NVIDIA’s next-gen RTX 40 series “Ada Lovelace” graphics cards are slated to launch in the third quarter of the year (August/September). According to various rumors and leaks, these GPUs will roughly double the floating-point capabilities, thanks to TSMC’s N5 (5nm EUV) node and a robust microarchitecture. The AD102 die will reportedly accommodate up to 18,432 FP32 cores (ALUs), nearly twice as much as the Ampere top-die (GA102). In addition to this, the core clocks will be able to break the 2GHz barrier (even at stock), and level with AMD’s RDNA 2/3 offerings, with boost clocks approaching the 2.5GHz mark.
18K shaders pulsating at 2.5GHz resulting in a mammoth compute rating of 90 TFLOPs (FP32), 2.5x more than the RTX 3090, and a whopping 3x more than the RTX 3080. All these extra frame rates will come at a cost though. Even with the most advanced process node, and the world’s best engineers, squeezing in this kind of compute into a consumer die will result in a TDP of over 500W. Although not necessarily true for all models, the AIB overclocked models of the RTX 4090/4090 Ti might even exceed the 600W limit.
Read more on NVIDIA and AMD’s next-gen graphics cards below: