Ed,

>>2. The best hardware solution is to have two hard drives on two controllers. Large table copies, joins, etc. should always read from one physical drive and write to the other. Can cut real-world times by 50% or more.
>>
>
>Absolutely not a correct statement. A RAID solution is going to outperform this by a wide margin.

I have to assume you are correct. A hardware-based RAID solution is not in my budget, but adding a mere second hard drive is. I ran an overnight test having Foxpro repeatedly copy a 370MB between all combinations of four partitions on three physical units on two controllers. The most dramatic performance was copying between two separate physical drives on two separate controllers. The slowest result was copying from any partition to itself. Within my budget, this is conclusive.

But let's continue...

>
>>3. If there are a lot of linear reads and writes, Win 98 with FAT32 is actually faster than WinNT. (For that matter Fox/DOS on OS/2 is faster than any possible windows platform, but this is not practical in most cases).
>>
>
>Absolutely, positively, beyond any shadow of a doubt wrong, especially with a dual processor configuration. NT with NTFS and stripe sets (or better, hardware RAID) and correct tuning of NT is going to leave the biggest, fastest baddest Win9x box bleeding an inch or two from the starting line.

Can you please define "properly tuned?" What I'm wondering is this: If I have two SCSI drives on a single controller (Adaptec 2940), currently partitioned completely separately as F and G, will I see a measurable performance gain if I repartition them as one unit, courtesy WinNT? This is my test box and I would be happy to do the repartition and run the tests if it yields a demonstrable result to management.