Thermal performance of Office Server Cabinets
The effect to server operating conditions, provided that the thermal loading and installation guidelines are adhered to, is so slight as to be insignificant.
To understand how effective the PSEs thermal management system is, please consider the following example of a Office Server Cabinet Pro RM v7.6 model PSE18 enclosure’s impact to server CPU temperatures, under normal office environmental conditions, compared to operation of those same servers in free space.
- Cabinet: Model PSE18 v7.6, maximum recommended thermal load 1.2 kW
- Server 1: HP Proliant DL380 G4 2U rack-mount servers with 2 x 3.2 GHz Intel Xeon processors
- Server 2: HP Proliant DL380 G4 2U rack-mount servers with 2 x 3.2 GHz Intel Xeon processors
- Server 3: HP Proliant ML370 G4 5U rack-converted server with 2 x 3.2 Intel Xeon GHz processors
Incidental hardware also present in Office Server Cabinet Pro during example test:
- APC Smart-UPS 3000 VA 3U rack-mount UPS
- 3 x Netgear network switches
- 2 x 1U climate monitoring devices
Test and measurement method
CPU temperature measurements were taken using HP Systems Insight Manager software
Continuous CPU loads were generated using BurnInTest software by Passmark
Room ambient temperature measurements were taken by recording the average reading of 2 x digital thermometers Room ambient air temperature
The test room air temperature was maintained at a constant 24°C / 75°F (+/- 0.5°C) throughout the test period (equivalent to a moderately warm office)
Procedure used in example tests:
1. In the first instance the entire outer shell of the ComputerVault Pro was removed, and the CPU temperatures were recorded at idle (running but not processing client tasks) in free air space. Removing the ComputerVault Pro’s casing created thermal conditions identical to those in a conventional open-frame rack, or with servers not rack-mounted but resting on surfaces.
2. Identical and continuous processing loads, generated by the test software, were applied to all six CPUs simultaneously, such that utilisation in all CPUs was increased in steps from idle to 20%, 40%, 60%, 80% and 100%. After each step change in CPU loading, CPU temperatures were allowed to stabilize for one hour, and then the operating temperature of each of the six CPUs was measured/recorded using the server manufacturer’s own software.
3. The outer shell of the ComputerVault Pro was then fully reinstated, and the above test procedure was replicated.
The following results were recorded:
The average increase in CPU temperature attributable to the ComputerVault Pro, versus free space operation, was +3.2° Celsius
The electrical consumption of the combined systems, with all CPUs at 100% utilisation continuously, was 1.076 kW.