ExxonMobil to Deploy Discovery 6 Supercomputer with NVIDIA GH200 Superchips

American ExxonMobil, one of the world’s largest oil companies, has presented a project for a high-performance computing complex, Discovery 6. This supercomputer is planned to be used to support the operation of advanced 4D seismic exploration systems.

The complex is being created in partnership with HPE and NVIDIA. The system will be based on the HPE Cray Supercomputing EX4000 platform. 4032 NVIDIA GH200 Grace Hopper superchips will be used. Grace Hopper includes a 72-core NVIDIA Grace Arm processor, up to 480 GB of LPDDR5x and an NVIDIA H100 accelerator with 96 or 144 GB of HBM3(e). The use of the HPE Slingshot interconnect is discussed.

It is noted that the new supercomputer will be approximately four times more productive than its predecessor, the Discovery 5 system, which was put into operation in November 2022. This complex uses the HPE Cray EX235n platform with 32-core AMD EPYC 7543 processors and NVIDIA A100 SXM4 accelerators. In the November TOP500 rating, the Discovery 5 machine occupied the 46th position with a peak performance of 30.99 Pflops. Thus, the performance of Discovery 6 is expected to be at the level of 120 Pflops.

Image source: ExxonMobil

The supercomputer being developed is designed to use a highly efficient direct liquid cooling (DLC) system. With the launch of Discovery 6, ExxonMobil expects to reduce the processing time for complex seismic data from months to weeks. This will help improve the efficiency of oil and gas exploration, which will lead to increased resource production with lower capital investment. The installation of Discovery 6 is scheduled to be completed in the first half of this year.