Pentium vs. 486 speed
Stefan Raaijmakers
I'm thinking of buying a new pc in one or maybe two months, and of course
the first thing that comes to mind is a pentium. But now a friend of mine
says that it is better to buy a 486dx4/100 in stead of a p90/p100. Now i
was wondering wich is faster. A pentium has the option of doing 2
commands at once i beleive, but he claims that not may programms do that
and it is better to spend your money on more memory. And i was also
wondering if a pentium needs less clock ticks to prosses its 'commands'
(how do you call these?:) . My guess is go for a pentium with 8Mb instead
of a 486dx4/100 with more MB's, because i'm not running linux or windoze,
only a occational 3d game..
Tell me what YOU think.
Alex Bachin
>Hi evryone,
>
>I'm thinking of buying a new pc in one or maybe two months, and of course
>the first thing that comes to mind is a pentium. But now a friend of mine
>says that it is better to buy a 486dx4/100 in stead of a p90/p100. Now i
I think that Pentium 90/100 is better than 486dx4/100, but ...
>was wondering wich is faster. A pentium has the option of doing 2
>commands at once i beleive, but he claims that not may programms do that
>and it is better to spend your money on more memory. And i was also
>wondering if a pentium needs less clock ticks to prosses its 'commands'
>(how do you call these?:) . My guess is go for a pentium with 8Mb instead
>of a 486dx4/100 with more MB's, because i'm not running linux or windoze,
.. 486dx4/100 with 16 Mb is better than Pentium 90/100 with 8 Mb
when running Linux, Windows, OS/2, etc.
>only a occational 3d game..
>
>Tell me what YOU think.
--
Alex Bachin
bac...@garant.msu.ru
-
Le Fongus jaune
In article <3vpibb$eeg...@news.dnai.com>, vbaz...@imsisoft.com (Victor Bazarov) writes:
-=)> Well, 486DX4/100 has 25 (yes, twenty five!) MegaHertz bus clock (where Pentium
-=)> has 90/100MHz). Speed is quadruple only inside the main processor. So it's
-=)> even slower on some operations than 486DX/50, or 486DX2/66. We've checked,
-=)> we've got couple of those babies... :-(
No, no ,no... dx4 has a 33MHz bus, and a p90 has a 60MHz and a
P100 has a 66MHz bus clock... P120 should be a 60MHz.
-=)> >Tell me what YOU think.
-=)> Pentium would serve you longer than 486
Yep, go for a Pentium if you can afford it (but at least a
P75) forget the P60 or P66. A P90 is certainly the best choice
price/performances.
--
Þökk fyrir
---------------------------------------------------------
Le fongus jaune. gros...@epita.fr
stephane grosjean
Stefan Raaijmakers
: In article <3vm1p9$7...@liberator.et.tudelft.nl>,
: ste...@djo.wtm.tudelft.nl (Stefan Raaijmakers) wrote:
: >says that it is better to buy a 486dx4/100 in stead of a p90/p100. Now i
: Well, 486DX4/100 has 25 (yes, twenty five!) MegaHertz bus clock (where Pentium
: has 90/100MHz). Speed is quadruple only inside the main processor. So it's
: even slower on some operations than 486DX/50, or 486DX2/66. We've checked,
: we've got couple of those babies... :-(
Well.. that's one of the delemma's i've got, I Know the dx4 has a 33 Mhz
bus (it only triple's) But i've heard that the p100/90 does the same
thing.. A 50 Mhz. bus is probably the fastet one around.. but too fast
for many localbus cards . So there was realy one of my problem's bu i
didn't seem to get it on my screen..
: >(how do you call these?:) . My guess is go for a pentium with 8Mb instead
: 'Instructions'. Well, there are some tricks, where Pentium beats 486. And one
: of these is existence of 2 pipes (like you said, two instructions in a clock
: tick).
Ofcourse, pretty stupid of me, I think got a little confused wit
'opcodes' and 'mnemonics'.. sorry :)
: >of a 486dx4/100 with more MB's, because i'm not running linux or windoze,
: >only a occational 3d game..
: >
: >Tell me what YOU think.
: Pentium would serve you longer than 486, since it was on market shorter
: period. You have made the right decision to buy Pentium with 8M, go for it.
: And you can upgrade your computer to more memory later.
Had a lot of the same advice lately, Thank's evryone
: Victor.
: --
: -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
: If it's stupid, but works, it isn't stupid.
: Murphy
: -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Everett Johnson
- Snip -
>Well, 486DX4/100 has 25 (yes, twenty five!) MegaHertz bus clock (where Pentium
>has 90/100MHz). Speed is quadruple only inside the main processor. So it's
>even slower on some operations than 486DX/50, or 486DX2/66. We've checked,
>we've got couple of those babies... :-(
- Snip -
Actually, DX4 are TRIPLE speed not QUADRUPLE speed internally. A
DX4/100 has a bus speed of 33 MHz.
Christopher HILL
>I'm thinking of buying a new pc in one or maybe two months, and of course
>the first thing that comes to mind is a pentium. But now a friend of mine
>commands at once i beleive, but he claims that not may programms do that
>and it is better to spend your money on more memory. And i was also
>wondering if a pentium needs less clock ticks to prosses its 'commands'
>only a occational 3d game..
In most situations, a 486dx4/100 will give you roughly the same performance
as a P5/60. A P5/90 should be noticeably faster.
Christopher Hill
Aiea, Hawai`i (O`ahu)
Victor Bazarov
ste...@djo.wtm.tudelft.nl (Stefan Raaijmakers) wrote:
>
>I'm thinking of buying a new pc in one or maybe two months, and of course
>the first thing that comes to mind is a pentium. But now a friend of mine
>says that it is better to buy a 486dx4/100 in stead of a p90/p100. Now i
Well, 486DX4/100 has 25 (yes, twenty five!) MegaHertz bus clock (where Pentium
has 90/100MHz). Speed is quadruple only inside the main processor. So it's
even slower on some operations than 486DX/50, or 486DX2/66. We've checked,
we've got couple of those babies... :-(
>was wondering wich is faster. A pentium has the option of doing 2
>commands at once i beleive, but he claims that not may programms do that
>and it is better to spend your money on more memory. And i was also
>wondering if a pentium needs less clock ticks to prosses its 'commands'
>(how do you call these?:) . My guess is go for a pentium with 8Mb instead
'Instructions'. Well, there are some tricks, where Pentium beats 486. And one
of these is existence of 2 pipes (like you said, two instructions in a clock
tick).
>of a 486dx4/100 with more MB's, because i'm not running linux or windoze,
>only a occational 3d game..
>
>Tell me what YOU think.
Pentium would serve you longer than 486, since it was on market shorter
period. You have made the right decision to buy Pentium with 8M, go for it.
And you can upgrade your computer to more memory later.
Victor.
Unknown
> Well, 486DX4/100 has 25 (yes, twenty five!) MegaHertz bus clock (where Pentium
No, twenty five! Yes, thirty three! DX4 chips are clock _tripled_, not quadrupled.
Ask Intel why the name's weird, I dunno. AMD & Cyrix simply followed suit.
Pentiums are clock multiplied, as well. The bus runs at 25 Mhz for the P75(x3),
30 Mhz for the P60(x2), P90(x3), & P120(x4), and 33 Mhz for the P66(x2),
P100(x3), & P133(x4).
I'd be interested in finding out the relative performance of the AM486DX4-120,
which clocks the bus at 40Mhz, and the P120 with its slower bus. It seems
possible that some applications that aren't cpu-intensive might actually do
better on the 486. Also, the P120 m/b (including cpu) costs about four times
as much as a comparable 486 m/b; is the performance gain comparable to
the price gain? I.e. is it at least four times faster?
I recently bought a 486DX4-120 because it seemed like the best buy. Not
the fastest, but the best value for the money.
Grinch
Victor Bazarov
>- Snip -
>>Well, 486DX4/100 has 25 (yes, twenty five!) MegaHertz bus clock (where
Pentium
>>has 90/100MHz). Speed is quadruple only inside the main processor. So it's
>>even slower on some operations than 486DX/50, or 486DX2/66. We've checked,
>>we've got couple of those babies... :-(
>- Snip -
>Actually, DX4 are TRIPLE speed not QUADRUPLE speed internally. A
>DX4/100 has a bus speed of 33 MHz.
>
Really? 8-(... Whatever :-/
--
Victor Bazarov, International Microcomputer Software, Inc.
San Rafael, California. (415) 257-3000 -=-=-=-=-=-=-=-=-=-
Jim Grossl
>From: vbaz...@imsisoft.com (Victor Bazarov)
>Subject: Re: Pentium vs. 486 speed
>Date: Thu, 03 Aug 95 23:58:21 GMT
>>- Snip -
>>Actually, DX4 are TRIPLE speed not QUADRUPLE speed internally. A
>>DX4/100 has a bus speed of 33 MHz.
>>
>Really? 8-(... Whatever :-/
>--
>Victor Bazarov, International Microcomputer Software, Inc.
>San Rafael, California. (415) 257-3000 -=-=-=-=-=-=-=-=-=-
Yes really, I have a DX4-100 the board runs at 33 Mhz. To all of you who said
to buy the Pent machines you were right. The DX4-100 is an excellent machine,
the Pentium (90 and above) are better. Memory upgrades are reasonably cheap.
****************************************************************
jg...@micron.net Jim Grossl /
----------------------------------------------------------------
"Experience is the best tool, at least that's what she said." /
"Welcome to my town, now please leave." /
****************************************************************
Garrett Drysdale
ste...@djo.wtm.tudelft.nl (Stefan Raaijmakers) wrote:
> Well.. that's one of the delemma's i've got, I Know the dx4 has a 33 Mhz
> bus (it only triple's) But i've heard that the p100/90 does the same
> thing.. A 50 Mhz. bus is probably the fastet one around.. but too fast
> for many localbus cards . So there was realy one of my problem's bu i
> didn't seem to get it on my screen..
>
Pentium 90's run a 60 MHz chip to L2 bus, and a 30 MHz PCI bus.
Pentium 100's run a 66 MHz chip to L2 bus, and a 33 MHz PCI bus.
===========================================================
Garrett Drysdale
NSP Applications Lab - Intel Corp.
I am not an Intel spokesperson.
===========================================================
Gavin Estey
gdry...@sedona.intel.com "Garrett Drysdale" writes:
> Pentium 90's run a 60 MHz chip to L2 bus, and a 30 MHz PCI bus.
> Pentium 100's run a 66 MHz chip to L2 bus, and a 33 MHz PCI bus.
>
I've got a Pentium 75, how much different is this?
--
Gavin Estey
=================================================================
E-MAIL ga...@senator.demon.co.uk PHONE/FAX +44 (01276) 476503
S-MAIL 2 Lovells Close, Lightwater, Surrey, GU18 5RS, ENGLAND
=================================================================
Maintainer of the rec.games.programmer FAQ. This can be found at:
ftp://x2ftp.oulu.fi/pub/msdos/programming/faq/rgp_faq.126
Tony Doimeadios
: In article <3vm1p9$7...@liberator.et.tudelft.nl>, Stefan Raaijmakers wrote:
: >I'm thinking of buying a new pc in one or maybe two months, and of course
: >says that it is better to buy a 486dx4/100 in stead of a p90/p100. Now i
: >and it is better to spend your money on more memory. And i was also
: >wondering if a pentium needs less clock ticks to prosses its 'commands'
: >(how do you call these?:) . My guess is go for a pentium with 8Mb instead
the 2-commands-at-once is called pipe-lining the instructions and MOST
compilers do not support it (prob. will in a few years tho). For
Windows, if you get more memory, it will outperform a newer CPU with less
RAM. Computer Shopper did some test and a DX4-100 with 16 megs of RAM
out-did a P5-90 with 8 megs of RAM (not by much tho).
I, too, am shopping for a new computer. Right now, I can buy a DX4-100
motherboard (CPU included) for $200. A P5-100 will cost me over $800!
If I got the DX4-100, I'd save enough money to get 16 megs of ram, a 1.25
gig HD, etc... Look for prices to DROP (ie: plummet) around Christmas
time. With the new P6 coming out in September, they're gonna have to
drop the prices or they will be stuck with some "old" cpu's....
-Sandy
Harald Feldmann
>Christopher HILL (ch...@BIX.com) wrote:
>: In article <3vm1p9$7...@liberator.et.tudelft.nl>, Stefan Raaijmakers wrote:
>: >I'm thinking of buying a new pc in one or maybe two months, and of course
>: >the first thing that comes to mind is a pentium.
>:> My guess is go for a pentium with 8Mb instead
>: >of a 486dx4/100 with more MB's, because i'm not running linux or windoze,
>: >only a occational 3d game..
Wow, and you need a Pentium to do that ?! :)
>the 2-commands-at-once is called pipe-lining the instructions
No it's not. It's called dual piped execution and is only vaguely
related to pipelining. Pipelining is also done on the 486 (and earlier
processors in a certain way).
The Pentium can organize this by itself and also uses it with older
software, however the old software is not *optimized* to use it to its
full extent. 'Old' software (like for instance MSDOS 5.0 ...) also has
jump targets on odd boundaries, a no-no in decent optimization.
>and MOST compilers do not support it
True. Most compilers do not have instruction scheduling optimized for
the Pentium.
>For Windows, if you get more memory, it will outperform a newer CPU with
>less RAM.
Depends on what you're trying to run.
>-Sandy
The 486 FPU unit is significantly slower than the Pentium one. But
then again, I don't think Doom et al use a lot of FP math do they :)
Go for a 486. Perhaps you can get a nice bargain DX2-66. ?
Harald.
Mike Schmit
writes:
>Christopher HILL (ch...@BIX.com) wrote:
>: In article <3vm1p9$7...@liberator.et.tudelft.nl>, Stefan Raaijmakers wrote:
>: >I'm thinking of buying a new pc in one or maybe two months, and of course
>: >says that it is better to buy a 486dx4/100 in stead of a p90/p100. Now i
>: >was wondering wich is faster. A pentium has the option of doing 2
>: >commands at once i beleive, but he claims that not may programms do that
>: >and it is better to spend your money on more memory. And i was also
>: >wondering if a pentium needs less clock ticks to prosses its 'commands'
>: >only a occational 3d game..
>
>compilers do not support it (prob. will in a few years tho).
This is not a correct statement (the way you've worded it.) The Pentium
and the 486 have pipelining. Pipelining is where an instruction moves
from stage to stage until all operations required for that instruction
are completed. The Pentium has 2 pipelines (the 486 has one). The
Pentium will automatically "pair" two instructions and execute them
at the same time, if they meet the critieria.
The compilers on the market do attempt to arrange instructions so that
they will meet the pairing criteria to some degree. On the whole, they
are not really good at doing this, but they do it.
Mike Schmit
-------------------------------------------------------------------
msc...@ix.netcom.com author:
408-244-6826 Pentium Processor Programming Tools
800-765-8086 ISBN: 0-12-627230-1
-------------------------------------------------------------------
Philip O'Carroll
>In <40mv50$q...@blackice.winternet.com> san...@winternet.com (Tony Doimeadios)
>writes:
>>
>>Christopher HILL (ch...@BIX.com) wrote:
>>: In article <3vm1p9$7...@liberator.et.tudelft.nl>, Stefan Raaijmakers wrote:
>>: >I'm thinking of buying a new pc in one or maybe two months, and of course
>>: >the first thing that comes to mind is a pentium. But now a friend of mine
>>: >says that it is better to buy a 486dx4/100 in stead of a p90/p100. Now i
>>: >was wondering wich is faster. A pentium has the option of doing 2
>>: >commands at once i beleive, but he claims that not may programms do that
>>: >and it is better to spend your money on more memory. And i was also
>>: >wondering if a pentium needs less clock ticks to prosses its 'commands'
>>: >(how do you call these?:) . My guess is go for a pentium with 8Mb instead
>>: >of a 486dx4/100 with more MB's, because i'm not running linux or windoze,
>>: >only a occational 3d game..
>>
>>the 2-commands-at-once is called pipe-lining the instructions and MOST
>>compilers do not support it (prob. will in a few years tho).
>This is not a correct statement (the way you've worded it.) The Pentium
>and the 486 have pipelining. Pipelining is where an instruction moves
>from stage to stage until all operations required for that instruction
>are completed. The Pentium has 2 pipelines (the 486 has one). The
>Pentium will automatically "pair" two instructions and execute them
>at the same time, if they meet the critieria.
Lets try again, what you mean is that pipelining is a method of breaking
up instruction execution into smaller steps which can be performed
simultaneously on sequential instructions, like a factory assembly line.
Usually the processor has steps Instruction Fetch(IF), Instruction Decode(DE),
Execute(EX) and Write Back Result (WB). All of these can be done by separate
functional units in the processor. Thus there is no reason why the processor
fetch logic can't fetch the next instruction while the Execute logic is
executing a previous instruction.
Consider:
MOV R1, (R3) Move the contents of memory addressed by R3 into register R1
MOV R2, (R4) ditto but R2 and R4
ADD R3, R1, R2 Add R1 and R2 and put the result in R3
Assume each stage takes one clock cycle.
Without Pipelining we have 3 instructions x 4 Clock cycles = 12 clock cycles
to complete.
MOV IF DEC EX WB
MOV IF DEC EX WB
ADD IF DEC EX WB
With pipelining, the stages can overlap and the total time is 6 clock cycles
this is quite a saving. We are achieving 1 instruction per clock cycle
instead of four.
MOV IF DEC EX WB
MOV IF DEC EX WB
ADD IF DEC EX WB
It is not this easy in reality (of course) because sometimes (for example)
the execute stage of one instruction will depend on the WB stage of a
previous one and the pipling must "stall" the second instruction until
the first instruction completes its WB stage.
>The compilers on the market do attempt to arrange instructions so that
>they will meet the pairing criteria to some degree. On the whole, they
>are not really good at doing this, but they do it.
I have written a small applet to read the pentium performance counters
in real time while in windows and it seems on everyday unoptimised code, you
rarely get more than about 20-25% sustained parallelism. I have run DESCENT on
in Windows 95 with the performance monitor in the background and again, I
don't get any real change in this figure.
The method is to count all instructions in one perf counter and just
V-pipe instructions in the other perf counter (see Byte July 1994 for details
on how to do this) and calculate:
VPipe
-------------
Total-VPipe
Of course the relevance of this approach is questionable because I am
running the test under Windows and consequently I don't know exactly what
tast is running. It is also averaged over 55 milliseconds (Windows timer
resolution) during which time the Pentium is capaple of executing several
million instructions. However, I think it yields a nice rough and ready
figure. The highest sustained MIPS figure I have generated in this way is
90 MIPS for a P-100 during a VC++2.0 compile. That is nearly one instruction
per clock cycle under an ordinary uncontrived load. I think that is fairly
impressive.
A simpler test of the Pentiums parallelism (and current compilers)
would be to time execution with Single pipe execution turned on and off.
YMMV
Philip
PS if anyone wants a copy of this applet it is available for free on FTP
from Simtel in /pub/win3/sysutil It is called p5mon.zip
--
-----------------------------------------------------------------------------
Philip O'Carroll email: p...@maths.tcd.ie Tel: 353-1-8324970
u800946 ee
: In article <3vr84c$c...@liberator.et.tudelft.nl>
: ste...@djo.wtm.tudelft.nl (Stefan Raaijmakers) wrote:
: > Well.. that's one of the delemma's i've got, I Know the dx4 has a 33 Mhz
: > bus (it only triple's) But i've heard that the p100/90 does the same
: > thing.. A 50 Mhz. bus is probably the fastet one around.. but too fast
: > for many localbus cards . So there was realy one of my problem's bu i
: > didn't seem to get it on my screen..
: >
: Pentium 90's run a 60 MHz chip to L2 bus, and a 30 MHz PCI bus.
: Pentium 100's run a 66 MHz chip to L2 bus, and a 33 MHz PCI bus.
Why p5-90 runs a 30MHz PCI bus ?
Isn't it dependent on mainboard ?
In accordant with PCI spec. v2.1 , a PCI bus clock has
nothing to do with host bus (local bus) clock . There is
a bridge between them handling data flow .
I've never measure the PCI bus clock over p5-90 computer ,
but I think it also has a 33MHz freq.
--
Frank Tuan
u80...@Oz.nthu.edu.tw
Kevin Michelizzi
u800946 ee <u80...@Oz.nthu.edu.tw> wrote:
>Garrett Drysdale (gdry...@sedona.intel.com) wrote:
>: In article <3vr84c$c...@liberator.et.tudelft.nl>
>: ste...@djo.wtm.tudelft.nl (Stefan Raaijmakers) wrote:
>
>
>: > Well.. that's one of the delemma's i've got, I Know the dx4 has a 33 Mhz
>: > bus (it only triple's) But i've heard that the p100/90 does the same
>: >
>: Pentium 90's run a 60 MHz chip to L2 bus, and a 30 MHz PCI bus.
>: Pentium 100's run a 66 MHz chip to L2 bus, and a 33 MHz PCI bus.
>
> Why p5-90 runs a 30MHz PCI bus ?
> Isn't it dependent on mainboard ?
> In accordant with PCI spec. v2.1 , a PCI bus clock has
> nothing to do with host bus (local bus) clock . There is
> a bridge between them handling data flow .
> I've never measure the PCI bus clock over p5-90 computer ,
> but I think it also has a 33MHz freq.
>
the motherboard manufacturers make the two boards. They use the same
board for both versions, and they derive their system frequencies from
a common clock source. This allows them to quickly (and cheaply) change
the main clock, and the processor, and have a 100.
It saves them manufacturing costs, component costs, and quality costs.
They don't have to reload different parts into the assembly line to
switch boards when demand dictates.
manufacturing c
--
-------------------------------------------------------------------------
Kevin Michelizzi Any comments are my own, and nobody elses.
bios...@rahul.net They are also guaranteed to not mean what
ke...@neomagic.com you think they mean.
Jen-Chien Tuan
: In article <417qt6$h...@eusapia.nthu.edu.tw>,
: u800946 ee <u80...@Oz.nthu.edu.tw> wrote:
: [snip]
: > Why p5-90 runs a 30MHz PCI bus ?
: "Typically", the 90's are at 30, and the 100's at 33 because of the way
: the motherboard manufacturers make the two boards. They use the same
: board for both versions, and they derive their system frequencies from
: a common clock source. This allows them to quickly (and cheaply) change
: the main clock, and the processor, and have a 100.
: It saves them manufacturing costs, component costs, and quality costs.
: They don't have to reload different parts into the assembly line to
: switch boards when demand dictates.
: manufacturing c
Hm ... You mean the mainboard use half of the xtal osc
clock for PCI bus clock ? ( xtal osc for p5-90 = 60MHz , for p5-100
= 66MHz , maybe it don't use a crystal ..... I don't know )
If I runs a p5-75 ( xtal osc = 50 MHz ) , than I'll got
a 25MHz PCI bus clock ?
: --
: -------------------------------------------------------------------------
: Kevin Michelizzi Any comments are my own, and nobody elses.
: bios...@rahul.net They are also guaranteed to not mean what
: ke...@neomagic.com you think they mean.
--
/| MM /\
|| || ||
|| || || Jen-Chien Tuan
|| /||_ || (035)715131-7572 (Taiwan)
(\||_ ( || ) /||\
)||( )||( /_||_\ tu...@athletes.ee.nctu.edu.tw
(____) (____) (/ \)