Replies to post #13868 on Intel Corp (INTC)

[chipguy]

SpecInt correlates well with majority of the commercial apps, server and client oriented, including the most commercial server app - database servers.

That is absolutely false. SPECint and OLTP/DBMS workloads
are *very* different in nature. Every architectural performance
analysis paper for x86, RISC, or IPC chips I have ever seen
shows this clearly.

have hard time imagining the power going up 3x. As far as cost (silicon cost), again, Dothan has somewhere around 3x advantage today.

Montecito has a 100W TDP with *two* CPUs. How much do
you figure a dual Dothan with 6.4 GB/s system bus and full
blown RAS capability would burn at 2.3 GHz?

[wbmw]

Joe, Re: SpecInt correlates well with majority of the commercial apps, server and client oriented, including the most commercial server app - database servers.

Yet in the case of SPECint, an HP integrity rx5670 with 1.5GHz Madison 6M scores up to 1312, while an HP ProLiant DL580 G2 with 3.0 GHz Xeon MP scores 1455.

In the case of TPC-C, however, that HP Xeon MP system scores 95k TpmC, while the Madison based system scores 121k TpmC using the same database and operating system.

This is the difference between a 10% deficit in SPECint and a 27% advantage in TPC-C. Moreover, if you compare higher end systems with up to 32 CPUs, the difference is even more distinct.

I think it's obvious from that data that SPECint is *not* a good database performance indicator.