[simpits-tech] Monitor question

Alan D. Mazurka simpits-tech@simpits.org
Wed, 11 Dec 2002 10:36:38 -0500


--=====================_1711278==_.ALT
Content-Type: text/plain; charset="us-ascii"; format=flowed

Hi, Ido

thanks for the links, i will devour them.

for interested readers, in an attempt to answer my own question(s) i offer 
the following AGP/PCI articles from Tom's Hardware:

a 1977 article describing the intent of AGP (and some comparisons to PCI)
http://www17.tomshardware.com/graphic/19970805/agp-01.html

a 2000 article describing the impact of AGP speed (same one Ido mentions)
http://www17.tomshardware.com/mainboard/20000214/index.html

bottom line: AGP is a dedicated bus: solely for the graphics processor. if 
you get onto PCI, you get in line with any and all other devices which also 
want to use the bus. this has unpredictable real-life considerations from 
system to system, since everyone has a different set of cards plugged into 
their computers, PLUS varying types of DMA controllers, memory speeds, etc.

there are other, more technical issues (plus some goodies used by various 
graphics manufacturers) to get extra duty over AGP.

real-life use will effect PCI performance in a manner other than can be 
indicated by raw bandwith numbers alone. the problem is made a bit more 
apparent when flightsimmers overcome the "speed" of the original product 
with their custom designed 2-gazillion poly airplane landing at the 
9-gazillion poly airport.

it gets MUCH worse when you hook 100 2-gazillion poly freight cars (with 
their accompanying 16 MEG textures) together in train-sim, and expect it to 
run like out-of-the-box. these are situations where people are creating 
huge throughput not necessarily envisioned by the program's designers, but 
DO take advantage of AGPs capabilities.

by the way, that Radeon 9700 card sounds unbeeeeelieeeeeevable.

  - adm -

At 04:36 PM 12/11/02 +0200, you wrote:

>There is a premium - the AGP is faster - 2GB with AGP x8 as opposed to PCI
>with 133MHz , and it has it's own bus - it's not on the PCI bus,
>read this article for the exact difference
>http://www.tomshardware.com/mainboard/20000214/index.html
>But ( big one ) for the high end cards, the memory throughput is not even
>close to what the GPU needs so you can only rely on the internal memory,
>in the low end cards you will see an improvement.
>read the conclusion here :
>http://www.tomshardware.com/graphic/200210041/nv18_nv28-08.html#new_cards_with_nvidia_geforce4_ti42008x_and_mx4408x
>if you read the entire article do note that the 440 new version has a
>higher clock not just AGP x8.
>
>Ido


----------

Alan D. Mazurka                    Webspace Design & Implementation
adm.design@verizon.net

--=====================_1711278==_.ALT
Content-Type: text/html; charset="us-ascii"

<html>
Hi, Ido<br><br>
thanks for the links, i will devour them.<br><br>
for interested readers, in an attempt to answer my own question(s) i
offer the following AGP/PCI articles from Tom's Hardware: <br><br>
a 1977 article describing the intent of AGP (and some comparisons to
PCI)<br>
<a href="http://www17.tomshardware.com/graphic/19970805/agp-01.html" eudora="autourl">http://www17.tomshardware.com/graphic/19970805/agp-01.html</a><br><br>
a 2000 article describing the impact of AGP speed (same one Ido
mentions)<br>
<a href="http://www17.tomshardware.com/mainboard/20000214/index.html" eudora="autourl">http://www17.tomshardware.com/mainboard/20000214/index.html</a><br><br>
bottom line: AGP is a dedicated bus: solely for the graphics processor.
if you get onto PCI, you get in line with any and all other devices which
also want to use the bus. this has unpredictable real-life considerations
from system to system, since everyone has a different set of cards
plugged into their computers, PLUS varying types of DMA controllers,
memory speeds, etc.<br><br>
there are other, more technical issues (plus some goodies used by various
graphics manufacturers) to get extra duty over AGP.<br><br>
real-life use will effect PCI performance in a manner other than can be
indicated by raw bandwith numbers alone. the problem is made a bit more
apparent when flightsimmers overcome the &quot;speed&quot; of the
original product with their custom designed 2-gazillion poly airplane
landing at the 9-gazillion poly airport.<br><br>
it gets MUCH worse when you hook 100 2-gazillion poly freight cars (with
their accompanying 16 MEG textures) together in train-sim, and expect it
to run like out-of-the-box. these are situations where people are
creating huge throughput not necessarily envisioned by the program's
designers, but DO take advantage of AGPs capabilities.<br><br>
by the way, that Radeon 9700 card sounds unbeeeeelieeeeeevable.<br><br>
&nbsp;- adm -<br><br>
At 04:36 PM 12/11/02 +0200, you wrote:<br><br>
<blockquote type=cite class=cite cite>There is a premium - the AGP is
faster - 2GB with AGP x8 as opposed to PCI<br>
with 133MHz , and it has it's own bus - it's not on the PCI bus,<br>
read this article for the exact difference<br>
<a href="http://www.tomshardware.com/mainboard/20000214/index.html" eudora="autourl">http://www.tomshardware.com/mainboard/20000214/index.html</a><br>
But ( big one ) for the high end cards, the memory throughput is not
even<br>
close to what the GPU needs so you can only rely on the internal
memory,<br>
in the low end cards you will see an improvement.<br>
read the conclusion here :<br>
<a href="http://www.tomshardware.com/graphic/200210041/nv18_nv28-08.html#new_cards_with_nvidia_geforce4_ti42008x_and_mx4408x" eudora="autourl">http://www.tomshardware.com/graphic/200210041/nv18_nv28-08.html#new_cards_with_nvidia_geforce4_ti42008x_and_mx4408x</a><br>
if you read the entire article do note that the 440 new version has
a<br>
higher clock not just AGP x8.<br><br>
Ido</blockquote>
<x-sigsep><p></x-sigsep>
<hr>
<br>
Alan D.
Mazurka&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
Webspace Design &amp; Implementation<br>
adm.design@verizon.net&nbsp;&nbsp;&nbsp;&nbsp; <br>
</html>

--=====================_1711278==_.ALT--