Advanced Micro Devices Inc (AMD)

[Dan3]

Re: So how does all this Opteron fabulousness result in a herculean 193 sq. mm. die size, 2 extra metal layers, 8 extra masking layers, an extremely expensive SOI process that results in 18% yields and sub 2 GHz clcok speeds and produce "improving performance and lowering costs,"

18% yields? You must be thinking of Itanium.

Check out the performance charts that are linked to from this review. 1.8ghz Opteron, with integrated chipset, is roughly twice as powerful in server applications as 2.8ghz Xeon:
http://www.pcmag.com/article2/0,4149,1061852,00.asp

We used our 32-bit Nile application server test (which simulates ordering transactions for visitors to a fictitious Web bookstore) to pit the two servers against one another. The test is both CPU- and disk-intensive. The Nile test uses Oracle 9 i as the back end, and it has BEA WebLogic Server 7.0.2 application server software running on the test equipment. Each tested server was running Windows 2000 with SP3.

As our scatter charts show, the dual Opteron system outperformed the dual Xeon server by a fairly wide margin. For example, across a 300- to 500-user load (where transaction processing was stabilized with high disk and CPU utilization) on the Pages Received test, the Xeon machine averaged 7.6 pages per second, and the Opteron entry averaged 15.2 per second. On the Search Response Times test, the two servers were fairly close with a load of 200 users. But as the number of virtual users increased, the Xeon server needed increasingly more time than the Opteron model to service the requests.