In short, at the end of October IBM published a TPC-C benchmark on dual-core Xeons. It was the fastest four-way¹ TPC-C benchmark recorded. A couple of slight problems though, it came in at $8.27/tpmC (price/performance ratio), when competing benchmarks on Opterons were about $2.3/tpmC and even on Itaniums and POWER5 were more like $4/tpmC. Also the listed availability date for the hardware was 31st of March 2006.
At the start of December, HP published a benchmark on their dual-core Opteron DL585. It beat the IBM system, it was $2.02/tpmC, and there was immediate availability of hardware. The real kicker? Just like the IBM benchmark, HP used the IBM DB2 database system running on Windows.
¹ Counting lumps of silicon not independent processors. Multi-core CPUs have really confused the way people categorise multiprocessor benchmarks.