NVIDIA Announces Next-Generation Pascal GPU

According to NVIDIA, Pascal GPUs will have three main characteristics that will make them faster and more accurate for training artificial neural networks. They are mixed-precision computing, 3D memory and NVLink.

The addition of mixed-precision computing enables Pascal GPUs to compute at 16-bit floating point accuracy at twice the rate of 32-bit floating point accuracy.

With 3D memory, users can access three times the bandwidth and approximately three times frame buffer capacity of Maxwell GPUs. The hardware of Pascal also changes, with its memory chips stacked on top of each other adjacent to the GPU.

Furthermore, the addition of NVLink will let data move five to 12 times faster than today's current PCIe standard, NVIDIA states. It allows for double the number of GPUs in a system to work on deep learning computations.

“It will benefit from a billion dollars worth of refinement because of R&D done over the last three years,” said Jen-Hsun Huang, co-founder and CEO of NVIDIA in his keynote address.

The Pascal GPU is set to debut next year.

For more information, visit NVIDIA.

Learn more about the GPU Technology Conference.

Sources: Press materials received from the company and additional information gleaned from the company’s website.