27 Dec
1998
27 Dec
'98
9:23 a.m.
Hello all,
At 11:16 AM 12/24/98 -0500, you wrote:
I did however do some major tuning tricks on the cmos
and things like
windows swap drive. Some of those old boards if you take the cmos defaults
you get a very pokey system often 1/3 the performance or worse.
Allison
Yes. I've been using a 386DX with 128k cache with Win3.1 and Netscape 3.0.
Doesn't seem significantly slower than my friend's P90 with W95. Sometimes
Netscape crashes with some .JPG files (GP fault in the VGA driver) and I
have to start over.
Haven't tried much newer, it doesn't seem right to require a 300MHz
processor and 64MB ram to mostly read email, much less type mail messages!
I did adjust several things, like remove smartdrv's "buffering" for the HD,
etc.
-Dave
27 Dec
27 Dec
5:51 a.m.
Yes. I've been using a 386DX with 128k cache with
Win3.1 and Netscape 3.0.
Doesn't seem significantly slower than my friend's P90 with W95. Sometimes
Netscape crashes with some .JPG files (GP fault in the VGA driver) and I
have to start over.
Haven't tried much newer, it doesn't seem right to require a 300MHz
processor and 64MB ram to mostly read email, much less type mail messages!
I did adjust several things, like remove smartdrv's "buffering" for the
HD, etc.
-Dave
That's why it's Windows 95. It takes forever to boot. Anyway, cache's the
only thing that makes the K6-2 slower than a PII at the same clock speed.
If I remove my cache entirely on my 200MHz MMX system, then it performs
considerably slower than my 486 with 136K (IE 8K on chip, 128K on the
motherboard) cache. Also the bus speed matters a lot, especially on a
pre-486 (or 486SX with 487) system if you're doing heavy math, etc. I
don't know why Intel's 'low cost' processors are always so bone headed:
486SX, which removed the one true thing that made it a 486, 8088, removing
the crucial 16-bit bus of the 8086, 386SX, which worked pretty well, but
still halfed the external bandwidth (did Intel ever make a cheap version of
a 286?), and now we've got Celeron: Until the Celron A, no cache at all...
Tim
------------------------------------------------------*
*Ever onward, always forward. *
*Tim D. Hotze Panel Member, The Ultimate Web Host List*
*http://www.webhostlist.com worldsfate(a)geocities.com *
------------------------------------------------------*
1:01 p.m.
Tim Hotze <tim(a)thereviewguide.com> wrote:
I don't know why Intel's 'low cost'
processors are always so bone headed:
486SX, which removed the one true thing that made it a 486, 8088, removing
the crucial 16-bit bus of the 8086, 386SX, which worked pretty well, but
still halfed the external bandwidth (did Intel ever make a cheap version of
a 286?), and now we've got Celeron: Until the Celron A, no cache at all...
Well, playing Devil's Advocate for a moment...
Pretend you're Intel. (Only smarter.) How would YOU make a low cost version
of the 8086, in a non-bone-headed way?
The 386?
The 486?
The Pentium II?
Bear in mind that you've already done all of the expensive engineering
necessary for the expensive full-performance processors; those are sunk
costs. But now you want to offer a lower-priced processor, without
seriously diminshing sales of your full price model.
It's not a simple problem with an easy answer. I'm very reluctant to claim
that Intel did it wrong.
And in case you're wondering, no, I've never worked for Intel. I personally
find the x86 architecture revolting. Despite that, I have a great deal of
respect for the engineers that managed to take a toaster controller from
1978 and turn it into a world-class 32-bit superscalar processor. Imagine
what they could have accomplished if they'd started from a GOOD architecture.
Eric
7:48 p.m.
Well, playing Devil's Advocate for a moment...
Sounds like fun. ;-) With my computer going haywire (Right now, I'm on mu
'auxiliary' 486 computer... I suspect a virus, but my antivius software
didn't find one, but I don't really trust it. I only really trust it if it
find a virus...)
Pretend you're Intel. (Only smarter.) How would
YOU make a low cost
version of the 8086, in a non-bone-headed way?
Slower clock speeds. This means that you literally can have a plug in
upgrade. You could also make one with a larger micron process and more
vertical and not worry about heating, but that gets to be bone headed. A
better idea would be to learn from competitors, and start making a more
economical way of producing chips.
But I see your point... but remember, before being used on the IBM PC the
8088 had been on the market for 3 years, and probably in OEM's hands a year
before that.
The 386?
Lower clock speeds... kind of a all-solving solution.
The 486?
The 486 was just plain dumb. You didn't really need a 486SX at all. You
could keep on making a 386, but Intel was afraid of AMD catching up to them
(with their 40MHz 386's and all). You could have also made a 386 to go in
a 486 socket.
The Pentium II?
The Celeron was actually a pretty good idea, too good. They should have
made the 128K cache at half clock speed like in the PII. A Celeron 333
will perform better than a PII 300 in the same system, and you can
overclock the Celeron to compete with almost all Intel processors.
Bear in mind that you've already done all of the
expensive engineering
necessary for the expensive full-performance processors; those are sunk
costs. But now you want to offer a lower-priced processor, without
seriously diminshing sales of your full price model.
You make a bone headed one, to keep your 'get me a computer now' segment
happy, possibly gaining repeat customers, and you keep prices high on your
upper model saying 'well... sure, but look at this performance! It really
is aimed at...)
It's not a simple problem with an easy answer.
I'm very reluctant to
claim that Intel did it wrong.
Well, they did have a lot of proffessionals a lot smarter than I am working
on the problem, and I'm willing to bet they had good ideas, but may have
been shot down somewhere between 'cost' and 'competion'.
And in case you're wondering, no, I've never
worked for Intel. I
personally find the x86 architecture revolting. Despite that, I have a
I'll say that I don't really like it, and that we could have done much
better. MIPS, SPARC, Alpha, all better than x86, and if they had
competition like the x86 market does, then it'd be a good assumption to
guess that the cost might even be halfed and performance considerably faster.
great deal of respect for the engineers that managed to
take a toaster
controller from1978 and turn it into a world-class 32-bit superscalar
processor. Imaginewhat they could have accomplished if they'd started
from a GOOD architecture.
Well, I'm glad that Merced means the end to this lousy architecture. It
wasn't until this year that we began seriouly getting rid of ISA, which,
aside from going from 8 to 16 bit, hasn't changed all that much since the
days of the XT. Note this is good for the collector side of me, but very
bad for the 'innovative, creative, etc.' side of me.
Eric
Tim
------------------------------------------------------*
*Ever onward, always forward. *
*Tim D. Hotze Panel Member, The Ultimate Web Host List*
*http://www.webhostlist.com worldsfate(a)geocities.com *
------------------------------------------------------*
28 Dec
28 Dec
5 a.m.
New subject: Virus Checking
Tim Hotze wrote:
Sounds like fun. ;-) With my computer going haywire (Right now, I'm on mu
'auxiliary' 486 computer... I suspect a virus, but my antivius software
didn't find one, but I don't really trust it. I only really trust it if it
find a virus...)
One of the ways I use to check for a virus is to run CHKDSK and look at the
total memory. For a number of viruses I have seen, the normal number of
655360 gets reduced to some other value. Not the best way of checking of
course, but with many of the viruses I have seen, it gives a fast
indication.
8:03 a.m.
Pretend
you're Intel. (Only smarter.) How would YOU make a low cost
version of the 8086, in a non-bone-headed way?
Lower clock speeds... kind of a
all-solving solution. 
29 Dec
29 Dec
12:44 p.m.
Tim Hotze wrote:
But I see your point... but remember, before being
used on the IBM PC the
8088 had been on the market for 3 years, and probably in OEM's hands a year
before that.
The 8086 had been around that long, the 8088 was done at IBM's
request -- the marketing concept worked well enough that the later
SX processors were done on purpose.
--
Ward Griffiths <mailto:gram@cnct.com> <http://www.cnct.com/home/gram/>
WARNING: The Attorney General has determined that Alcohol, Tobacco,
and Firearms can be hazardous to your health -- and get away with it.
9342
days inactive
9344
days old
6 comments
6 participants
participants (6)
-
cisin＠xenosoft.com -
ddameron＠earthlink.net -
eric＠brouhaha.com -
gram＠cnct.com -
marvin＠rain.org -
tim＠thereviewguide.com