Table 7.

Extended performance results obtained on equivalent hardware configuration used in the iDASH competition, more specifically Intel^® Xeon^® Gold 6140 (see “Computation environments” section for more details). The results illustrate how latency and throughput scales when more than 4 CPU cores are utilized for parallel processing of the workload

	Latencya(s)				Throughput2(q/s)
	# Cores				# Cores
Method	4	8	16	32	4	8	16	32
“Unsupervised method” section	6.24	3.43	2.44	2.17	64	117	164	184
“Clustering-based supervised method” section	7.53	4.26	2.83	2.38	53	94	141	168
“Linear regression method” section	5.92	3.06	2.26	1.85	68	131	177	216

Disclaimer: Performance varies by use, configuration and other factors. Learn more at www.Intel.com/PerformanceIndex.

Latency of the different approaches described in “Methods” section running on multiple cores.

Throughput of the different approaches described in “Methods” section running on multiple cores