1 min readTimeLogic Achieves the Performance of 471 CPU-Cores for HMM Searches Using Single FPGA Accelerator
According to the announcement, search tests were run to compare the performance range, flexibility and scalability of the new SeqCruncher field programmable gate array (FPGA) hardware and improved DeCypherHMM software against the established standard open-source software, HMMER from Dr. Sean Eddy (HHMI). SeqCruncher provided 471x, 863x, and 1151x faster performance per CPU core with 1, 2 and 3 cards when searching an HMM database.
“With one SeqCruncher accelerator installed in a 2 CPU server, we were able to compare 1,000 proteins from the JCVI Global Ocean Sampling Expedition to the Washington University St. Louis Pfam-fs database in 44 seconds. Running HMMER software on the same 8-CPU core server took 43 minutes. This performance makes our system ideal for any laboratory engaged in metagenomics or high-throughput protein annotation,” commented Dr. Darryl Leon, Product Manager, Informatics at Active Motif.
The SeqCruncher also supports BLAST, Smith-Waterman and gene-modelling tools for bioinformatics, and Active Motif has begun exploring ways to apply this TimeLogic technology to computing burdens relating to next-generation sequencing.