To evaluate the Doubleshot's performance, the Windows NT Magazine Lab used the BAPCo benchmarking suite introduced in the January issue. These benchmarks use applications, such as Microsoft Word and Excel, to provide a real-world evaluation of system performance. To find out more about these benchmarks, visit BAPCo's Web site at http://www.bapco.com.
The baseline machine for comparison is a 90-MHz Pentium Telos system with 32MB of RAM, a 1GB SCSI-2 internal hard disk (with a PCI Adaptec 2940 controller), a 2X CD-ROM drive, a Diamond Stealth 64 PCI video card, and a 3Com Etherlink III ISA network card. It has a street price of $2800.
BAPCo benchmarks measure performance using SYSmarks and runtime. The SYSmark rating is a composite score based on popular applications which are run to simulate the activities of actual users. SYSmark ratings are linear, so a system with a SYSmark rating of 200 is twice as fast as a system with a rating of 100. The Doubleshot's ratings, with and without one of the CPUs disabled--as well as those for the baseline system--are shown in the SYS-mark Results graph (Graph A). The Runtime graph (Graph B) shows how long each system took to complete the tests (three runs per application).
As you can see, the difference between running one and two processors seems hardly noticeable. The applications used in the BAPCo benchmarks aren't explicitly designed to take advantage of SMP; thus the addition of a second CPU didn't yield overwhelming performance gains. (Further details will be covered in the review of Photoshop 3.0.5 in the March issue of Windows NT Magazine.)
This version of the Doubleshot only comes with two processors (a different board is available for a single CPU system), so I can't offer a cost/benefit analysis of one CPU vs. two. The one vs. two tests I conducted were arbitrary. They were only meant to show the benefits--or lack thereof--of an SMP workstation.