On 26 Feb 98 at 22:58, Rick Poepping wrote:

> I'm getting ready to do various upgrades to my 586-p133

  This is not a very clear designation; no CPU manufacturer really
uses "586" as a designation.  [Some motherboard manufacturers use
this to identify motherboards designed for the Pentium (Socket 5 or
Socket 7).

  My first thought was that maybe you have a 5x86-133, which is an
AMD version of the 486 with a 4x clock multiplier.  Many 486
motherboards supported 256K of cache, and some may have supported
512K.

  It's more likely that you have a 133MHz Pentium, often abbreviated
as "P133".


> and I would like to know the relation of Cache and RAM.

  Cache is fast SRAM instead of slower main RAM.  At any given
moment, some of the locations in main RAM are duplicated in cache
RAM, and so may be referred to quickly if needed again soon.

> Specifically, how much RAM can 512K cache?

  That's up to the motherboard designer, subject to limitations of
the chipset they've chosen to use.
  Many popular Intel chipsets are limited to caching values in the
first 64MB of RAM installed.  Some are not.
  On some machines that are not so limited, a 256KB cache will handle
only values in the first 64MB, and a 512KB expanded cache will have
the larger span (typically 512MB).
  I happen to have a machine where this is ordinarily true, but of
which units equipped with 512KB at the factory do not cache beyond
64MB without the addition of a tag RAM chip.  It's theoretically
possible that there are other scenarios where 512KB of cache still
only spans 64MB of main RAM -- including boards using chipsets that
are limited to that.

> And what benefits and pitfalls are there to going beyond that
> amount (of RAM).

  A chipset limited to caching 64MB of main RAM faces a problem when
more is installed.  One simple way out of the jam is to disable the
L2 cache entirely....

> I've heard all kind of things -like: Win95 uses the highest memory
> addresses first, so if you have more RAM than you can cache, you'll
> always be operating out of the cached area (to begin with, at
> least)  and so on.

  The address space that Win95 is allocating out of is one step
removed from the physical addresses dealt with by the cache hardware.
It's often asserted that the OS will wind up not being cached
*because of its high addresses*, but I am not yet convinced that this
actually happens.

> Is this exactly the same for laptops?

  There are a few chipsets that are specifically designed for
laptops; I don't know exactly what they do in this regard, but it's
probably similar.  [Odds are very good that support for caching more
than 64MB of RAM is not regarded as a necessary feature by developers
of laptop chipsets who are usually looking for wastes of space and
power to eliminate....]

David G