Unicode with MSSQL performance

I was wondering if anyone has any real-world experience of the performance difference in running unicode & non-unicode business data on SQL Server 2005. I understand there's a performance hit from JDE converting non-unicode data to unicode & vice versa, however everything I know about SQL Server performance (not that much, admittedly) is telling me that the increasing the page count on all of those string indexes is going to drastically increase logical IO and hammer performance.

I'm aware there's a whitepaper around showing 2.6% performance increase with unicode on iSeries, however I'd expect the iSeries probably has better disk IO than our SAN. Has anyone done any research on unicode performance on a non-iSeries platform?

Sam,

We had that conversation with the Oracle consultant that did the conversion on our test box. Our test box is MS Wondows 2003, MS SQL 2005 and 8.12. We asked about the pros and cons of converting the business data over to unicode. Since our test box is not using any of the double byte languages, we were advised to not bother with the conversion. The performance hit is too small to notice.

Gregg

From what I've been told in the past there is a huge performance impact if running 8.9 and up with a non unicode database because JDE always processes its data in unicode format so it has to constantly convert the data it reads from non-unicode to unicode at runtime. I've only worked with a unicode database with those releases so I can't say from experience how dramatic the difference is.

I attached a slide that gives some bullet points about this topic.