home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.os.linux
- Path: sparky!uunet!munnari.oz.au!hp9000.csc.cuhk.hk!a080700
- From: a080700@hp9000.csc.cuhk.hk (Stephen Wong S M)
- Subject: Re: Performance Enhanced or Worsened by Cache Addition? On Zeos 486DX2-66 system?
- Message-ID: <1993Jan7.113057.15502@hp9000.csc.cuhk.hk>
- Organization: The Chinese University of Hong Kong
- References: <FRIEDMAN.93Jan7000112@khan.trincoll.edu>
- Date: Thu, 7 Jan 1993 11:30:57 GMT
- Lines: 53
-
- In article <FRIEDMAN.93Jan7000112@khan.trincoll.edu> mark.friedman@mail.trincoll.edu writes:
- >On some PC-Clone systems, when running some versions of Unix, users
- >have found that performance actually increases after disabling the
- >hardware cache perhaps because the cacheing algorithm in the kernel
- >was better tuned to Unix access than the hardware cacheing algorithms.
- >
- >Can anyone argue through an understanding of Linux's kernel and
- >hardware cacheing, or through experience, or best yet actual
- >measurements whether the performance of Linux with X, and say GNU
- >programs such as emacs and gcc increases or decreases with the
- >addition of a hardware cache?
- >
- >In particular, I'd be interested in knowing if any Zeos 486DX2-66
- >system users have measured/experienced their systems running Linux
- >with and without the 128K and 256K hardware caches and experie nced a
- >significant difference in performance in either direction.
- >
- >I am trying to decide whether to purchase a 128K cache upgrade to go
- >with a system with 16M RAM.
- >
- >Thanks,
- >
- >-- Mark
- >--
- >
- >
- >Mark A. Friedman mark.friedman@mail.trincoll.edu
- >Engineering and Computer Science or friedman@starbase.trincoll.edu
- >Trinity College Phone: (203) 297-2519
- >Hartford, Connecticut 06106 Fax: (203) 297-2569
-
- You must be mixed up disk cache with CPU cache. In general, no flame, no
- flame please, disk cache can be implemented quite good by kernel code, as
- in linux, and therefore, you may be better to use software disk cache,
- rather than cached harddisk controller. However, CPU cache is totally
- different story. The kernel cannot speed up main memory (RAM) access by
- software method! Unless the CPU cache design is really bad.
-
- You will be quite sure to gain some speed in increasing CPU cache size.
- However, there is deminishing return in adding CPU cache. Let say, if
- you add extra 32kB secondary cache on top of the 8kB internal cache in
- 486DX, you will really feel the difference in speed. On the other hand,
- if you add extra 256kB CPU cache on top of a system which already has
- 256kB, you may not tell the difference!
-
- *><*><*><*><*><*><*><*><*><*><*><*><*><*><*><*><*><*><*><*><*><*><*><*
- v Stephen Wong Siu-ming v internet: stephenwong@cuhk.hk v
- ^ Computer Services Center ^ bitnet : a080700@cucsc.bitnet ^
- * Chinese University of * phone : (852) 609-8904 *
- v Hong Kong v fax : (852) 603-5001 v
- *><*><*><*><*><*><*><*><*><*><*><*><*><*><*><*><*><*><*><*><*><*><*><*
-
-
-