San Jose, Calif. : NVIDIA announced another big leap in producing state-of-art GPUs. Introducing the new NVIDIA-NVLink, which will be integrated in its future GPUs, the company promised to have the faster data transfer in between CPU and GPU. Making it about 10 times faster, the new GPUs will eliminate the issues with bottleneck and will make super computers even more powerful.
“NVLink technology unlocks the GPU’s full potential by dramatically improving data movement between the CPU and GPU, minimizing the time that the GPU has to wait for data to be processed,” said Brian Kelleher, senior vice president of GPU Engineering at NVIDIA.
Due to the bottleneck in between CPU and GPU, the yield from the current set of computer is very low. The exploitation of PCIe results in the delay to access CPU’s memory and the results provided are comparatively slower by 5-10 times.
Being developed in collaboration with the IBM, company’s spokesperson said, “NVLink makes it easier for developers to modify high-performance and data analytics applications to take advantage of accelerated CPU-GPU systems. We think this technology represents another significant contribution to our OpenPOWER ecosystem.”
Bradley McCredie, vice president IBM, also said that, “NVLink enables fast data exchange between CPU and GPU, thereby improving data throughput through the computing system and overcoming a key bottleneck for accelerated computing today,”
The GPU will be coming with the concept of unified memory which allows the programmer to treat CPU and GPU memory as a single unit. A programmer than need not to worry about the data storage, whether its in CPU’s or GPU’s memory is not much of concern.
NVLink which is differential with embedded clock will be based on the PCIe programming model with DMA+ or Direct Memory Access. It also has an unified memory with a cache coherency in Gen 2.0.
These new GPUs can be used to produce a huge computation power and will boost the performance of super computers by 50-100x times.
[Pictures : legitreviews ]