home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.ibm.pc.hardware
- Path: sparky!uunet!sdl!stu
- From: stu@Warren.MENTORG.COM (Stu Brown)
- Subject: Re: Is a 486-33 with 0 ws cache faster than a 486-50 with 1 ws ?
- Message-ID: <1992Aug21.144712.26345@Warren.MENTORG.COM>
- Organization: Mentor Graphics, Silicon Design Division
- References: <1992Aug16.060405.15165@ariel.ec.usf.edu>
- Distribution: usa
- Date: Fri, 21 Aug 1992 14:47:12 GMT
- Lines: 64
-
- From article <1992Aug16.060405.15165@ariel.ec.usf.edu>, by fevrier@sunburn.ec.usf.edu. (Ian Fevrier (EE)):
- >
- >
- >
- > HELP!!!
- > I just upgraded from a 486-33 to a 486-50 and have noticed a
- > few strange things that I can't understand.
- >
- > The 486-33 cache operated at zero wait state using 25ns
- > sram; at 33mhz that works out to 30ns per hit .
- >
- > On the 486-50 the cache (20ns chips) operates at 1 wait
- > state. At 50mhz that's 40ns per hit ??????
- >
- > Does that mean that the 486-33 will out-perform the 486-50
- > on memory-intensive operations ???? (both machines have 70ns
- > Dram).
- >
- > Another curiosity, I disabled both internal and external
- > caches on the 486-50 and ran the Norton benchmark; the
- > speed rating dropped from 107.2 to 15.
- >
- > With just the 256k external cache enabled the speed
- > increased to 37.
- >
- > With just the 8k internal cache, the speed was 106 !!! ,
- > almost the same as having both caches enabled.
- >
- > Does this mean that my 256k external cache has been a waste of
- > hard earned money????
- >
- >
- > Any feedback will be appreciated, especially on ways to run
- > my cache at zero wait state. Are there even 486-50's with zero wait
- > caches ?.
- >
-
- As of yesterday I was considering a 486/50 over a 486/33, but
- I read some things which changed my mind.
-
- An article in Sept Computer Shopper said that some vendors were
- using 33MHz board designs and just upgrading chups to 50MHz. I would
- certainly want a board that was designed for 50MHz is I paid that
- premium. Since there is no way to verify that a vendor is doing
- this until after you've bought the board (and can't return it),
- I was really bothered by that.
-
- The vendors I called quoted the Landmark speeds for various boards.
- 386/33 (which I have now) is 68MHz (mine measures 48MHz, but I have
- 80Ns RAM and 32K cache, so I probably have a RAM wait state). The
- 486/33 is 151MHz (a 3x performance boost to me). But 486/50 was
- just 200MHz - not anywhere near what I will see for 486/33 over 386/33.
-
- Another article I read in Byte said that their tests showed very little
- difference between 64K cache and 256K cache in a 486 machine.
-
- I was suprised to learn that a 486/33 was so much faster than a 386/33
- and that a 486/50 was not that much faster than a 486/33 (considering
- the >$200 price difference).
- --
- Stuart Brown
- Mentor Graphics Corporation
- I.C. Group, Warren NJ
- stu@warren.mentorg.com or uunet!sdl!stu
-