Nvidia just had their RTX 4000 series keynote, and I watched the whole thing so you don’t have to—here’s the rundown in under six minutes, workstation-builder style. The big headlines were the RTX 4090 and the RTX 4080 (two variants: 12GB and 16GB). On the 4090, the specs look seriously promising: it’s packing a huge CUDA core bump over the 3090 Ti, and Nvidia’s talking about ~2.5GHz boost clocks (and even claims of 3GHz overclocks in their lab). Jensen also threw out a pretty wild ray tracing uplift claim—roughly 2–4x over the 3090—which, if it holds up in real workloads, is a big deal for anyone doing GPU-heavy rendering. What surprised me most is power. Everyone was expecting absurd numbers, but Ada Lovelace is being positioned around 450W for the 4090—still a lot, but not the “700W space heater” rumor mill. A lot of that efficiency story comes from the process shift: Samsung 8nm to TSMC N4. I also covered the architectural upgrades: DLSS 3’s frame prediction (which can boost FPS and offload work from the CPU), shading execution reordering (still waiting on more details), and improvements to RT cores, tensor cores, and shader cores. On pricing, I get why people are mad—$1600 for a 4090 is expensive. But I don’t automatically call it “overpriced” when you factor inflation, TSMC costs, and that this is a premium tier product most people don’t even need.