Looking for GPU Benchmarking

[KindnessCUDA]

Hello community,

According to the official documentation, Parabricks is compatible with these GPU models:

• V100
• T4
• A10, A30, A40, A100, A6000
• L4, L40
• H100, H200
• Grace Hopper Superchip

I am looking for benchmarking results applied to the same input file that demonstrate the time and performance. Having such benchmarks is necessary for choosing some GPUs with a reasonable price/performance ratio.

Thank you

[KindnessCUDA]

Hello community,

Is there any technical data, benchmarking, or compression information available to help choose a GPU for Parabricks?

[kuhnsa3]

I have some benchmarking information using different allocations of A100 GPUs here for the parabrick’s germline workflow:

Very low Recall for HG002 germline benchmark

Is there anyone on NIVIDA’s side that can look into this? This seems like a potential issue with the tool that may need to be resolved. Have you benchmarked parabricks germline callers against any of the GIAB samples using hap.py? This is much lower than I would have expected. If I run a vanilla GATK germline workflow (using The Broad’s GATK) against the same input, I am seeing higher recall (see screen shot below). I also benchmarked the parabricks deepvariant germline caller and its performan…

Have you benchmarked any of the germline workflows (either GATK or deepvariant) against the GIAB samples? I am seeing very low recall with parabricks when compared to the tools available from The Broad or Google.

[KindnessCUDA]

Hi @kuhnsa3, Thank you for your reply!
Yes, I did run the germline pipeline with an A100 GPU against HG002, but I’m not sure how to assess the performance and accuracy of the outputs. Could you please provide some instruction?