In article <[log in to unmask]>,
[log in to unmask] says...
>
>Mike from SMS did have a spread sheet that he had generated from
another
>customer that did measure reponse time changes in a before and after
>situation.
 
I can vouch for a similar situation - I may be the shop that Mike
refers to, because I wanted to find the knee in the performance curve.
Here are the details:
        Before     After
917LX   24MByte    56MByte
 
Single user system, mostly software development.
 
 
Same situation, 4.0 performance was find, but 5.0 required large
quantities of air freshener. Adding 32MBytes  dropped compile times
SIGNIFICANTLY! - back to 4.0 times. Going from 56 MBytes to 88MBytes
had no increase whatever for performance, so I stayed with 56.
I can endorse the <PLUG>, I got a good price out of Mike, with the same
offer to remove the memory if it didn't meet my needs.
:-)  :-)