LOL, plenty of BS in this thread to keep us all occupied.
I think some of you guys would be surprised how much effort goes into testing ALL software, especially games, since it is not so simple to upgrade them, and manufacturers only really get one shot at producing it before their reputation is set in stone! Most software spends more time in the test lab than in development!
More memory would be very useful for those people who are trying to port Windows or Linux, as well as future emulation needs. Bear in mind that Mame is still evolving, and they are currently running into memory problems, so how do you think it will cope in a few more iterations?
If a CPU upgrade was offered, it would be a simple task to set it to run at the slower clockspeed, so that older games were unaffected, so optimisation doesn't make any difference.
And who said upgrade to a Celeron?

ROTFLMAO
Do you think clockspeed is the only factor in determining the actual speed of running code?
Although the CPU is running at 733MHz, it has a different architecture to any of the standard Pentium chips, although it is similar to the PIII. The amount of level 1 and 2 cache will be a greater factor in overall speed, as graphics processing tends to be memory speed dependant. The PIII had an architecture that made it very fast in multitasking environments, using predictive computing, although it wouldn't be much faster than a PII in a linear test.
The Celeron has always been the donkey of processors. It is slow, has poor cache performance, and limited multitasking capabilities. However, it is cheap, and because of its simplicity, it is easy enough for Intel to keep booting the clockspeed, with relatively little redesign.