home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!mcsun!uknet!cam-cl!doc.ic.ac.uk!doc.ic.ac.uk!not-for-mail
- From: ntc@doc.ic.ac.uk (N T Cheung)
- Newsgroups: comp.sys.mac.hardware
- Subject: Re: Mac IIsi, RAM and cache
- Message-ID: <185cgoINN16i@oak23.doc.ic.ac.uk>
- Date: 3 Sep 92 15:55:04 GMT
- References: <1992Sep2.135813.16797@doug.cae.wisc.edu>
- Organization: Department of Computing, Imperial College, University of London, UK.
- Lines: 41
- NNTP-Posting-Host: oak23.doc.ic.ac.uk
-
- In article <1992Sep2.135813.16797@doug.cae.wisc.edu> jimy@neuro.ece.wisc.edu (Jim K. H. Yu) writes:
- >I use a Mac IIsi with 17MB RAM and System 7.
- >That is plenty of RAM for my application (at least for now) :)
- ...
- >
- >To begin with, I could increase the cache size to, say, 2MB.
- >That would definetely solve the bottle-neck created by the built-in video.
- >However, too big a cache may not improve the hit ratio and can
- >even slow down the Mac (cache management overhead increases with
- >cache size).
-
- When I did some cache size testing on a IIci using MPW builds as a benchmark, I
- found that the optimum cache size was 96-128K. Any larger and performance is
- decreased. Thus I am very interested to note that people are now advocating
- much larger cache sizes. Perhaps it is something to do with System 7 (my testing
- was in the bad old 6.07 days). Does anyone have any figures they could post?
-
- >The question is: Does anyone know what kind of cache the Memory
- >control panel creates?
- >In other words, If it's a *data cache* for the HD, then it's existence is still
- >necessary, even with a RAM disk.
- >However, if it's an *instruction cache*, then is it correct to assume that
- >running the whole system from a RAM disk is equivalent to 100% hit ratio,
- >and therefore, I can turn the cache to the minimum size and save 2MB for other use?
- >
- I would guess it's a data cache for the HD. After all, it was the 68040's
- instruction cache that caused some not-very-well-written apps to break.
-
- Besides, you wouldn't get a 100% hit rate for an instruction cache just because
- the program you are running is memory-resident. The cache will still miss as
- it overflows and blocks are replaced.
-
- >
- >Jim
- >
- >
- >=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=--=-=-=-=-=-=-=-
- >Jim K. H. Yu
- >jimy@eckert.ece.wisc.edu
-
- Tseung
-