SPEC® MPIM2007 Result

SGI

SGI ICE X (Intel Xeon E5-2690 v2, 3.0 GHz)

SPECmpiM_peak2007 = 103

SPECmpiM_base2007 = 100

MPI2007 license:	4	Test date:	Dec-2013
Test sponsor:	SGI	Hardware Availability:	Sep-2013
Tested by:	SGI	Software Availability:	Nov-2013

Results Table

Benchmark	Base							Peak
Benchmark	Ranks	Seconds	Ratio	Seconds	Ratio	Seconds	Ratio	Ranks	Seconds	Ratio	Seconds	Ratio	Seconds	Ratio
Results appear in the order in which they were run. Bold underlined text indicates a median measurement.
104.milc	800	11.4	138	11.4	138	11.3	138	800	11.4	138	11.4	138	11.3	138
107.leslie3d	800	34.2	153	34.2	153	34.3	152	800	34.2	153	34.2	153	34.3	152
113.GemsFDTD	800	215	29.4	216	29.2	215	29.3	160	156	40.3	156	40.3	157	40.2
115.fds4	800	12.5	156	11.3	173	12.5	156	640	11.6	167	11.5	170	11.8	166
121.pop2	800	105	39.3	105	39.4	105	39.3	800	105	39.3	105	39.4	105	39.3
122.tachyon	800	23.2	120	23.6	119	23.3	120	800	23.2	120	23.6	119	23.3	120
126.lammps	800	102	28.6	102	28.6	102	28.6	160	102	28.7	101	28.8	101	28.8
127.wrf2	800	35.2	222	35.4	220	35.2	222	800	35.2	222	35.4	220	35.2	222
128.GAPgeofem	800	10.5	197	11.0	187	11.1	186	800	10.5	197	11.0	187	11.1	186
129.tera_tf	800	23.8	116	23.4	118	23.7	117	800	23.8	116	23.4	118	23.7	117
130.socorro	800	41.1	93.0	41.1	93.0	41.3	92.4	800	41.1	93.0	41.1	93.0	41.3	92.4
132.zeusmp2	800	25.1	123	25.0	124	25.1	123	800	25.1	123	25.0	124	25.1	123
137.lu	800	25.6	144	25.5	144	25.7	143	800	25.6	144	25.5	144	25.7	143

Hardware Summary
Type of System:	Homogeneous
Compute Node:	SGI ICE X IP-113 Compute Node
Interconnect:	InfiniBand (MPI and I/O)
File Server Node:	SGI Modular InfiniteStorage Server
Total Compute Nodes:	40
Total Chips:	80
Total Cores:	800
Total Threads:	1600
Total Memory:	2560 GB
Base Ranks Run:	800
Minimum Peak Ranks:	160
Maximum Peak Ranks:	800

Software Summary
C Compiler:	Intel C++ Composer XE 2011 for Linux, Version 14.0.0.080 Build 20130728
C++ Compiler:	Intel C++ Composer XE 2011 for Linux, Version 14.0.0.080 Build 20130728
Fortran Compiler:	Intel Fortran Composer XE 2011 for Linux, Version 14.0.0.080 Build 20130728
Base Pointers:	64-bit
Peak Pointers:	64-bit
MPI Library:	SGI MPT 2.09
Other MPI Info:	OFED 1.5.2
Pre-processors:	None
Other Software:	None

Node Description: SGI ICE X IP-113 Compute Node

Hardware
Number of nodes:	40
Uses of the node:	compute
Vendor:	SGI
Model:	SGI ICE X IP-113 (Intel Xeon E5-2690 v2, 3.0 GHz)
CPU Name:	Intel Xeon E5-2690 v2
CPU(s) orderable:	1-2 chips
Chips enabled:	2
Cores enabled:	20
Cores per chip:	10
Threads per core:	2
CPU Characteristics:	Ten Core, 3.0 GHz, 8.0 GT/s QPI Intel Turbo Boost Technology up to 3.60 GHz Hyper-Threading Technology enabled
CPU MHz:	3000
Primary Cache:	32 KB I + 32 KB D on chip per core
Secondary Cache:	256 KB I+D on chip per core
L3 Cache:	25 MB I+D on chip per chip
Other Cache:	None
Memory:	64 GB (8 x 8 GB 2Rx4 PC3-14900R-13, ECC)
Disk Subsystem:	None
Other Hardware:	None
Adapter:	Mellanox MT27500 with ConnectX-3 ASIC (PCIe x8 Gen3 8 GT/s)
Number of Adapters:	2
Slot Type:	PCIe x8 Gen3
Data Rate:	InfiniBand 4x FDR
Ports Used:	2
Interconnect Type:	InfiniBand

Software
Adapter:	Mellanox MT27500 with ConnectX-3 ASIC (PCIe x8 Gen3 8 GT/s)
Adapter Driver:	OFED-1.5.2
Adapter Firmware:	2.11.312
Operating System:	SUSE Linux Enterprise Server 11 SP2, Kernel 3.0.80-0.7-default
Local File System:	NFSv3
Shared File System:	NFSv3 IPoIB
System State:	Multi-user, run level 3
Other Software:	SGI Tempo Compute Node 2.7.3, Build 708rp14.sles11sp2-1305311204

Node Description: SGI Modular InfiniteStorage Server

Hardware
Number of nodes:	1
Uses of the node:	fileserver
Vendor:	SGI
Model:	SGI Modular InfiniteStorage Server
CPU Name:	Intel Xeon E5-2670
CPU(s) orderable:	1-2 chips
Chips enabled:	2
Cores enabled:	16
Cores per chip:	8
Threads per core:	2
CPU Characteristics:	Intel Turbo Boost Technology up to 3.33 GHz Hyper-Threading Technology enabled
CPU MHz:	2600
Primary Cache:	32 KB I + 32 KB D on chip per core
Secondary Cache:	256 KB I+D on chip per chip
L3 Cache:	20 MB I+D on chip per chip
Other Cache:	None
Memory:	128 GB (8 * 16 GB 2Rx4 PC3-12800R-11, ECC)
Disk Subsystem:	64.8 TB RAID 6 72 x 900 GB SAS (Western Digital, 10K RPM)
Other Hardware:	None
Adapter:	Mellanox MT27500 with ConnectX-3 ASIC (PCIe x8 Gen3 8 GT/s)
Number of Adapters:	2
Slot Type:	PCIe x8 Gen3
Data Rate:	InfiniBand 4x FDR
Ports Used:	2
Interconnect Type:	InfiniBand

Software
Adapter:	Mellanox MT27500 with ConnectX-3 ASIC (PCIe x8 Gen3 8 GT/s)
Adapter Driver:	OFED-1.5.0
Adapter Firmware:	2.11.312
Operating System:	SUSE Linux Enterprise Server 11 SP3 Kernel
Local File System:	xfs
Shared File System:	--
System State:	Multi-user, run level 3
Other Software:	SGI Foundation Software 2.9, Build 700r3.sles11-1004061553

Interconnect Description: InfiniBand (MPI and I/O)

Hardware
Vendor:	Mellanox Technologies and SGI
Model:	None
Switch Model:	SGI FDR Integrated IB Switch Blade 2SW9x27 with Mellanox SwitchX device 51000
Number of Switches:	12
Number of Ports:	36
Data Rate:	InfiniBand 4x FDR
Firmware:	07130007_LLR and 08130007_LLR
Topology:	Enhanced Hypercube
Primary Use:	MPI and I/O traffic

Submit Notes

The config file option 'submit' was used.

General Notes

130.socorro (base): "nullify_ptrs" src.alt was used.

 Software environment:
   export MPI_REQUEST_MAX=65536
   export MPI_TYPE_MAX=32768
   export MPI_BUFS_THRESHOLD=1
   export MPI_IB_RAILS=2
   export MPI_CONNECTIONS_THRESHOLD=0
   ulimit -s unlimited

 BIOS settings:
   AMI BIOS version 3.0
   Hyper-Threading Technology enabled (default)
   Intel Turbo Boost Technology enabled (default)
   Intel Turbo Boost Technology activated in the OS via
     /etc/init.d/acpid start
     /etc/init.d/powersaved start
     powersave -f

 Job Placement:
   Each MPI job was assigned to a topologically compact set
   of nodes, i.e. the minimal needed number of switches was
   used for each job: 2 switches for up to 180 ranks,
   4 switches for up to 320 ranks, 8 switches for 640 ranks,
   10 switches for 800 ranks, 16 switches for 1280 ranks,
   22 switches for 1920 ranks, and 30 switches for 2560 ranks.

 Additional notes regarding interconnect:
   The Infiniband network consists of two independent planes,
   with half the switches in the system allocated to each plane.
   I/O traffic is restricted to one plane, while MPI traffic can
   use both planes.

 Peak run:
   In the peak run, some benchmarks used different number of ranks
   from base. It is the only difference between base and peak.

Compiler Invocation

C benchmarks:

icc

C++ benchmarks:

126.lammps:

icpc

Fortran benchmarks:

ifort

Benchmarks using both Fortran and C:

icc ifort

Portability Flags

121.pop2:	-DSPEC_MPI_CASE_FLAG
127.wrf2:	-DSPEC_MPI_CASE_FLAG -DSPEC_MPI_LINUX
130.socorro:	-assume nostd_intent_in

Base Optimization Flags

C benchmarks:

-O3 -xAVX -no-prec-div

C++ benchmarks:

126.lammps:

-O3 -xAVX -no-prec-div -ansi-alias

Fortran benchmarks:

-O3 -xAVX -no-prec-div

Benchmarks using both Fortran and C:

-O3 -xAVX -no-prec-div

Peak Optimization Flags

C benchmarks:

104.milc:	basepeak = yes
122.tachyon:	basepeak = yes

C++ benchmarks:

126.lammps:

-O3 -xAVX -no-prec-div -ansi-alias

Fortran benchmarks:

107.leslie3d:	basepeak = yes
113.GemsFDTD:	-O3 -xAVX -no-prec-div
129.tera_tf:	basepeak = yes
137.lu:	basepeak = yes

Benchmarks using both Fortran and C:

115.fds4:	-O3 -xAVX -no-prec-div
121.pop2:	basepeak = yes
127.wrf2:	basepeak = yes
128.GAPgeofem:	basepeak = yes
130.socorro:	basepeak = yes
132.zeusmp2:	basepeak = yes

Other Flags

C benchmarks:

-lmpi

C++ benchmarks:

126.lammps:

-lmpi

Fortran benchmarks:

-lmpi

Benchmarks using both Fortran and C:

-lmpi

The flags file that was used to format this result can be browsed at
http://www.spec.org/mpi2007/flags/SGI_x86_64_Intel14_flags.html.

You can also download the XML flags source by saving the following link:
http://www.spec.org/mpi2007/flags/SGI_x86_64_Intel14_flags.xml.