NVIDIA Volta GPUs and IBM Power9 CPUs To Deliver Up To 300 PetaFlops of Performance in 2022 With Summit and Sierra Supercomputers

On 29th Oct 2022, Oak Ridge National Laboratory made functional their Titan supercomputer which housed xviii,688 Tesla K20X GPUs and 18,688 AMD Opteron 6274 CPUs, built just for scientific research. So amazing was its compute capability that NVIDIA's next big graphics card was named afterwards information technology. Two years take passed since the supercompter became functional and work is underway on edifice the side by side large matter after Titan.

NVIDIA Volta GPUs To Power Next Big Supercomputers in 2022

In 2022, nosotros will be looking forward to ii new supercomputers, the Top from Oak Ridge National Laboratory and Sierra from Lawrence Livermore National Laboratory. Now both of these supercomputers accept 1 thing in common, both of them will feature several next generation IBM POWER9 CPUs and besides several NVIDIA Volta GPUs. But before we go into details, in that location's ane interesting affair I like to point out. We all know Maxwell is here to stay for at least first half of 2022 before NVIDIA transitions to their adjacent architecture code named Pascal in second one-half of 2022. Pascal is expected to stay its term for two years till 2022 when NVIDIA will replace information technology with their next generation Volta architecture. However, according to the data from NVIDIA, Volta volition make its way to the professional market i twelvemonth early in 2022 which ways NVIDIA is on track with their GPU roadmap that was displayed at GTC 2022.

That'southward good news for all and next upward, we will detail some of the performance and specifications of the Summit supercomputer which is a calculating beast eclipsing everyone behind. Rated at a summit performance of 150-300 PFLOPS, the Elevation supercomputer volition be based on more than 3400 compute nodes with each node consistin of several next generation IBM POWER9 CPUs and NVIDIA Volta architecture based Tesla GPUs. Each node will deliver around twoscore TFLOPs of compute functioning and is said to exist enough to outperform an entire rack of top of the line Haswell based x86 CPU servers.

With Summit come up a few other enabling technologies such every bit NVLINK which is expected to debut effectually 2022. The Pascal GPU would be the first to innovate NVLINK which is the side by side generation Unified Virtual Retention link with Gen 2.0 Cache coherency features and 5 – 12 times the bandwidth of a regular PCIe connection. This will solve many of the bandwidth problems that high performance GPUs currently face up.

Offset applied science we'll announce today is an of import invention chosen NVLink. It's a chip-to-chip communication channel. The programming model is PCI Limited but enables unified memory and moves 5-12 times faster than PCIe. "This is a large leap in solving this bottleneck," Jen-Hsun says. NVIDIA

On summit, each compute node will characteristic 512 GB of memory through dense server DDR4 retentivity and high bandwidth HBM stacked designs which will work in complete coherency. Extending the impressive amount of memory is an additional 800 GB of NVRAM per node, which can be configured either as a burst buffer or every bit extended memory. The system is interconnected with dual-rail Mellanox EDR InfiniBand, using the full, non-blocking fat-tree design. Summit volition as well include the GPFS parallel file system with 1 TB/s of I/O bandwidth and 120 PB of disk capacity. NVIDIA Tesla K80 Slide_6

NVLink is an energy-efficient, high-bandwidth communications aqueduct that uses up to three times less energy to move data on the node at speeds five-12 times conventional PCIe Gen3 x16. Starting time bachelor in the NVIDIA Pascal GPU architecture, NVLink enables fast communication between the CPU and the GPU, or between multiple GPUs. Figure 3: NVLink is a key building block in the compute node of Pinnacle and Sierra supercomputers.

VOLTA GPU Featuring NVLINK and Stacked Memory NVLINK GPU high speed interconnect eighty-200 GB/s 3D Stacked Retentiveness 4x Higher Bandwidth (~1 TB/s) 3x Larger Chapters 4x More Energy Efficient per chip.

NVLink is a key applied science in Tiptop's and Sierra's server node compages, enabling IBM Ability CPUs and NVIDIA GPUs to admission each other'due south memory fast and seamlessly. From a developer's perspective, NVLink erases the visible distinctions of data separately attached to the CPU and the GPU by "merging" the memory systems of the CPU and the GPU with a loftier-speed interconnect. Because both CPU and GPU take their ain retentiveness controllers, the underlying memory systems can be optimized differently (the GPU's for bandwidth, the CPU'southward for latency) while all the same presenting as a unified memory system to both processors. NVLink offers ii singled-out benefits for HPC customers. First, it delivers improved awarding performance, only by virtue of greatly increased bandwidth between elements of the node. Second, NVLink with Unified Retentiveness technology allows developers to write lawmaking much more seamlessly and nonetheless achieve high performance. via NVIDIA News

The most impressive feature about Meridian will exist that while information technology will consume ten% more than power than the Titan supercomputer, it will also deliver an impeccable 5 times more than system to application power than the previous fastest supercomputer. While Titan was rated at 25-30 PETAFLOPs, Sierra will be >100 PFlops while Summit will be 150-300 PFlops at any given circumstance.