MSN This startup has produced a way to use your gpu for higher bandwidth in RAID for SSDs, for the server market in particular. Its not yet compatible with Intel and AMD who, along with Nvidia, are still pounding out their own universal memory standard. Intel is winning in that department, but AMD is making significant contributions, and Nvidia is always on Mars, and unpredictable. Anywho, instead of the usual four SSDs, they can use your GPU to combine 32, and putting the pressure on the Big Boys, to finally come out with a single standard everyone else can adopt. They're all going for creating an All-In-Wonder chip, with both cheap and expensive memory combined, with distributing computing, and will come out with watered down and souped up versions, for servers. Basically, this gives you some idea of how much more powerful the next gen computers will be with serious AI circuitry added to boot. Of course, it will all be multiplexed to death, but if you add something like a NPU, neural processor, capable of 100,000 nodes, that's enough to seriously imitate the human brain, in a laptop with a thumb drive. The requirements for running a Star Trek holodeck have come down considerably, and its no longer a good standard for computing comparisons. They can hardwire a lot the holodeck now on the next generation video cards, requiring considerably less memory and processing power, and our chips are starting to resemble the way the brain is organized.
