Using its Second Generation cloud, Oracle Cloud Infrastructure (OCI) has completed the best-known benchmark for NVIDIA Parabricks on any cloud. This speed and compute performance supports quick and accurate genome analysis for healthcare and researchers, which can help save lives.
NVIDIA Parabricks is the only GPU-accelerated computational genomics toolkit that delivers fast and accurate analysis for sequencing centers, clinical teams, genomics researchers, and next-generation sequencing instrument developers. Parabricks offers up to 80-time speedups over CPU-only solutions, reducing computing costs by up to 50%. The toolkit includes full compatibility with workflow languages and managers, including WDL, NextFlow, and Cromwell, to easily intertwine GPU- and CPU-powered tasks.
Major cloud providers benchmarking NVIDIA Parabricks use all eight NVIDIA A100 Tensor Core GPUs in a single node. The standard is based on germline variant calling with HaplotypeCaller, going from FASTQ to VCF, for the 30-time coverage whole human genome sequence data set HG002.novaseq.pcr-free.30x.
Record-setting benchmark results
Utilizing all eight NVIDIA A100 GPUs in a single node, Oracle achieved a record-breaking 19.2 minutes for running the whole germline pipeline. Using four NVIDIA A100 GPUs, Oracle achieved a record breaking 32.9 minutes. In considering production deployments of Parabricks, Oracle also produced a record-breaking 33.1 minutes using four NVIDIA A10 Tensor Core GPUs.
By combining industry leading performance for genome analysis with OCI’s lower, simpler, and more predictable pricing, Oracle can help organizations deliver better patient outcomes and at a lower cost.
Want to learn more?
Join us at HIMSS 2023, the Global Health Conference and Exhibition, from April 17–21. Visit the Oracle booth at HIMSS to learn more about NVIDIA Parabricks, these record-breaking results, and how OCI is using AI and machine learning to transform healthcare. For more information on Oracle Cloud Infrastructure Compute, see OCI Compute: GPU instances.