Discussion in 'Article Discussion' started by Gareth Halfacree, 26 Feb 2014.
Gen2 spec doubles the interface rate.
Nvidia stated they would be using the cubes for Maxwell, I assume AMD will do the same - looks like V2 could even make it in time for a Maxwell refresh later in the product life cycle.
Will be interesting to see how much all this additional bandwidth improves performance.
Memory can quite easily be a bottleneck in a GPU, especially in GPGPU applications.
I'm quite ignorant of the technology, here ... so forgive if this is a foolish question: can this tech be used to replace L2 or L3 cache on a chip, and if so, wouldn't that give some awesome performance gains for CPU / GPGPU tech? Or is this only relevant to DIMM-type memory?
Vaporware performance doubles, amazing!
Separate names with a comma.