"Enable write caching on the disk"

" Enable advanced performance (recommended only for disks with a backup power supply."

  The raid adapter takes controls of both disk cache/advanced performance, not the OS....  turned off as this interferes with raid safety/performance.

As to the other setting, the default values are generally best unless you have a specialized/dedicated server. You can play with WB, WT, read  and I/O policies with out doing damage.

For a general use server the default is good, if you bench mark with the different read policies you will see minuscule differences. If you have a server dedicated to a specific program or to a specific task such as a group of users retrieving similar data, then the read policies may help. Daily defragging will give you 4-5% over an array which is defragged once a week or more. 

I do not see that much difference, except in the bust rate which means very little; your average speed and CPU usage are more important. You will always get variations (like in the IO test), but the read policy changes shows trade offs..what goes up or down in certain data size areas will be offset by changes in other data size areas; what do you think happens when the cache is filled with read ahead info.. something needs to be forced out of the cache. Secondly, most benchmarks highly stress arrays,  flooding the cache..unless you have a couple hundred users hammering a server, your arrays will not blaze away or over stress the cache like benchmarks cause. If an array's disk LEDs are on as much as most benchmarks cause, it is time for an added server ( and a few extra cooling fans or "The Idiot's Guide to Cooking On a Hard Drive"). 

To get any meaningful benchmarks, you need to run multiple benchmarks, HD Tune/Tach and ATTO are not greats benchmarks, especially ATTO ( though good for cache speed comparisons). IOMeter, stop watch timed disk copies of averaged sized files within arrays (and over the network), SQLIO, and COSBI File Copy Benchmark, v0.52 are more realistic.

Check this out, for Win 2008, but unlike other previous MS tunning guides, it is accurate

http://www.microsoft.com/whdc/system/sysperf/Perf_tun_srv.mspx