>Nvidia releases V100 Volta GPU featuring tensor cores>Those are advertized as "AI accelerators" for Neural network processing>AI is still not really cared about in 2017, ton of money is still in the desktop sector, while nvidia spent billions of making volta>Gpu's specialized die is expensive to manifacture, yet they need to somehow sell this to consumers>Idea! Take the most inefficent way of calculation (Ray tracing), and then push AI upscaling as the crutch>Repackage volta with "RT cores" (Aka what turing is) last moment and push rushed brute force raytracing that only runs passably with dedicated denoising and AI assisted>Charge double the prices of the last gen, while craming AI upscaling and ray tracing down the thorat of gamers that don't want it>Inspite of early resistance as this is when GPU prices started spiking and when games started slowly deteriorating in performance, goysumers slowly started expecting all the AI goyslop and RT in their games>AMD caught offguard cause they weren't prepared for making AI accelerators fused with GPUs, they were making simple GPUs (as they should be)Turing was a way for NVIDIA to pay off what they were building up to right now, AI. That's why GPUs grew in price year over year, most of the real development went into AI workloads, with you paying off it's early stages.Now that this poison is in the GPUs (muh upscaling, muh framgen), no company will drop this to make GPUs 50-40% cheaper again, until die manufacturing itself becomes cheaper.This is what they wanted physx to become btw, but they failed and had to drop this grift.
