Do you have FeatureCaching enabled? Try disable it before production runs.

How long does it take when you only read all features? How long does it take when you read and write without featurecaching? This to narrow down where the bottleneck is.

Are you implementing any restrictions, like a where clause or a spatial filter, in your reader? Is your db indexed correctly? Do you really need to also read geometry in your workspace if you only want to change an attribute name?

Similar comments:  All 3 of those components can have settings to make them slow.  It could be 1 or more combinations of things.

As @nielsgerrits highlights, make sure Feature Caching is turned off.

Then first diagnostic is to disable the FeatureWriter and SubstringExtractor (Ctrl-E toggles) and just run the workspace with the Reader only.  Then the second diagnostic is to run with only the FeatureWriter disabled.  Review the comparative performance and this should pointer to which workspace components are bottlenecking performance.

Note that string manipulation Transfers like SubstringExtractor, AttributeTrimmer, StringCaseChager etc.are all relatively expensive transformers in my experience and do cause noticeable bottlenecks in the workspace although not to the degree that you have indicated.

I would not think it alone would cause the bottleneck indicated, and instead believe the majority is more likely in the Writer or the Reader,  but using SQL Server function SUBSTRING() and getting the server to extract the substring rather than do it in the Workspace can yield some performance gains.