True I forgot to mention eGPU use cases. But realistically how many of those actually exist today? The cost of Thunderbolt eGPU enclosures has always been completely mad or involved extremely convoluted construction processes, not to mention that until Thunderbolt 3 ships in meaningful quantities there will be significant performance penalties. Seems kinda silly to me to bandwidth limit a flagship GPU part over TB2. Big enough market to justify paying a team of six-figure salaried engineers and QAs in Santa Clara to build and maintain? I really don't see it.
Our only hope is that Nvidia feel they need to keep maintaining the driver in the hope of an Apple design win for Pascal in the future. That's a much more solid business justification to my eyes.
Codding horror recently ran a great
blog post on high-end eGPU gaming over TB3. While it's technically completely viable, the TB3 dock was $499(!).