Our CEO, Jen-Hsun Huang, sketched out today a future where mobile, desktop and supercomputer technologies intersect in powerful and surprising ways.
Speaking at the GPU Technology Conference, Huang described how GPUs are already finding their way into applications that were undreamed of a decade ago.
NVIDIA’s Kepler GPU architecture — introduced last year — is not only a runaway hit with gamers. It underpins a new generation of hyper-efficient supercomputers, including Titan, the world’s fastest system.
Next up: new GPUs and mobile processors that promise to put the efficiency and speed of our massively-parallel GPU architectures into an ever-broader array of devices.
- Maxwell will offer unified virtual memory, giving CPUs access to the speedy memory built into GPUs, and vice versa.
- In Volta, due after Maxwell, memory modules will be piled atop one another and placed on the same silicon substrate as the GPU core itself – a radical idea called ‘stacked DRAM’ — giving these new GPUs access to up to one terabyte per second of bandwidth. That’s enough to move the equivalent of a full Blu-Ray disc worth of data through a chip in just 1/50th of a second. “Unbelievable stuff,” Huang said.
In mobile processors:
- Logan will pair ARM-based mobile processor cores with our powerful Kepler GPUs, putting technologies now found in high-performance PCs and workstations – such as PhysX, CUDA 5, and Open GL 4.3 – into mobile devices.
- Parker will join new 64-bit ARM compatible CPU cores with our next-generation Maxwell GPU architecture – combining the ability to gulp down big chunks of data, like a server, with Maxwell’s ability to mix and match memory resources with CPUs.
Note: an earlier version of this post described Volta’s use of ‘stacked DRAM’ incorrectly.