18-Apr-2024 02:02 GMT.		 UNDER CONSTRUCTION
Anonymous, there are 89 items in your selection (but only 39 shown due to limitation) [1 - 50] [51 - 89]
[Forum] Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?]ANN.lu
Posted on 29-Nov-2003 16:30 GMT by Shotgun (Edited on 2003-11-29 21:16:02 GMT by Christophe Decanini)89 comments
View flat
View list
I am an old time amiga user and have decided to look at what option is available. A part of Amiga hardware and modern pcs which i believe is very important are graphics chips. I was shocked to heard the pegasos2/marvel will only support sub AGPx1 performance. AGP 2x is today the minimal supported standard supported by modern graphics cards why release something below the minimal accpeted standard in 2003? Modern games will have no ability to reach anywhere near the performance levels modern chipsets allow. The pegasos2 already hopeles crippled before even released! People may claim IBM already has 1500 boards but why not show at least one in public then? The people think others are blind followers seem to be acting like this themself. And before anyone claiming i am a blind amiga inc followers. I am not although mainly amigaos4 interests me. The amigaone is just to expensive for me. DE lacks software and support to. Maybe I will own a Pegasos3 with good specs and amigaos4. Or maybe a amigatwo or amigalite if they finally get the prices back down to earth.
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 51 of 89ANN.lu
Posted by hammer on 01-Dec-2003 03:21 GMT
In reply to Comment 12 (Anonymous):
>If somebody made a Pepsi test with you to tell the difference between AGPx2 vs >AGPx8 or an AMD 1400 vs an AMD 2200, you would fail.

AMD K7 Athlon XP @ 2200Mhz 400FSB vs AMD K7 Athlon T-Bird @1.4Ghz DDR266FSB would make a difference in media encoding i.e. rip/encode/burn market and any computable extensive tasks e.g. AI, game physics, software T&L, emulators(e.g. WinUAE), raytracing, software digital video editors(e.g. Ulead VideoStudio 7, Sonic Foundry Vega 4.0), software PVR/DVRs and 'etc'.

Refer to the minimum vs recommended system requirements for Max Payne 2 as an example.
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 52 of 89ANN.lu
Posted by Raffaele on 01-Dec-2003 03:22 GMT
In reply to Comment 20 (Neko):
mr neko wrote:


>Okay, Raffaele -
>[... and more...]


O__o -(Gah???)


O__O' -(ahem...)


>Buffer time between waking up and posting on ANN.lu
>has hereby been increased to 25 minutes,
>and wearing glasses has been made mandatory
>for sitting at the computer.
>
>Sorry, Raff :)
>
>=Neko=


^__^ -(No probs)
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 53 of 89ANN.lu
Posted by hammer on 01-Dec-2003 03:45 GMT
In reply to Comment 23 (Eva):
> Your ignorance in this sector his so huge m8.
>Simpky take a PC, go in the Bios and switch you Agp Bus speed from 4-8x to 1-2x.
>You will notice only a so little slow down that they can be mesaured only with >benchmarks programs (and the difference in 3dMark2003 with a Radeon 9660 from Agp >1x to Agp 8x is 3%).
>People is so ignorant in these times.
3dMark2003 is not a true game per se i.e. unlike UT2003 (play it with a custom map like 2019). Try with AquaMark3, UT2003, Quake3, RTCW, 'etc'. The importance of AGP transfers speed comes into play when said data exceeds the onboard memory of the AGP card i.e. in serious Open GL tasks.

AGP8X acts as insurance for future developments (i.e. "future proofing") i.e. reducing the need of video card upgrade occurrence.
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 54 of 89ANN.lu
Posted by hammer on 01-Dec-2003 04:42 GMT
In reply to Comment 53 (hammer):
Addendum;

The importance of AGP transfers speed comes into play when game's graphic data is stored AGP Aperture area. This is where your BIOS’s AGP Aperture setting comes in.
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 55 of 89ANN.lu
Posted by hammer on 01-Dec-2003 05:18 GMT
In reply to Comment 54 (hammer):
Addendum II

There's some chipset issues between NVidia's nForceII vs Intel's Granite Bay chipset in relation to AGP8X transfers and memory conrollers i.e. Dual channel memory controllers aren’t made equal.

Refer
http://www.lostcircuits.com/video/quadro_980/8.shtml

[quote]
nForce2 and AGP 8X

Both the Granite Bay and the nForce2 chipset are using what falls under the general umbrella of Dual Channel DDR, however, as we explained in our Granite Bay article, aside from the name and a total of 128 bit memory bus width, the two technologies have very little in common. The nForce2 SPP uses two independent memory controllers in crossbar configuration whereas the Granite Bay combines two banks of memory to one 128 bit wide "Überbank". The net effect is higher granularity on the nForce2 chipset vs. higher total bandwidth on the Granite Bay. In terms of functionality, the difference is that the Granite Bay can only address one bank at the time whereas the two independent controllers of the nForce2 chipset can set up independent DMA engines or else serve the CPU. For AGP that needs to contrive transfer of both geometry data from the CPU and texture data from the graphics aperture, this higher granularity could be a prerequisite for utilizing the increased bandwidth of the 8X protocol.
...
The Pitfalls
...
Again, that does not mean that there will be a difference if all the switches are working. In fact, our findings that there was no difference whatsoever in any other benchmarks even with hardware-forced AGP4X only reinforces our hypothesis of bandwidth vs. granularity as the major deciding factor for the ability to capitalize on AGP8X. It appears as if at this point nVidia's nForce2 chipset is the only player in the field that really can take advantage of AGP8X and that evaluations of AGP8X vs. 4X on other platforms may miss the point altogether.
[/quote]
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 56 of 89ANN.lu
Posted by hammer on 01-Dec-2003 05:54 GMT
In reply to Comment 37 (Eva):
A slight edge over competition is important when HW is commoditized.
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 57 of 89ANN.lu
Posted by hammer on 01-Dec-2003 06:30 GMT
In reply to Comment 43 (Eva):
>So there is zero differences for the enduser from 1x to 8x (a part marketing) >as a lot of people can read in various article on Hardware Upgrade and
>DriverHeaven.

On GPUs with beefy on-card memory(e.g. 128MB), it’s a just insurance for future sloppy programming. Owning an ATI 9600 wouldn’t expose the need for AGP8X for awhile.

>Enstead there is a Big difference in using slow and chunky Sdram versus DDR
>(2100 too), and if some troll here is capable to conclude that Aone is a
>FASTER machine of Peggy2, is really ridicolous and must be deny.

Without at least EV6 style bus, the current FSB of PowerPC G3 and Power G4 will not support the PC2100 level of bandwidth**. **Unless you want to use that excess bandwidth for SMP or/and concurrent AGP<>Memory/CPU<>Memory functions***(IF Marvell’s NB chipset supports such a function). ***MAI claims that their chipset can also do concurrent AGP<>Memory and CPU<> Memory functions.

PS; I’m interested in anandtech style benchmarking review between the two vendors e.g. AmigaOne XE vs PegyI vs PegyII.
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 58 of 89ANN.lu
Posted by Eva on 01-Dec-2003 10:34 GMT
This are the point m8.
1) T&L calculation? M8, during the job a modern Gfx card with all possible transfer of Vertex Information, Pixelshader requests, TexuterShader requests, T&L commands and so on ... also a PcI Bus can manage it at a decent speed. If you find some articles about Ati 8500 AGP VS ati 8500 PCI you will be surprised to see a so small (5%) difference in the framerate in games like UT2003.

2) You speack abou Nforce 2 ... Let's speack so about the limit of this arguings.
Let's compare the FASTEST integrated GFX chipset for P4 (Ati 9100 IGP) with a standard Geforce 4 Gfx Mx card.
Integrated cards use Agp transfer to use Main memory (so it really heavely depends on AGP bus speed)

These are some bench that can clear the truth about the AGP speed, when it is needed and why in today market all games use only GFX memory and minimize AGP use.

*******
(A) ATI 9100 IGP (AGP 8x - 64 MB shared using AGPBUS - DualCHannel bus speed - FastWrite Enabled)
(B) Nvidia 5200 FX (AGP 8x - 64 MB on GFX board - Fast Write enabled)
(C) ATI 9000 Pro (AGP 1x - 64 MB on GFX board - Fastwrite DISABLED)

3dMark 2001 (1024*768*32bit)
(A) 4778
(B) 5400
(C) 7171 (if 4x Agp is enabled 7202)

HALO (1024*768*32bit)
(A) 28.02 fps
(B) 28.06 fps
(C) 30.01 fps (if 4x Agp is enabled 30.08 fps)

SeriousSam2 (1024*768*32bit)
(A) 30.40 fps
(B) 42.10 fps
(C) 40.12 fps (if 4x Agp is enabled 41.23 fps)
********

Some of these results can be found on http://www.rage3d.com/content/reviews/mobo/ati9100igp/

Be clear that in all tests (UT2003, CallOfDuty, 3dMark2003, Aquamark, ShaderMark) the difference in speed of Agp1x and Agp 8x is max 3%.

The really importance of an Agp bus is its reliablity.
And sure with its "NOT fully 2.0 AGP compliant" specification of Articia chip, Peggy 2 is already better.

PS
http://www6.tomshardware.com/motherboard/20000214/agp-08.html
and to see the "Huge" difference from AGP over simple PCI here you will find some benchmark of an ATI 9000 PCI (unfortunatley these tests are done on an AMD Athlon XP 2000 processor while all previous one are done on a Intel Pentium 4 2400). They show that the performance of a relatively modern card like Ati Radeon 9000 is similar if you pass form an AGP4x bus to a standard PCI bus:
http://www5b.biglobe.ne.jp/~pika2/9000pci.html

PPS
Mmm I'm starting to think that all the topic is simply a way of some red party troll to fire shit on an harware manufacted and created (not relabelled) by a good company like Genesi. really depressive.
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 59 of 89ANN.lu
Posted by Anonymous on 01-Dec-2003 12:12 GMT
In reply to Comment 53 (hammer):
>The importance of AGP transfers speed comes into play when said data exceeds>the onboard memory of the AGP card i.e. in serious Open GL tasks. Are AOne/Pegasos designed for such "serious Open GL tasks"?Would anyone buy a 600-1000MHz machine with lackluster driver support as anOGL workstation?
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 60 of 89ANN.lu
Posted by Neko on 01-Dec-2003 12:21 GMT
In reply to Comment 49 (strobe):
The whole point of hardware T&L in modern GPUs is to reduce load on the CPU
(which no longer has to calculate) and conserve bandwidth across the bus.

Quartz doesn't stress AGP any more than it would stress a PCI card - the
actual vertex upload requirements of a Quartz desktop (each window is a
polygon mesh, with a texture applied) is actually quite tiny.

Keeping those texturemaps for the window contents to allow proper and speedy
compositing is what slows Quartz down on slow machines. Quartz gets around the
slow bus and crappy read/write performance of AGP and PCI by abstracting all
the graphics operations so that they all execute solely on the card.

Under Windows, MacOS 9.x, AmigaOS, MorphOS, Linux - you can get the address of
the graphics card framebuffer and poke it manually. It's slow but useful
sometimes. Under MacOS X, you can't do it - because the display subsystem is a
bunch of beziers, meshes, matrices and composited textures. This is how it
keeps it's speed on more modern machines, all graphics operations are REQUIRED
to be done in hardware on the graphics card for performance..

No CPU fallbacks without a bitmap download/upload. That's what's slow. Not
vertex buffers.

=Neko=
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 61 of 89ANN.lu
Posted by Eva on 01-Dec-2003 14:59 GMT
In reply to Comment 60 (Neko):
Exacltly, Neko!
Actually the only true research limit in actuall Video subsystem are Bus and Run speed ON the GFX board (not on the external BUS).
Ati Point to Point, Nvidia MegaBusPump, DDR2, WFRam ... these are the future in the GFX market in the performance field, not Agp 16x or PCI-X or PCI-Expresse (that are completly different BUSes) that will be (in the case of PcI-Express) a more elant and fast way to tell to GFX board "Take those textures and execute these commands"
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 62 of 89ANN.lu
Posted by Alkis Tsapanidis on 01-Dec-2003 19:50 GMT
In reply to Comment 61 (Eva):
The problem is that there is software that pushes an ENORMOUS amount of vertices, commands and textures.
Ok, at least DOOM3 is not one of them as most of the real calculations there
are done in hardware.
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 63 of 89ANN.lu
Posted by Eva on 01-Dec-2003 21:28 GMT
In reply to Comment 62 (Alkis Tsapanidis):
The problem is that there is software that pushes an ENORMOUS amount of vertices, commands and textures.
______________

Alkis, actually I can't name a software that can strech also a PCI 66mhz bus.
Can you tell us one?
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 64 of 89ANN.lu
Posted by Alkis Tsapanidis on 01-Dec-2003 22:47 GMT
In reply to Comment 63 (Eva):
HL2 probably. SOF2 would, if they kept the original textures (even the leaves on
the trees had ENORMOUS textures).
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 65 of 89ANN.lu
Posted by Eva on 01-Dec-2003 23:13 GMT
In reply to Comment 64 (Alkis Tsapanidis):
In Reply to Comment 63:
HL2 probably. SOF2 would, if they kept the original textures (even the leaves on
the trees had ENORMOUS textures).
______________

Na, none of these games need more than a Agp1x bus.
HalfLife2 BETA is equally fast on my Ati900Pro with AGP 4x + Fastwrite and with AGP1x - Fastwrite.
For Sof2 the trend is absolutely similar. Same fps, speed.

PS
Bench of HF2 beta ANON (I put all quality to Maximum, a part FSAA+AF, just to stress the entire system, hoping to see a decent difference in the agp performance):

ATI 9000PRO (800*600*32) AGP4x
18.6 FPS

ATI 9000PRO (800*600*32) AGP1x
18.5 FPS
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 66 of 89ANN.lu
Posted by strobe on 02-Dec-2003 03:40 GMT
In reply to Comment 60 (Neko):
Neko,

When you change the "texture" of a Quartz window it is done by the CPU not the GPU and therefore has to be pushed through the bus. Any TINY animation like a small QuickTime movie COMPLETELY SATURATES THE PCI BUS as a result.

Also the point of T&L is to conserve CPU usage at the expense of bus bandwidth. How you think this conserves bus bandwidth is beyond me. With T&L you push MORE verticies through the bus, not less.
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 67 of 89ANN.lu
Posted by strobe on 02-Dec-2003 03:46 GMT
In reply to Comment 60 (Neko):
BTW Neko if you think bezier paths are rendered with the GPU with QE, you're an idiot.

I thought you were smarter than that. The GPU only takes care of composition and possibly affine transformations, not the rendering of the textures. In order for the GPU to handle anything further Apple would have to write Apple-specific OpenGL extensions for such things. There are Apple extensions, but not for rendering the contents of windows. Geez.
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 68 of 89ANN.lu
Posted by hammer on 02-Dec-2003 07:09 GMT
In reply to Comment 58 (Eva):
Games mentioned wouldn’t even stress the 64MB on-board memory i.e.
1. Halo - Combat Evolved's system requirement is "32MB 3D graphics card"
(X-BOX like footprint).
2. 3DMark 2001's system requirement is "32MB 3D graphics card".
3. Serious Sam 2's system requirement is 16MB 3D graphics card.

My old test PC delivers ~3105 Marks from ~3D Marks 2001 SE
This PC has
+Geforce2 MX (NV11) 32MB SDR**
+nForce 2
+512MB PC2100
+Athlon XP 1900+
**GPU considered to be inferior to Geforce 2 MX 400/Geforce 4 MX 4x0.
The said test PC has no trouble running these games.

As for PCI vs AGP;
The said test PC's Quake III FPS has halved to ~86 FPS compared to 162 FPS(with AGPX4). This was attained by the removal of the nForce’s chipset driver(includes AGP drivers). One should make sure that processor is the same for all of the benchmark cases.
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 69 of 89ANN.lu
Posted by hammer on 02-Dec-2003 08:27 GMT
In reply to Comment 65 (Eva):
Note that HalfLife 2(X86-32) will be available for X-BOX.

Reference;
http://www.gamespot.com/xbox/action/halflife2/
http://www.gamersstop.com/games3/Half-Life_2_xbox_game.htm

The killer performance about Half Life 2 is its use of pixel shaders. I don’t think they would be aiming for PC with 1GB memory (and similar texture data consumption).

There’s a reason why serious OpenGL card are equipped with 512MB of on-board memory i.e. the serious applications can be bloated up to use that amount of texture data. "AGP texture" is an insurance for users who can’t afford/don't have access to a 512MB of on-card memory video card.
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 70 of 89ANN.lu
Posted by hammer on 02-Dec-2003 08:50 GMT
In reply to Comment 63 (Eva):
Try with serious applications which can bloat up the texture usage beyond 512MB** and try it with a late model 64MB on-board video card. AGP8X transfers and AGP texture memory will play critical role in this insurance scheme.

PS; Don’t forget to add PC3500 3GB of main memory while you at it.
**IF texture compression schemes are in use (e.g. 20:1, 10:1, 5:1 and 'etc') adjust texture consumption accordingly.
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 71 of 89ANN.lu
Posted by hammer on 02-Dec-2003 09:12 GMT
In reply to Comment 58 (Eva):
>1) T&L calculation? M8, during the job a modern Gfx card with all possible
> transfer of Vertex Information, Pixelshader requests, TexuterShader
> requests, T&L commands and so on ... also a PcI Bus can manage it at a
> decent speed. If you find some articles about Ati 8500 AGP VS ati 8500 PCI
> you will be surprised to see a so small (5%) difference in the framerate in
> games like UT2003.
They don't use custom maps such as 2019 (detailed Bladerunner style city maps). Most of UT2003’s shown benchmarks only test with stock maps and stock settings i.e. aimed at boxed target specs**.

**UT2003 has system requirements of 16MB TNT(need) up to 32MB/64MB/128MB VRAM(Want). My test NV25 GPU only has 64MB VRAM (I made sure it was crippled when running 2019 map). The graphic setting for running this UT2003 test was on "holy shit" mode at 1024x768x32 resolution. To minimize virtual memory disk trashing 1GB of PC3200 memory was installed (about 760MB Free physical memory after WinXP login).

PS; UT2003 is even available for the lowly X-BOX (System spec that exceeds 16MB TNT requirements).
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 72 of 89ANN.lu
Posted by hammer on 02-Dec-2003 09:17 GMT
In reply to Comment 71 (hammer):
Addendum
For scope limited AGP transfer benchmarks refer to;

1. http://cwdohnal.home.mindspring.com/var_mem_bench/var_mem_bench.html
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 73 of 89ANN.lu
Posted by hammer on 02-Dec-2003 09:32 GMT
In reply to Comment 72 (hammer):
>Let's compare the FASTEST integrated GFX chipset for P4 (Ati 9100 IGP) with a >standard Geforce 4 Gfx Mx card.

Note that one could duplicate AGP8X’s transfer (~2GB/s) with Video card’s on-card memory i.e. by down clocking the VRAM’s clock speed. Then there’s another issue with AGP<>Memory/CPU<>Memory concurrency. ATI’s technology is highly based on Intel’s technology.
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 74 of 89ANN.lu
Posted by hammer on 02-Dec-2003 09:48 GMT
In reply to Comment 73 (hammer):
>PS
>http://www6.tomshardware.com/motherboard/20000214/agp-08.html

In this reference;
http://www17.tomshardware.com/motherboard/20000214/agp-07.html
it shows the impact of AGP transfer rates.
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 75 of 89ANN.lu
Posted by priest on 02-Dec-2003 10:14 GMT
and IMHO, that (this http://www17.tomshardware.com/motherboard/20000214/ ) exact article describes what kind of rough performance we ought to get on A1, Peg1 and Peg2.

With Radeon 8500 the AGP speed should matter less than on Radeon 7500.

And with either one of those cards (+600-1000Mhz PPC) we should beat those x86 machines that were used by tomshardware in that test.

And still the main point is that it takes long time before we have SW that utilizes the available performance.

A1/Peg1/Peg2 will be a giant performance leap for classic Amiga users ... for x86 lovers it just "y2k performance".
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 76 of 89ANN.lu
Posted by Anonymous on 02-Dec-2003 11:45 GMT
In reply to Comment 64 (Alkis Tsapanidis):
>HL2 probably. SOF2 would, if they kept the original textures (even the leaves>on the trees had ENORMOUS textures). What about texture compression then?
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 77 of 89ANN.lu
Posted by Anonymous on 02-Dec-2003 11:52 GMT
In reply to Comment 66 (strobe):
>Any TINY animation like a small QuickTime movie COMPLETELY SATURATES THE PCI BUS as a result. How is that? 320x240 x 16Bit x 30fps is just 4.6MB/s.Macs must have a very slow PCI(-X) bus if this saturates it already?
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 78 of 89ANN.lu
Posted by Anonymous on 02-Dec-2003 12:01 GMT
In reply to Comment 71 (hammer):
>To minimize virtual memory disk trashing 1GB of PC3200 memory was installed>(about 760MB Free physical memory after WinXP login). The OS eats 1/4 of your Ram? Man, that's just plain sick.
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 79 of 89ANN.lu
Posted by Eva on 02-Dec-2003 13:27 GMT
In reply to Comment 74 (hammer):
Hammer this is the point!
Actually there are ZERO application that use AGP transfer too much, and this just for Performance!
Also an AGP16X is slower than 2000 Generation Onboard GFX-Ram bus!
It's pathetic the try to convince people that a machine is slow because "don't have a 28x Agp bus".
All user application (Lightwave, 3dstudio, Alias) and gamez (Ut 2004, HalfLife2, doom3) will simply require a 1x AGP bus to run decently and a decent Gfx card with at east 64Mb of on board ram.
Can you ask us and yourself, why Top Gfx cards use 256 or 512 Mb of Onboard Ram enstead of their "super fast 16x Agp buses" and cheap common DDR ram on the Mainboard?
I will ask for you: for performance.
Also a 16x Agp bus is slower than a Gfx Bus on a Geforce 2 card (aka the Bus that is on the GFX board that transfer data from tha GFX chip to the Gfx memory, if I was no clear).
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 80 of 89ANN.lu
Posted by Anonymous on 02-Dec-2003 14:44 GMT
In reply to Comment 79 (Eva):
> Can you ask us and yourself, why Top Gfx cards use 256 or 512 Mb of Onboard Ram enstead of their "super fast 16x Agp buses"

You make the mistake of not lookig beyond the texture issue. Yes, for textures it is much better to use on-board RAM rather than main memory (as long as you have enough of it). But games still run on the CPU and the data for each frame still goes over the AGP bus. There you have the nasty word: AGP bus, not telepathy, is used to feed the card with data. With many 3D shooters, the bus speed does not matter so much because loading textures is a rare event and each frame does not need much bandwidth in form of 3D geometry data. But there is a life beyond current 3D shooters: Both, 2D games and future 3D games, will create a significant load on the AGP bus, especially at high resolutions and high frame rates.
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 81 of 89ANN.lu
Posted by minator on 02-Dec-2003 15:02 GMT
In reply to Comment 66 (strobe):
>When you change the "texture" of a Quartz window it is done by the CPU not the GPU
>and therefore has to be pushed through the bus. Any TINY animation like a small
>QuickTime movie COMPLETELY SATURATES THE PCI BUS as a result.

Hardly, PCI can do around 132MB / Second which is 1280 X 1024 @ 50 FPS (theoritical max).
Your CPU is more likely to run out of steam before the PCI bus gets satuted.

>Also the point of T&L is to conserve CPU usage at the expense of bus bandwidth. How
>you think this conserves bus bandwidth is beyond me. With T&L you push MORE
>verticies through the bus, not less.

But the T&L unit is onboard the Gfx card...
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 82 of 89ANN.lu
Posted by Eva on 02-Dec-2003 15:17 GMT
In reply to Comment 80 (Anonymous):
Again?
No games present or future will neeed anything better of Agp1x.
The future is first of all PXI-Express so think about the future on Agp 4x- 8x is a bit unproof.
Second: also future games like HF2, Ut2004 and Doom3 had EQUAL performance with AGP1x and AGP8x.
Simply try all beta and alpha around.
Anyway there is NO game reason to tell to a machine "You are a crappy machine" when it use a Full AGP 2.0 compliant Bus.
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 83 of 89ANN.lu
Posted by Anonymous on 02-Dec-2003 16:04 GMT
In reply to Comment 82 (Eva):
>No games present or future will neeed anything better of Agp1x

No computer present or future will need more than 640K RAM. Bill Gates.
No games present or future will neeed anything better of Agp1x. Eva.

>Second: also future games like HF2, Ut2004 and Doom3 had EQUAL performance with AGP1x and AGP8x.

Actually, if you care to look up the numbers, the step from AGPx2 to AGPx4 makes a 10% difference for some of these games (no idea what difference AGPx8 makes). Not much but not nothing: if a chipset or CPU is 10% faster, it usually gets the gold medal of the review.

Besides, I notice that you carefully pick 3D games. 3D games using hardware acceleration are unproblematic (so far). Your nice theory on the irrelevance of AGP speed comes to an end with 3D software renderers (not so uncommon on Amiga-like platforms) and HDTV video.
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 84 of 89ANN.lu
Posted by Olegil on 02-Dec-2003 16:39 GMT
In reply to Comment 32 (Anonymous):
The reason why new cards "require 4x or 8x" to work is that this is AGP3.0, which specifies 0.8V signaling (you need this to get 8x), thus these cards WON'T work in an AGP1.0 (1x-2x) or AGP2.0 (1x-4x) slot.

That's the point.
At least some Radeons support 1-8x, which means they have the ability to work from 0.8 to 2.5V or something (major headache, so I'm not entirely certain right now, could be lower). It's actually the standard version number, NOT the speed, that is important here.

AGP at 1x speed CAN be AGP2.0. But it cannot be AGP3.0, because AGP3.0 specifies signaling levels so low that they cut the 1x and 2x support. This obviously doesn't mean some people decided to implement it BETTER, but the standard only speficifies 4x and 8x... Because of the low voltage ;-)
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 85 of 89ANN.lu
Posted by hammer on 02-Dec-2003 20:20 GMT
In reply to Comment 79 (Eva):
>Hammer this is the point!
>Actually there are ZERO application that use AGP transfer too much, and this >just for Performance!

Note that I stated it’s was an "insurance" for future when one exceeds the amount of VRAM. One could effectively turn off AGP Aperture from BIOS IF you want. Late Windows releases handles VRAM and AGP texture memory similar to waterfall effect i.e. it will use VRAM first before it use AGP texture memory. With a decent GFX card, most titles don’t trigger this insurance.

>Also an AGP16X is slower than 2000 Generation Onboard GFX-Ram bus!
>It's pathetic the try to convince people that a machine is slow >because "don't have a 28x Agp bus".

I didn't state it was not useable. It would be unwise to claim a machine with AGP1X DirectX8 GPU will be able to beat a machine with AGP8X DirectX8 GPU for ALL of the cases.

>All user application (Lightwave, 3dstudio, Alias) and gamez (Ut 2004, >HalfLife2, doom3) will simply require a 1x AGP bus to run decently and a >decent Gfx card with at east 64Mb of on board ram.

I didn't state it was not useable. 64MB of VRAM wouldn’t be enough for serious OpenGL usage(unless you dabble with G-MAX’s texture footprint). Having the insurance of AGP8X and AGP texture memory is better than nothing/little at all. IF the game is designed for X-BOX in mind, it’s more than likely it will run on PCs that equal or exceed it.
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 86 of 89ANN.lu
Posted by hammer on 02-Dec-2003 20:39 GMT
In reply to Comment 78 (Anonymous):
That value includes "System cache". Also PVR/DVR, Internet Firewall, Internet Gateway, LAN Bridge, Creative Media Source(Audigy2 ZS 7.1), Norton Antivirus 2003 Pro, Windows Messenger 5.0, and etc services are loaded during boot time. I can tweak WinXP for 128MB/256MB machines IF required.
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 87 of 89ANN.lu
Posted by hammer on 02-Dec-2003 20:55 GMT
In reply to Comment 82 (Eva):
>The future is first of all PXI-Express so think about the future on Agp 4x- >8x is a bit unproof.

My "Future proofing" is in relation to the sustain performance through the useful life of the consumable asset. My purchasing of PCI-Express motherboards(and 64bit processors) is scheduled for next year.

>Anyway there is NO game reason to tell to a machine "You are a crappy >machine" when it use a Full AGP 2.0 compliant Bus.

Its insurance capability would be less desirable. Note that I didn’t state PegyII was a "crap" machine. "AGP 2.0" refers to voltage and power requirements.
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 88 of 89ANN.lu
Posted by hammer on 02-Dec-2003 22:01 GMT
In reply to Comment 79 (Eva):
>Also a 16x Agp bus is slower than a Gfx Bus on a Geforce 2 card (aka the Bus
>that is on the GFX board that transfer data from tha GFX chip to the Gfx
>memory, if I was no clear).

Which Geforce 2(NV11 or NV15)?

Geforce2 MX(NV11)(depending on the variant).
183Mhz SDR x 128bit/8 = 2928MB/s
166Mhz SDR x 128bit/8 = 2656MB/s
140Mhz SDR x 128bit/8 = 2240MB/s

AGP16X = ~4224 MB/s
AGP8X = ~2112 MB/s

I don't think NV11's FSB could beat AGP16X.

This particular NV11 250Mhz SDR x 128bit/8 = 4000MB/s setup could rival it.

With nForce2, a single memory controller (CH) can be hog tied to AGP transactions and while independent of the processor’s transactions e.g.
CH1, PC3200 (3.2GB/s) SPP (2.1GB/s) GPU.
CH2, PC3200 (3.2GB/s) SPP (FSB) CPU.

AGP transactions would be unhindered by the CPU transactions.
CH1 has enough bandwidth left over for MCP-T(APU)(800MB/s) SPP link.

Nforce2’s dual channel controller doesn’t form a straight 128bit "uberbus" like Intel’s core logic technology (includes Intel’s Extreme Graphics IGP i.e. which "interlaces" with CPU and GPU transactions).

At the moment I’m attempting to obtain ATI’s core logic architecture information (I’m certain it’s based on Intel’s core logic technology).

Note that ATI's 9100 may have Hyper-Z and better texture compression regime than any of NV11, NV15 and NV18 (flaws on the 1st generation NV1x GPU families). The product that matches ATI’s IGP 9100 in terms DirectX 8(just 3D side) feature set is the nForce chipset within the X-BOX i.e. NV2A (the second generation IGP family).
Pegasos 2 AGP support ? [Was Pegasos2 sub standard hardware?] : Comment 89 of 89ANN.lu
Posted by hammer on 02-Dec-2003 22:31 GMT
In reply to Comment 88 (hammer):
Addendum;

With VIA and ATI based solutions, the VIA 4in1 driver changes the default size of the AGP aperture to 32MB i.e. problems with 1GB or more of system memory and an ATI graphics card with AGP aperture has been set above 32MB.

Reference
http://www.theinquirer.net/?article=12951
http://www.theinquirer.net/?article=12968
Anonymous, there are 89 items in your selection (but only 39 shown due to limitation) [1 - 50] [51 - 89]
Back to Top