Just think about it. PS5's SSD raw/uncompressed transfer rate of 5.5 GB/s. There is no way in hell that the SSD is going to be able to immediately access a game's data for rendering, A.I and gameplay fast enough in a given area, much less for the entire game. The CPU and GPU can't use the SSD to render or process from, because it's too slow; particularly when it comes to latency. Bandwidth is one thing, and latency is another. DRAM is hundreds of thousands of times faster (measured in ns) than an SSD (measured in ms) when it comes to latency. As for the how: 1) Massive multicore CPUs. We already have up to 16 core CPUs for mainstream parts using the Zen2 core. Future Intel and AMD CPUs will have even more cores and be more powerful. 2) DDR5. DDR5 will have twice the capacity DIMMs as DDR4, plus much greater bandwidth. 3) PCIe 16x bridge between VRAM and system memory. With PCIe 5.0, that's 128GB/s of bidirectional transfer and with PCIe 4.0, it's 64GB/s. Either way, it's far faster than the PS5's SSD from both a bandwidth and latency perspective. 4) Huge VRAM buffers solely for the GPU. Nuff said! 5) Faster SSDs. I think PCIe 4.0 x4 will top out at 8GB/s, while PCIe 5.0 x4 will be 16GB/s. 5) Predictive algorithms. Obviously this isn't just a hardware issue. Lower level APIs like Vulkan and DX12 makes it possible to have much more intelligent streaming algorithms. For a PC that uses disparate memory pools, this can really improve the overall efficiency and speed of the streaming. 6) Volume tiled resources. As more games use this technique, you can expect less pop in, faster streaming and more efficient use of VRAM. Current examples of this off the top of my head are Gears 4 and Gears 5, both of which had no noticeable pop in and performed excellently on PC with moderate VRAM usage.