Table 12.

Same as Table 11, but for the RIB benchmark.

No. of nodes	Processor(s) Intel	GPUs, IB	DD		Grid	N PME /node	N th		DLB	P (ns/d)	E
			x	y	z
1	E3‐1270v2	770,	1	1	1	–	8	(ht)	–	0.91	1
2	(4 cores)	QDR[a]	2	1	1	–	8	(ht)	(✓)	1.87	1.03
4			4	1	1	–	8	(ht)	(✓)	2.99	0.82
8			8	1	1	–	8	(ht)	(✓)	4.93	0.68
16			16	1	1	–	8	(ht)	(✓)	4.74	0.33
32			16	2	1	–	8	(ht)	(✓)	10.3	0.35
1	E5‐2670v2	780Ti×2,	8	1	1	–	5	(ht)	✗	4.02	1
2	(2×10 cores)	QDR	20	1	1	–	4	ht	✗	6.23	0.77
4			8	5	1	–	4	ht	✗	10.76	0.67
8			16	10	1	–	2	ht	✗	16.55	0.51
16			16	10	1	–	2		✗	23.78	0.37
32			16	10	2	–	2		✗	33.51	0.26
1	E5‐2670v2	980×2,	8	5	1	–	1	ht	(✓)	4.18	1
2	(2×10 cores)	QDR	20	1	1	–	4	ht	✗	6.6	0.79
4			8	5	1	–	4	ht	✗	11	0.66
1	E5‐2680v2	–	10	3	1	10	1	ht	✓	1.86	1
2	(2×10 cores)	FDR‐14	10	3	1	5	2	ht	✓	3.24	0.87
4			10	2	3	5	2	ht	✓	6.12	0.82
8			8	5	3	5	2	ht	✓	12.3	0.83
16			10	8	3	5	2	ht	✓	21.8	0.73
32			10	7	7	4.69	2	ht	✓	39.4	0.66
64			16	10	6	5	1		✓	70.7	0.59
128			16	16	8	4	1		✓	128	0.54
256			16	17	15	4.06	1		✓	186	0.39
512			20	16	13	1.88	2		✗	208	0.22
1	E5‐2680v2	K20X×2	20	1	1	–	2	ht	✗	3.99	1
2	(2×10 cores)	(732 MHz),	10	8	1	–	1	ht	✗	5.01	0.63
4		FDR‐14	10	8	1	–	2	ht	✗	9.53	0.6
8			16	10	1	–	2	ht	✗	16.2	0.51
16			16	10	1	–	2		✗	27.5	0.43
32			8	8	1	2	5		✗	49.1	0.38
64			16	8	1	2	5		✗	85.3	0.33
128			16	16	1	2	5		✗	129.7	0.25
256			16	8	4	2	5		✗	139.5	0.14

[a] Note: These nodes cannot use the full QDR IB bandwidth due to insufficient number of PCIe lanes, see “Strong Scaling” section.