The US has issued $258m in research contracts to develop the first Exascale supercomputer
The US has issued $258m in research contracts to develop the first Exascale supercomputer
Now the US’ Department of Energy has announced that they will be delivering six research grants that are worth a total of $258 million over three years to six different companies. These companies are AMD, Intel, Nvidia, IBM, Cray Inc and Hewlett-Packard Enterprise.
Companies that receive this funding will need to provide a minimum of 40% of their total project costs, which means that the total investment in Exascale computing is $430 million.
An Exascale computer is crazy fast, especially when considering the power of modern hardware, imagine AMD’s Radeon Vega, a GPU that offers roughly 12.5 TFLOPS of FP32 compute performance and 25 TFLOPS of FP16 compute performance. A single ExaFLOP of GPU performance would require a total of 40,000 Radeon Vega GPUs, assuming that every GPU runs at peak efficiency.
Right now, the world’s fastest supercomputer is the Sunway TaihuLight, which runs at a max speed of 93.01 PetaFLOPS (thousand TFLOPS), which is a long way from what is required to achieve the magic ExaFLOP performance milestone. A lot of innovation is required over the next few years to make Exascale computing happen, but a lot of what is learned will likely help companies develop more powerful cunsumer products moving forward.
You can join the discussion on the US’ plans to create an Exascale Computer by 2021 on the OC3D Forums.
The US has issued $258m in research contracts to develop the first Exascale supercomputer
Now the US’ Department of Energy has announced that they will be delivering six research grants that are worth a total of $258 million over three years to six different companies. These companies are AMD, Intel, Nvidia, IBM, Cray Inc and Hewlett-Packard Enterprise.
Companies that receive this funding will need to provide a minimum of 40% of their total project costs, which means that the total investment in Exascale computing is $430 million.
An Exascale computer is crazy fast, especially when considering the power of modern hardware, imagine AMD’s Radeon Vega, a GPU that offers roughly 12.5 TFLOPS of FP32 compute performance and 25 TFLOPS of FP16 compute performance. A single ExaFLOP of GPU performance would require a total of 40,000 Radeon Vega GPUs, assuming that every GPU runs at peak efficiency.
Right now, the world’s fastest supercomputer is the Sunway TaihuLight, which runs at a max speed of 93.01 PetaFLOPS (thousand TFLOPS), which is a long way from what is required to achieve the magic ExaFLOP performance milestone. A lot of innovation is required over the next few years to make Exascale computing happen, but a lot of what is learned will likely help companies develop more powerful cunsumer products moving forward.
You can join the discussion on the US’ plans to create an Exascale Computer by 2021 on the OC3D Forums.