home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!usc!cs.utexas.edu!rutgers!njitgw.njit.edu!hertz.njit.edu!dic5340
- From: dic5340@hertz.njit.edu (David Charlap)
- Newsgroups: comp.os.os2.misc
- Subject: Re: Does VESA Local Bus put a huge strain on a CPU?
- Message-ID: <1992Nov5.083832.13479@njitgw.njit.edu>
- Date: 5 Nov 92 08:38:32 GMT
- References: <1992Nov3.195101.7330@njitgw.njit.edu> <Bx63uw.JrD@news.cso.uiuc.edu>
- Sender: news@njit.edu
- Organization: New Jersey Institute of Technology, Newark, N.J.
- Lines: 42
- Nntp-Posting-Host: hertz.njit.edu
-
- In article <Bx63uw.JrD@news.cso.uiuc.edu> nap42487@uxa.cso.uiuc.edu (Nishith A. Patel) writes:
- >dic5340@hertz.njit.edu (David Charlap) writes:
- >>ISA runs at a theoretical 8MHz at a 16-bit datapath, making an
- >>effective 16MB per second if every bus-cycle would be used. This is
- >>almost never the case, though. Most ISA machines max out at about
- >>5-10MB/s. The fastest I've ever seen an ISA card do was 4MB/s. I
- >>think a bus-mastering card can approach the maximum speed, since it
- >>can shut the CPU out of the bus.
-
- One minor correction: ISA can only access the bus every other clock
- cycle, so an 8MB computer has a theoretical maximum at 8MB/s, not 16.
-
- >>LocalBus, I believe, is 32-bits wide, and runs at the CPU's internal
- >>speed. This would deliver a theoretical 100MB/s if every clock was
- >>used for transfer. This, just like with MCA and ISA, is unrealistic.
- >>I think you can expect about 20-50MB/s with an LB slot and an LB card
- >>that keeps the CPU and other cards off the bus.
- >
- >Now one thing I don't understand: If MCA can come close to its theoretical
- >peformance of 20, 40, or 80 MB/s, depending on the card, then why can't LB
- >do the same if it does the ideal things that MCA does to reach the limit?
- >if LB can theoreticsally get 100 MB/s, why is the expected only 20-50 if it
- >does everything right? What is it wasting all those clock cycles on
-
- It probably could, but the CPU would be doing nothing but servicing
- the bus every clock cycle. And there's the other problem that your
- RAM won't be able to keep up with 100MB/s, since most RAM is about
- 60-80ns. With bus-mastering devices, it is theoretically possible for
- two cards to talk to each other at full speed while the CPU goes off
- and does something else.
-
- >And if LB is running at 33 MHz, isn't the max more than 100 MB/s?
-
- LB runs at processor speed. I was assuming a 25MHz system, such as
- the one I have. (No, my system is not LB).
-
-
- --
- David Charlap |"there aren't 50,000 things for which it's worth writing
- dic5340@hertz.njit.edu| software; and the computer industry doesn't have enough
- ----------------------+ programmers to create that much good software.
- Therefore, most of it must be worthless" -- Boris Beizer "The Frozen Keyboard"
-