home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!uwm.edu!ee!bloc1469
- From: bloc1469@ee.ee.uwm.edu (Gregory R Block)
- Newsgroups: comp.sys.amiga.advocacy
- Subject: Re: Future Amiga chipsets
- Date: 23 Dec 1992 23:49:18 GMT
- Organization: Electrical Engineering Dept. University of Wisconsin - Milwaukee
- Lines: 118
- Message-ID: <1hattuINNbok@uwm.edu>
- References: <1992Dec22.173828.5167@oracle.us.oracle.com> <1992Dec22.200816.25990@sol.ctr.columbia.edu>
- NNTP-Posting-Host: 129.89.2.33
-
- In article <1992Dec22.200816.25990@sol.ctr.columbia.edu> jerry@msi.com (Jerry Shekhel) writes:
- >David Navas (dnavas@oracle.uucp) wrote:
- >: Well, personally, I would define crippled as any machine whose principle
- >: communication's bus cannot transfer memory at anything even approaching a
- >: rough approximation of processor speed.
-
- >Does anyone else define it that way? Face it, your definition is totally
- >arbitrary.
-
- No, it's not, it actually has a large basis in reality. Let's take
- the explanation a little further. A bus, in theory, should be fast
- enough to allow the CPU to transfer at speeds approaching or equal to
- the rough approximation of the maximum speed at which the CPU can
- transfer data. Anything less, and the bus becomes a bottleneck which
- must be dealt with. When that bottleneck becomes large, such as a
- 33mhz i486 sitting on an ISA architecture, that becomes a crippling
- bottleneck. For ISA transfers, or anything of a similar reign whose
- bus speeds are minute in size compared to the throughput that is
- needed, the bottleneck cripples the system.
-
- It's why so much research has gone into advanced bus technology, it's
- why you're seeing some of the brand new innovations like SCI and other
- scaleable buses come into existence, testing theories of operation.
-
- >Really? Well, in *my* universe the average hard drive isn't capable of
- >more than, say, 1.5-2MB/sec, which ISA can easily deal with, with plenty of
- >room to spare. What about your universe?
-
- So I suppose that just because your hard drive can go that fast, you
- should be happy? Look at the TIME for that transfer. If you've got
- an i486, you'd see much better use out of it--you're transferring
- PACKETS of data, not byte by byte. Those packets transfer faster, and
- you can get back to REAL work with the CPU. In the end, you're
- spending less time waiting for your bus.
-
- >I define "everyday tasks" for an expansion bus as "interfacing with normal
- >peripherals found in most personal computers". Is that an unfair definition?
- >What good is a 550 GB/s bus if all your peripherals being cranked at once
- >don't add up to 5MB/s?
-
- Because unless you're doing polled io waiting for the bus to give you
- each byte, you can spend less time transferring data and more time
- doing things. Don't make the mistake of believing that there's no
- point to a faster bus when your throughput isn't there. After all,
- wouldn't it be nice to actually add memory? Wouldn't it be great if
- you could access your VRAM-based graphics card at something faster
- than 8mhz? Those kinds of things are REAL things. They're not things
- made up, people use buses for these kinds of things NOW.
-
- >Where are you getting your information? LocalBus supports any type of
- >peripheral, not just video. Look at a Gateway ad; most of their PC's now
- >ship with LocalBus disk controllers. Besides, the fact that a bunch of
- >frustrated Amiga fanatics consider PC's "banally mediocre" has no bearing
- >on reality. Your attempt to pass it off as common knowledge is typical as
- >well as laughable.
-
- There are also serious problems with tying a bus too closely to any
- particular processor. Especially now that the 586 is on the way.
- It's riscy, to coin a term. :)
-
- Frankly, I'd have liked to see a better bus standard come out of the
- PC industry.
-
- >Nice speech, but totally pointless. Technology doesn't exist for technology's
- >sake -- it has to be useful for human beings, *today*. I have to have a
-
- Don't kid yourself by believing that people don't BUY technology for
- technology's sake, though.
-
- >computer *today*, and *today* there aren't any peripherals which would
- >seriously tax the ISA bus. Why should I spend extra for a fast bus and
- >get *nothing*, when I could spend the money on a 24-bit graphics card and
- >get *something*?
-
- We're not talking about taxing an already flimsy bus, pal. We're
- talking about the ISA bus being a BOTTLENECK. Something that would
- tax an ISA bus? The 24-bit graphics card you just mentioned. If
- implemented with VRAM, you'll spend your time waiting for your bus to
- do your transfers. Your solution? Stick it onto the VESA bus.
- LocalBus. And screw forward-compatibility. Because VESA is far, far
- too tied to the current intel style of 486... I can't venture to say
- how or what kinds of modifications, visible or not, will be needed to
- make it work with the Pentium.
-
- The nice thing about Zorro III is that it really ISN'T tied. It's a
- really, really good bus design. Something that is uncommon in the
- world of intel-based PC's, unfortunately.
-
- >Thanks for the advice :-P Come on, face it. Computers are tools. When I'm
- >in the market for a drill or a screwdriver, I'll take the one which is the
- >most capable of helping me with the tasks I need to accomplish, even if
- >there's another one whose specs look more macho on paper.
-
- Consider this: Your power screwdriver, your PC, has an engine which can
- deliver the torque you want, but not at a good, fast speed. Sure, it
- gets the short screws in just fine. But give a faster power screwdriver
- to someone else, with an engine capable of delivering a faster speed,
- and you'll see that person put that much more screws into the wall.
- Given a limited number of screws, let's say, oh, 2/minute, you'll be
- spending more time than he will using your power screwdrier on those
- screws. He'll have larger "gaps". And on computers, it's those
- "gaps" that let you get REAL work done.
-
- >I ask you again. What can't my ISA PC do directly because of ISA bus
- >limitations? Look, I'll even help you by mentioning one thing: DMA to
- >memory above 16MB can't happen due to the ISA's 24 address lines, but then
- >again, it's invisible at the end-user level. Is there anything else?
-
- Bottlenecks seem to be invisible to the user, too. You don't notice
- them unless you're on a system which makes it obvious...
-
- Greg
-
- --
- (: (: (: (: Have you overdosed on smileys today? Why NOT!?! :) :) :) :)
- (: "Our father, who art in Iowa, Hollow be thy head, Thy ideas run :)
- (: Thy will be done, At Commodore as it is at Apple" -Dan Barrett :)
- (: (: (: (: (: (: (: (: (: (: (: (: () :) :) :) Wubba, the Dark Angel :)
-