Skip to main content

Posts

Showing posts with the label pascal

CES 2016 - Nvidia Drive PX will bring 8 tflops to your future car

Nvidia has officially anounced its in car supercomputer called Nvidia Drive PX 2 . The system , designed to give self driving cars the next generational push it need is set up using Nvidias new Pascal gpu which boasts of 8 tflops - this power however comes from 2 pascal GPU cores and 2 ARM based Tegra cores .details about the two tegra cores are not listed , but all we can say is it is based on ARM's 64bit cortex A57 architecture and each of them are configured as a quadcore Soc . The two Pascal processors on tbe other hand are designed using a 16nm FinFet technique and they generate the same power as 150 macbook pro's while remaining the same size as a meduim size lunchbox.  Via -Nvidia The system , equipped with its own cooling system ,the platform requires 250w .But considering that automakers will add this system to the cars battery , it will not be a issue at all .Volvo is said to be the first automaker out to make use of the Drive PX car computer  .Th...

Nvidia next gen Pascal GPU architecture to be available in 2016

After announcing details about the Tegra Erista mobile processor and a $3000 graphics card , Nvidia published new details about their upcoming GPU architecture called "Pascal" .The GPU design was previously known as "Volta" and it has been slowly replaced by "Pascal" instead (fate of Volta is unknown as of now) A slide showing the earlier codename "volta" - circa 2013 (via Anandtech) A slide from GTC 2014 showing the "Pascal" architecture after Maxwell (via Anandtech) The Pascal GPU is said be the next step in avoiding data bottlenecks when data transferred between the CPU and GPU slows down the processors ability to perform tasks . The new Nvidia NVLink is a integrated chip to chip communications technology that will eventually replace the PCIe standard by about 80GB per second compared to the 16GB on PCIe. The new Pascal system will also have stacked DRAM chips which are placed in dense modules with wide interfaces...