Perlmutter is the most powerful computer dedicated to artificial intelligence in the world. Launched at the Berkeley Lab in the United States, this supercomputer will be used for scientific research. Inside? Hewlett Packard Enterprise Engineering, AMD processors and 6,000 NVIDIA A100 GPUs.
With the collaboration of different technological brands, the laboratory has the second most powerful supercomputer in the world, only overtaken by the Japanese Fugaku. Of course, it is the most powerful for processing artificial intelligence data.
As stated by NVIDIA in a statement, the supercomputer is capable of reaching 180 PetaFLOPS (the Fugaku reaches 442 PetaFLOPS), which directly makes it the second most powerful supercomputer in the world. In addition, under additional performance conditions, it can achieve all four ExaFLOPS in the processing of AI.
Inside the computer Highlights include AMD’s high-end 64-core EPYC 7763 processors, in addition to the aforementioned 6,000 NVIDIA A100 GPUs. To unify everything, HPE’s Cray Shasta architecture was used. Although there is an interesting detail: it starts in two phases.
The first phase counts 1536 knots. Inside each is a 64-core AMD EPYC 7763 processor with 256 GB of DDR4 SDRAM and four 40 GB NVIDIA A100 GPUs. In this first phase, the supercomputer achieves 60 FP64 PetaFLOPS of performance.
Once it is implemented the second phase by the end of the year will add 3072 knots which will contain AMD EPYC 7763 processors and 512 GB of memory per node. With the implementation of the second phase, it is expected to reach 120 performance PetaFLOPS. the combined performance of the entire supercomputer will be 180 PetaFLOPS throughput and up to four FP16 throughput ExaFLOPS for AI applications.
What good is giving 6000 NVIDIA A100 GPUs
Simulations This is what the Perlmutter supercomputer will mainly do. It is expected to be the fastest system on the planet when it comes to handling artificial intelligence workloads. In one of his first projects will help build the largest 3D map of the Universe visible to this day.
In addition to the Universe mapping project, researchers will use the supercomputer to study subatomic interactions. A supercomputer like this can accurately simulate the behavior of atoms over time periods a little longer than the nanoseconds currently reached.
Via | NVIDIA
45seconds is a new medium, do not hesitate to share our article on social networks to give us a solid boost. 🙂