Benchmarking mobile workstations

For today's review of the Dell Precision M6400 and HP EliteBook 8730w mobile workstations (see "Road warrior power trip"), I tested performance using three diverse sets of workloads. In each case, testing was conducted under Windows Vista Business (x64 Edition) with Service Pack 1, using the latest driver and utility stacks from Dell and HP.

First, to see how quad cores affect complex, multi-process workloads, I used the DMS Clarity Studio tool to configure a combination database/workflow scenario. Second, because virtualization has become such a big part of the modern application development cycle, I timed how fast each system to clone and snapshot large virtual machine images under VMware Workstation 6.5.1. Finally, I also ran the prerequisite SPECviewperf 10 (64-bit version), which measures OpenGL graphics rendering performance.

The Clarity Studio Database workload consists of copying several SQL Server 2008 tables from a source database to a target database and then running various SELECT statements against the target database. The tool uses ADO to access the databases and also employs MTS to encapsulate the database operations as distinct transactions.

The Clarity Studio Workflow workload consists of copying roughly 25MB of mixed message and attachment data from a single source folder to multiple target folders, using the MAPI/CDO libraries to accomplish each transaction. Both workloads execute in a continuous loop, with a 1 second delay between iterations. The workstations test scenario involved running 10 concurrent instances of each workload, for a total of 20 discrete workload tasks all executing in parallel.

All test results are presented in the table below. Figures for the VMware and Clarity Studio tests are the number of seconds required to complete the task or script (smaller is better). Results for the SPECviewperf 10 tests are based on the number of frames per second rendered during the tests (larger is better).

In every test, the HP ran behind the Dell. The differences in the SPEC results were most likely a product of the Dell unit's faster DDR-3 memory (1066MHz versus 800MHz DDR-2 in the HP) and perhaps the slightly newer certified nVidia driver stack (version 176.53 versus 176.06 for the 8730w). Both systems incorporated the same nVidia Quadro FX 3700M video card with a full gigabyte of video RAM.

The performance differentials in the VMware Workstation and Clarity Studio tests can be traced directly back to the RAID 0 configuration in the Dell, which gives the unit a tremendous advantage during disk write operations. You could achieve similar results with the HP by configuring it with a two-disk RAID. In other words, the big winner of these disk-intensive tests is RAID 0.