Colin,
  I have seen cases where the RAID controller cache has caused some
performance degredation.  This has to do with the interaction between the
checkpoint process and the controller cache.  The rate at which the
checkpoint process runs is regulated by the speed of writes to your disk
drives.  With a RAID controller cache these writes are very short, on the
order of 1-2 ms.  This causes the checkpoint process to immediately return
and write out more checkpoint data until the number of writes fills up the
controller cache, thus shutting it off.  This can degrade performance.  The
checkpoint process is regulated by the parameter MAX ASYNC IO.
I hope that this helps.
Ed
--
Edward Whalen
President
Performance Tuning Corporation

Phone: (281) 251-3495
Fax:   (281) 376-9987
Pager: (888) 912-3778

Web:   www.perftuning.com

Colin,

 This would usually only occur with a controller with a small amount of
cache, or one where the write-cache can use 100% of the entire cache.  
What controller are you using?

 Try running SQLHDTST.EXE with and without the cache on to see if any
errors are generated.

 Neil Pike MVP/MCSE
 Protech Computing Ltd