RAM question

[Mike Rambo]

Edward Glowacki wrote:
> 
> Quoted from Jo Dillon on Fri, Nov 16, 2001 at 03:31:50PM +0000:
> > On Thursday 15 November 2001 21:50, Benjamin Minshall wrote:
> > > Hi Mark,
> > >
> > > This could be an issue related to "high-density" or "low-density" 256MB
> > > DIMMs.  I don't know a lot about it, but I ran into some machines that were
> > > only rated for 128MB chips, yet could use low-density 256MB DIMMs.  Perhaps
> > > someone else could (in)validate my speculation?
> >
> >   Another possibility is it only has enough cache for 512MB worth of memory.
> > In that case it'd boot and recognise the memory but any access that happened
> > to hit the high 512MB of memory would be extremely slow. So the system would
> > appear (possibly intermittently) about as fast as a 386 ;)
> 
> I don't think this is correct, it doesn't make much sense.  Cache
> and main memory are independent, and caching algorithms generally
> work fine for any combination of the two.  You'll get better
> performance with bigger cache, but you won't get worse performance
> with bigger memory.
> 

Usually true. There are some exceptions to the rule though (like usual).
I have seen machines where performance does go down with additional ram
because, for whatever reason, it is not able to be cached. I've never
looked into why so I don't know. Just a couple of months back we had a
box in for some upgrades the crucial.com warned right on their web page
to not upgrade memory beyond a certain point (really low too, I think
128MB in this case) because even thought the motherboard would support
it there would be performance degredation because of (insufficient I
assume) cache.


-- 
Mike Rambo
mrambo@lsd.k12.mi.us