domain | gcore.com |
summary | **Summary:**
- **Goal:** To assess inference speed on NVIDIA's H100 GPU for genomic processing. - **Methodology:** Used an NVIDIA-developed model without customization. - **Collaboration Benefits:** - Up to 50x faster variant calling tasks. - Cost savings via on-demand GPUs, avoiding upfront hardware investments. - Scalable infrastructure with dynamic resource allocation. - Data sovereignty and security in a compliant cloud environment. - **Processing Times (H100 GPUs):** - Range from 13 minutes (haplotypecaller with 2 GPUs) to nearly 90 minutes (Deepvariant with 8 GPUs). - GPU overhead can slow processing time when memory limits are reached. - **Future Plans:** - RTU HPC plans to expand AI applications beyond genomics using Cloud GPUs. - Potential collaboration with Gcore for optimizing GPU configurations and exploring AI Inference opportunities. - **Global Impact:** BMC's work contributes to the international 1 Million Genomes Project, making genomic information more accessible. - **Benchmark Results (H100 GPUs):** - fq2bam and haplotypecaller showed shortest processing times with consistent results across 2, 4, and 8 GPUs. - **Future Vision:** Liepa sees the future of AI in healthcare as fast, flexible, and sovereign. |
title | Gcore | Global Hosting, CDN, Edge and Cloud Services |
description | We provide powerful solutions that will help your business grow globally. Try our superior performance for free |
keywords | cloud, performance, infrastructure, high, game, saber, streaming, platform, demand, edge, servers, more, scale, metal, team, network, time |
upstreams |
|
downstreams |
|
nslookup | A 5.188.126.61 |
created | 2025-04-12 |
updated | 2025-05-14 |
summarized | 2025-05-14 |
|
|