It’s one thing for blade server vendors to claim their blades have better power efficiency, and another thing entirely to prove it. Since the launch of our 12th generation blade servers, Dell has published SPECpower results on all our newest blade servers. Meanwhile, HP hasn’t published any SPECpower results for any model of Gen8 blade server; likewise, Cisco hasn’t published SPECpower results for any of their M3 blades.
To understand how the vendors stack up against each other for blade power efficiency in this generation, we assembled solutions based on similar blades from each vendor: the Dell PowerEdge M620, the HP ProLiant BL460c Gen8, and the Cisco UCS B200 M3. Each solution consisted of eight blades, that vendor’s blade enclosure, and a 10GbE internal IO module.
The solutions were tested using the industry standard SPECpower_ssj2008 benchmark, which produces an overall score for each setup based on its performance per watt ratio.
This graph shows the results of this comparison, which measured power for just the blades and internal chassis components. This comparison results in a decisive win for the Dell setup, with the Cisco solution scoring 16% lower and the HP solution scoring 24% lower.
We also compared the power efficiency of each blade solution with its vendor’s external 10GbE top-of-rack switch included in the power measurements. For the Dell solution, we used the Force10 S4810 switch; for the Cisco solution, the Nexus 5020; and for the HP solution, the ProCurve 5900 switch.
This graph shows the Dell solution again with a significant power efficiency advantage over the competition. Its overall SPECpower score is 26% higher than HP’s solution, and 45% higher than the Cisco solution with its Nexus 5020 switch and required Fabric Interconnect module.
I’ve just published a whitepaper with full details of the configurations and the test methodology here. I hope that when customers are making their purchasing decisions, this blade solution power efficiency comparison will help those looking to get the most performance out of every watt they buy to run their datacenter.