How I tested

IBM provided a late preproduction eServer 325, which was presented to me as identical to the production systems, with the exception of a smaller hard drive (40GB ATA) and some minor cosmetic differences. The system was preloaded with two versions of SuSE Linux 8.2, one 32-bit, the other, a 64-bit. The test server had dual 1.6GHz AMD Opteron processors, 2GB RAM, and the single ATA hard drive.

To test that the Opteron processor lets you run both 32-bit x86 software or 64-bit software written to use its AMD64 extended instruction set, I first booted the server using the 32-bit SuSE Linux, verified proper functioning of the operating system and hardware, and then repeated those tasks using the 64-bit SuSE Linux.

Using 64-bit Linux, I wrote and compiled several short C programs with GCC (GNU Compiler Collection) that verified the system was indeed running in 64-bit mode, and was using 64-bit pointers and the extended Opteron registers and instruction set.

The second phase of testing involved Microsoft's first beta of Windows XP 64-Bit Edition, build 1069. The beta installed directly from the bootable CD, and recognized all hardware in the server. With Windows XP installed, I ran several 32-bit and 64-bit applications, including a 64-bit version of Outlook Express, and a 32-bit version of SQL Server 2000. This early beta of Windows XP did not include the .Net Framework, or any .Net capabilities, but it did include IIS.

To test long-term hardware stability, I migrated a production Web site to IIS and ran it for six weeks on the server. The system never crashed or hung, even when an artificial load (generated using Rational's SiteLoad software) was placed on the Web site.