🚀 go-pugleaf

d
>chuckf@charm.net (Chuck Frain) writes:

>>I

>Rats! Are you sure about this? Way back I saw recommendations about 512k
>when using 16M+ with DX2 so I stuck to that when ordering my new Box
>as I'm starting with 16M but will go higher soon. According to you estimates
>I will have payed extra dough to get an extra sluggish PC, GREAT! Does
>anuone know wether it's possible to disable part of the cache to reduce
>its size to the optimum? Can anyone explain WHY a computer would run
>slower using a larger cache? Could it be that time is taken up updating
>the cache with bytes that will only be used once? Couldn't this mean that
>the performance would depend greatly upon the number of threads/processes
>that are running and using CPU time? In that case Win95 could possibly
>suffer less than Win3.1 from too large cache, and WinNT even less, at
>least when under a fair load. Just me guessing, please enlighten us
>if you REALLY know about computers and OS internals, ANYONE.

>Hans Lindberg, Phone: +46-31-183151 Mobile: +46-70-7813996
>EMail: d7hansl@dtek.chalmers.se

OK, when you have 256 K of cache you have 8 32K SRAM chips(8*32=256)
in 2 banks, with each bank having 4 chips...when you have 512K you
have 4 128K SRAM chips(4*128=512) in one bank...the access to the
motherboard is faster for the 256K configuration because the
motherboard can do bank interleaving...it can't do this with the 512K
configuration because all the SRAM is in one bank...the end result is
that 512K, which costs a lot more, offers only a slight increase in
performance