Wow, the new RTX cards have really gone all in on the ray tracing cores. The main tweak is the larger, more flexible RT cores—about a 40% increase in core count compared to the previous generation. That means each core can handle a ray intersection or shadow test in fewer cycles, so you get more rays per second, which translates to a noticeable lift in frame rates for ray‑traced scenes. But it’s not just raw count. They re‑engineered the cache hierarchy to keep the most frequently accessed geometry in a tighter, lower‑latency L1 cache. That reduces the bandwidth hammering on the memory bus, so you see a dip in power draw when you hit the same workloads. On top of that, the new micro‑architecture introduces a dynamic clock scaling for the RT units: when a frame has fewer rays, the cores can throttle down faster, cutting idle energy waste. The kicker is how they balanced the two. For full‑ray‑traced titles, you see a performance bump of around 15–20% over the older cards, while power consumption per ray falls by roughly 12%. In mixed workloads—say, when only a small portion of the frame is ray‑traced—those RT cores still sit in low‑power mode, so the overall energy budget doesn’t balloon. Of course, if you push the RT cores hard with insane scene complexity, you still hit the thermal ceiling, so the cooling solution needs to be top tier. So, in short: more cores, smarter cache, dynamic scaling = better performance for heavy ray tracing and cleaner power efficiency when you’re not fully engaged. If you’re building a system that will run those next‑gen titles, make sure the VRMs and cooling can handle that sweet spot.