Nvidia has released its Tesla k80 which is equipped with 24Gb GDDR5 VRAM also the world’s fastest in terms of 2.9 TFlops, pumps out impressive 480GB/s bandwidth. But in Scientific research, things are even far beyond this performance. For example, look at the Supercomputer, which is giving its services in the Oak Ridge National Laboratory, there Titan supercomputer which housed 18,688 Tesla K20X GPUs and 18,688 AMD Opteron 6274 CPUs, built just for scientific research. So the computing demand is increasing day by day and firms who are responsible to entertain those laboratories are coming with their bad boys in order to meet the demand. And the latest news give us the clever hit that after two year things would be even far better then this.
NVIDIA Volta GPUs and IBM Power9 CPUs to Deliver Up to 300 PetaFlops in 2017 with SuperComputers
In 2017, we will be looking forward to two new supercomputers, the Summit from Oak Ridge National Laboratory and Sierra from Lawrence Livermore National Laboratory. Now both of these supercomputers have one thing in common, both of them will feature several next generation IBM POWER9 CPUs and also several NVIDIA Volta GPUs. For now Maxwell is ruling from Nvidia side and it will dominate the market till 2016, then Nvidia will come with the pascal which will make its entry in the Q2 of 2016 and it will stay in the market for 2 years just like Maxwell. Between the rising day’s of Pascal, Nvidia will release the new architecture namely Volta. At first Volta will entertain the professional Graphics card Tesla for first in 2017. It seems like Nvidia has their own roadmap which we have seen in the GTC2014. But still they are deciding on the process node as till now we don’t know that in the next year their Next Geration Graphics card would feature 20nm process node or 16nm.
Well, let’s come to the main topic of the Supercomputer. Some of the specifications are disclosed in order to maintain the heat, Summit supercomputer which is a computing beast eclipsing everyone behind. Rated at a peak performance of 150-300 PFLOPS, the Summit supercomputer would based on 3400 compute nodes with each node consisting of several next generation IBM POWER9 CPUs and NVIDIA Volta architecture based Tesla GPUs. Each node will deliver around 40 TFLOPs of computing performance and is said to be enough to outperform an entire rack of top of the line Haswell based x86 CPU servers.
First technology, we’ll announce today is an important invention called NVLink. It’s a chip-to-chip communication channel. The programming model is PCI Express but enables unified memory and moves 5-12 times faster than PCIe. “This is a big leap in solving this bottleneck,” Jen-Hsun says. NVIDIA
Each node will pack with 512GB memory and they will support DDR4 memory and HBM stacked design, good news is that, SK Hynix begins shipping its next-Gen 8GHz GDDR5, but this SuperComputer will feature far stronger than this in terms of memory frequency.