Currently there would likely be little direct benefit for Terragen rendering...
However, as mass storage speeds increase and we are more and more able to access really large storage volumes without delays, we can then consider relying on this for aspects of rendering or other parts of the computation process. For example we currently load all scene elements into main memory for quickest access during the rendering process. This can mean that for really complex scenes featuring many high resolution external terrain, texture, and object files, memory use can get really high, so you might need 32GB or even more of main RAM to render some scenes. In the future it may not be a bottleneck anymore to just leave these things on the primary storage medium and access them directly, which would potentially allow for even more complex scenes, or today's most complex scenes rendering comfortably on lesser hardware (i.e. less RAM).
That being said the speeds and storage sizes that are quoted in that article really don't seem impressive as compared to even currently available NVme SSDs. Unless they're *per chip*, in which case it could be very impressive when e.g. 16x32GB chips are combined together to make a 512GB drive, especially if the 1.6GB/s transfer rate shown is again per-chip and it can be parallel when packaged together. That would make for a theoretical bandwidth of nearly 26GB/s. Since that is around current RAM speeds, I am guessing that's not actually possible, but if it were it'd be a big deal.
The long-term future of 3d xpoint does seem promising in any case because it offers some notable advantages over current NAND (what most SSDs are made from currently). If this first generation debuts at a data transfer rate similar to NVme, perhaps future generations can evolve more quickly and outpace NAND in the long-term while maintaining the benefits of longer lifespan and more fine-grained data access.
- Oshyan
