home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.arch
- Path: sparky!uunet!caen!nic.umass.edu!usenet
- From: HAGAN@deimos.ucc.umass.edu (Craig I Hagan)
- Subject: Re: CACHE MISS PENALTY FOR 386/486??
- In-Reply-To: eletanjm@nuscc.nus.sg's message of Wed, 19 Aug 1992 04:23:18 GMT
- Message-ID: <1992Aug19.200257.24361@nic.umass.edu>
- Sender: usenet@nic.umass.edu (USENET News System)
- Nntp-Posting-Host: deimos.ucc.umass.edu
- Organization: University of Massachusetts at Amherst
- References: <1992Aug19.042318.10163@nuscc.nus.sg>
- Date: Wed, 19 Aug 1992 20:02:57 GMT
- X-News-Reader: VMS NEWS 1.20
- Lines: 43
-
- In <1992Aug19.042318.10163@nuscc.nus.sg> eletanjm@nuscc.nus.sg writes:
-
- > Just what is a typical cache miss penalty on
- > (1) a 386 system with external cache
- > (2) a 486 system with only internal cache
- > (3) a 486 system with internal cache and external secondary cache
- >
- > I'm assuming an typical AT style system.
- >
- > Just how do you determine the miss penalty anyway?
-
- Well, i am going to attempt to not make a fool of myself.
-
- My guess would be that you would figure out the miss penalty (maximum
- penalty would probably be the best/easiest to computer) by taking the
- time to determine that there has been a cache miss, and add the
- memory access time of the main memory to that. If you had two
- caches (for example, an i486 with both an internal, and an external
- cache, then you would have:
-
- time to determine internal cache miss + time to determine external cache miss
- + time to retrieve the actual data from memory
-
- presumable the first two would be the shortest of the bunch as the
- the CAM in the cache controller can process a hit/miss fairly
- quickly (this would depend on the controller, naturally - unfortunately
- i am not sure of the current speeds of the controllers).
-
- the last, memory access would be dependant upon the number of wait
- states in the memory, as well as the physical speed and layout
- of the memory (probably worst case would be between the speed
- of the memory [e.g. 60ns, 70ns, 80ns, etc] to something like 10-20ns
- above that, but probably very close to the speed of the RAM itself.
-
- A guess on my part would say that a cache miss wouldn't be much
- worse than a normal memory access on a similar machine without
- the cache.
-
- ----------
-
- Craig Hagan
- University Computing Center
- UMASS/Amherst
-