Based on IBM actual, public results on HS22 & Intel internal analysis. 1U rack server configuration: 2x single-core Xeon (3.8GHz 2MB cache) with 8x 1GB memory and 1 HDD – total power: 382W under load. HS22 blade server configuration: 2x quad-core Xeon X5570 (2.93GHz 8MB cache) with 6x 2GB memory and 1 SSD – total power with chassis burden = 307W under load. Performance based on IBM internal projections comparing SPECint_rate2006 of 1U rack server to HS22, which yields approximately 11.17x better performance on HS22. For HS22, SPECint_rate2006 = 252 (published April, 2009 on SPEC.org). For 1U rack, estimated SPECint_rate2006 = 22.56. 252 / 22.56 ~ 11.17. Internal estimate on 1U rack to HS22 ratio made in March 2009. Source: www.spec.org. SPEC and SPECint are trademarks of the Standard Performance Evaluation Corporation (SPEC).
Power estimate assumes HS20 approximately equivalent to HS22 and HS22 is approximately 32% lower power per server versus a similarly configured rack-optimized server. Power cost estimate used is US$0.10 kW-Hr. Datacenter floor space cost estimate used is $500 per square foot, average 8 square feet per standard enterprise 42U rack. IT footprint compares 1U per rack server vs. 0.5U per HS22 blade–14 HS22 blades in 7U BladeCenter E chassis provides equal or greater performance than 137 1U rack servers.
Up to 65% less in connectivity costs based on IBM vs. HP web pricing.
Up to 84% fewer cables based on calculations of 1 rack of (6) fully populated BladeCenter E chassis vs. 1 fully populated rack (42) of competitive 1U servers.