In case you missed the news, Intel released a bunch of world record benchmarks for 2 socket x86 servers, and big ups to UCS for its three impressive world records—more than any other x86 vendor. Sure, benchmarks and statistics share damn lies together, and as a marketing communications professional, there is nothing that induces sleep faster than benchmark debate. Just the same, please allow me to expound on this particular set of world records because I do think there are some interesting comments to be made:
- Cisco Unified Computing System B250 M2 Server set a new record(1) with results of 35.83 @ 26 tiles in the VMware VMmark benchmark. While Cisco blades with extended memory and virtualized NIC controller are expected to win in virtualized environments, the VMmark benchmark is a significant win because it measure both virtualized system density and performance, and the UCS score represents a 43% improvement over the previous winner.
- Cisco Unified Computing System C250 M2 Server set a new record on the SPECjAppServer2004 benchmark, which measured application server performance, running a complete Oracle Enterprise Linux, Oracle 11g and Oracle WebLogic stack. While its no secret that UCS should do well in the app server category, the workload focus is still a nifty win for UCS—an opinion that Oracle clearly shares.
- Cisco also achieved top results(2) in several key High Performance Computing benchmarks including SPECfp_rate_base2006, SPECompM2001, SPECompL2001, and Linpack, as well as the compute intensive integer performance benchmark, SPECint_rate_base2006. Its nice to see UCS do well in computationally intensive areas.
Here, lifted from the Intel website, are all of the new world records:
| Highlights on Intel® Xeon® processor 5600 series-based server (March 16, 2010) | ||||
|---|---|---|---|---|
| Segment -- Benchmark | Platform/Partner | Software Details | Result (gain over Intel® Xeon® processor X5570) |
Importance |
| General Purpose Computing -- SPECint*_ rate_base2006 |
Dell PowerEdge R710* | Intel® C++ Compiler 11.1 -- SLES 11 | 355 (+40%) | Number one x86 two socket server result |
| Scientific Computing -- SPECfp*_rate_ base2006 |
Dell PowerEdge R710* | Intel C++ Compiler 11.1 -- SLES 11 | 248 (+25%) | Number one x86 two socket server result |
| Virtualization -- VMmark* |
Cisco UCS B250M2* | VMware ESX 4.0 U1* | 35.83@26 tiles (+42%) | Number one two socket result |
| Energy Efficiency -- SPECpower*_ ssj2008 (Single Node Server) |
IBM x3650 M3* | IBM* J9 JVM | 2,927 overall ssj_ops/watt (+42%) (gain over Intel® Xeon® processor L5530) | Number one two socket single node server result |
| Energy Efficiency -- SPECpower*_ssj2008 (Multi-Node Server) | IBM dx360 M3* | IBM* J9 JVM | 3,038 ssj_ops/watt (+31%) | Number one two socket Multi-node server result |
| eCommerce -- SPECjbb*2005 | Fujitsu PRIMERGY RX300 S6* | IBM* J9 VM* -- Microsoft Windows Server 2008* | 928,393 BOPS (+46%) | Number one x86 two socket server result |
| Application Server -- SPECjAppServer* 2004 |
Cisco UCS C250 M2* | Oracle WebLogic Server* SE Rel. 10.3 -- Oracle 11g Database*, Oracle Enterprise Linux* | 5,185 JOPS(+30%) | World record for the most JOPS (Java Operations Per Second) per Chip |
| Web Server -- SPECWeb*2005 | Fujitsu PRIMERGY RX300 S6* | Rock Web Server* v1.4.7 -- Red Hat* Enterprise Linux* 5.3 | 104,422(+25%) | Highest SPECWeb2005 result ever |
| Enterprise Resource Planning -- SAP-SD* 2-Tier |
Fujitsu PRIMERGY RX300 S6 | SAP ERP 6.0* (Unicode) | 4,860 Number of Benchmark Users | Number one two socket result |
| HPC -- SPEComp*Mbase 2001 |
Cisco UCS B200 M2* | RHEL 5.4*, Intel C++ Compiler 11.1 | 52,314(+20%) | Number one two socket result |
| HPC -- SPECMPI* 2007 |
Intel Result server cluster | Suse Linux Enterprise Server* | Excellent cluster scaling | Number one two socket cluster result |
| Integer Single Task -- SPECint*_ base2006 |
Fujitsu PRIMERGY* RX300 S6 | Intel C++ Compiler 11.1, SLES 11 | 38.9 (+10%) | Number one x86 two socket server result |
| Energy-Efficiency-Performance/watt using SPECint*_rate _base2006 |
Fujitsu PRIMERGY* RX300 S6 | Intel C++ Compiler 11.1, SLES 11 | Intel® Xeon® processor L5640 uses 30% lower power than Intel® Xeon® processor X5570 at SPECint*_rate_base2006 score of 250 | Same performance as Intel® Xeon® X5570 with 30% lower power results for
RX3000 S5
File Type/Size: ZIP 97.3KB results for RX3000 S6
File Type/Size: ZIP 101KB |
| HPC -- Black-Scholes* (Financial Services) |
Supermicro* | Black-Scholes | 11.5 seconds (+60%) | Exceptional gain over previous generation |
You can find all of the world records <a href=”http://www.intel.com/performance/server/xeon/summary.htm”>here</a>.
Comments Are Closed