Can you have too much ram?
Firstly, what type of RAM, as if it's the more expensive RDRAM, shelling out for that much may be an uncomfortable option.
I don't know too much about Intel (I'm, a AMD follower), but with "533FSB", you need a chipset that supports quad-pumped operation, using two banks of DDR266 (PC2100) interleaved, not sure if they ever got to using double banked DDR400 (PC3200) for 800FSB, or if that's only available on RAMBUS.
Aha -
Tom's Hardware looks at RAM
And note this page
http://www20.tomshardware.com/motherboard/20040119/index-04.html
Faster timings with 4x 256 modules than with 2x 512
NB. This is a poor config choice if you may want to upgrade above 1Gb RAM
Mind you, the conclusion does not come down strongly on the side of memory performance - looks like the results advocate a faster CPU and good name, but not superfast RAM - paying much extra for the fastest memory would seem not to be worth it.
More RAM, WILL take longer if you let the BIOS count and test every time, and I suppose the chance of a random soft error in memory is proportional to the amount of RAM actually in use - not so much the AMOUNT of memory though, on a system running constantly, I'd say fit ECC RAM if supported, and the price difference is not to great - think of ECC as insurance against undetected memory errors, as it's easy to prove memory is solidly BAD, but even many, MANY hours of ramtest running cannot prove that it will never error - ECC should correct any single bit (per location) error, and should detect most others - the pedantic may say that it COULD miss, but it's much harder to fool ECC than the old parity methods.
I have seen a system (not a PC) using ECC memory, continue running with a systematic 1 bit failure across an entire bank of RAM (the RAM card was then replaced as a matter of urgency) - I've also seen the same system throw a random memory error (no problem) and the memory didn't give any further problem.
The only downside to ECC on many DIMMS, there may be too many bits coming from each chip to survive a chip failure, thought ECC is not intended as an antidot to a memory module "going down" - though hopefully it should halt in a controlled manner, rather than making random screwups until it crashes (though a really gross error would crash it pretty quiclkly).
