For those of you wondering about what makes NVIDIA’s upcoming GK110 tick, NVIDIA has posed a complete architectural White Paper. There is some seriously interesting contained in there, enough to make us want GK110 to make it over into the GeForce lineup as soon as possible.
However, as NVIDIA discusses quite a few times in this White Paper: GK110 is geared towards a compute / HPC environment and not gaming. That doesn’t mean it can’t be adapted for gaming like Fermi was though.
For those of you that aren’t inclined to read thousands of words, here is a “greatest hits” collection. Let us tell you, some of these specs are downright staggering:
- 7.1 billion transistors
- Up to 3x the performance per watt of Fermi
- 15 SMX modules, 2880 cores
- Six 64-bit memory controllers (384-bit)
- An enhanced memory subsystem, offering additional caching capabilities, more bandwidth at each level of the hierarchy, and a fully redesigned and substantially faster DRAM I/O implementation.
- 960KB of shared memory
- 1536KB of dedicated L2 cache memory
You can read it HERE (PDF)