256K vs 512K cache
Garry Wiegand
256K to 512K pipeline burst cache. My machine is a P133, 60ns
EDO RAM, etc etc etc. I ran several compile-speed timing tests
just before making the change, and again just after. I chose
compiling because it's my most common CPU-intensive activity.
I'm running Visual C++; the code is template-intensive and is
almost entirely CPU (rather than disk) bound.
Net results: zilch. Within the limits of accuracy of my
measurements (maybe 2%), the extra cache memory makes no
detectable difference.
Yes, I believe the jumpers are set correctly; the bios does
report "512K" of cache at the appropriate moment during the
boot.
Oh well. If you've ever been tempted to get the extra cache...
better off buying bagels.
garry
Garry Wiegand
difference it made.
My machine is a P133, 60ns EDO RAM, etc etc etc. I ran of
compile-speed timing tests just before making the change, and
again just after. I chose compiling because it's my most common
CPU-intensive activity. I'm running Visual C++; the code is
template-intensive and is almost entirely CPU-bound (rather than
disk-bound).
Net results: zilch. Within the limits of accuracy of my
measurements (maybe 2%), the extra cache memory makes no
detectable difference in processor speed.
Yes, the jumpers are set correctly; the bios does report "512K"
Garry Wiegand
> Garry Wiegand <ga...@a.crl.com> wrote:
> >Hi all. I have a bit of info to share - I just upgraded from
> >EDO RAM, etc etc etc. I ran several compile-speed timing tests
> >...
> This ties in with other reports I've read. The speed increase between
> 256 and 512 is not noticible unless you are running with a LARGE
> amount of RAM (32 megs or higher).
Yes, even so: my machine *does* have 32 meg. (The memory wasn't saturated
though in the tests that I ran.)
garry
Fuchi
>garry
Hi Garry,
I heard that Cyrix6x86 could make good usage of cache over 256K even
the RAM is less than 32MB. If you have the chance to make the test
with Cyrix6x86, please post your observations. Thanks!
Greg Kemnitz
>Hi all. I have a bit of info to share - I just upgraded from
>difference it made.
>
>My machine is a P133, 60ns EDO RAM, etc etc etc. I ran of
>compile-speed timing tests just before making the change, and
>again just after. I chose compiling because it's my most common
>CPU-intensive activity. I'm running Visual C++; the code is
>template-intensive and is almost entirely CPU-bound (rather than
>disk-bound).
>
>Net results: zilch. Within the limits of accuracy of my
>measurements (maybe 2%), the extra cache memory makes no
>detectable difference in processor speed.
This actually makes sense - caching algorithms don't need all
that much cache to achieve high levels of cache hit-rates. Beyond a
certain (rather low) amount of cache memory, a doubling of size will
yield a very small increase in cache hit improvement.
If you have an utterly huge amount of memory (ie, 128MB), 512K cache can make
a measurable difference, especially if you have a multi-user system.
(I assume you are using pipeline cache - the difference between pipeline
and other cache *is* significant)
>Yes, the jumpers are set correctly; the bios does report "512K"
>of cache at the appropriate moment during the boot.
>
>Oh well. If you've ever been tempted to get the extra cache...
>better off buying bagels.
>
>garry
------------------------------------+----------------------------------
Greg Kemnitz | High-quality CD-ROM's and other
MultiSys, Inc, La Verne, CA | computer components at excellent prices.
Tel: (800) 316-8570 | Check out appropriate forsale groups or
Fax: (909) 392-3696 | send e-mail for a price list.
kem...@netcom.com | VISA/MC/Amex/Discover No Surcharge