled, but the clock speed will still be higher to reach 100 TFlops.
Currently, the RTX 3090 Ti takes about 40 TFlops of FP32 (single precision) computing power. It’s just that Nvidia isn’t expected to improve significantly in terms of computing power. AMD also wants to take a step with the RDNA 3 architecture. The downside is that new GPUs require even more energy. More than 600 watts of tdp is already spoken for the RTX 4090. For comparison: the TDP of the RTX 3090 Ti is 450 watts.
Source: Wccftech
Source: Hardware Info
