agp speed among other things...

[ple3003]

I recently made two benchmarks usind 3dMark on a duron 600, 256mb pc133 and ATI Radeon 8500. One with agp 4x enabled, and one with agp 4x disabled - guess that means 2x or 1x. Apparently it made no diffrence, cause i scored 4100 both times. That is good, as the A1 only use agp 1x. Not as good as my old XP1600+, with 512mb ddr266 together with that ATI Radeon of mine. Scored 9000 back then. guess that the CPU speed and/or memory speed. Certainly hope it wasn't the ddr though, cause the A1 is equipped with sdr. what do you think about this? Is the A1 mobo going to utilize all that power of the ATI Radeon 8500 to it's full potential?

[CodeSmith]

ple3003 wrote:
I recently made two benchmarks usind 3dMark on a duron 600, 256mb pc133 and ATI Radeon 8500. One with agp 4x enabled, and one with agp 4x disabled - guess that means 2x or 1x. Apparently it made no diffrence, cause i scored 4100 both times. That is good, as the A1 only use agp 1x. Not as good as my old XP1600+, with 512mb ddr266 together with that ATI Radeon of mine. Scored 9000 back then. guess that the CPU speed and/or memory speed. Certainly hope it wasn't the ddr though, cause the A1 is equipped with sdr. what do you think about this? Is the A1 mobo going to utilize all that power of the ATI Radeon 8500 to it's full potential?

Nothing uses AGPx1 any more - not even the AmigaOne (AGPx2).  When you disabled AGPx4, you set the AGP bus to AGPx2.  The reason you didn't see any difference is because AGP speed only makes a real difference for either huge scenes with millions of polygons, or textures that are too big to fit in the graphics card's internal RAM.  The nice thing about AGP is that if you run out of GFX RAM, you can use some of your main memory, but that's so slow compared to VRAM that the speed of the bus makes little difference.  The fact that the board has DDR and the mobo SDR does not impact anything, since the two sets of memory only communicate through the AGP bus (that's why you can use the same gfx card regardless of whether you have SDR, DDR, RAMBUS, whatever...)

[ple3003]

So i guess the CPU had some impact on my 3dMark scores right? Does anyone know how a g3 600 Mhz / g4 800 Mhz (sdr memory) compares to XP/P4 system using sdr/ddr?
I suspect numerous mac g3/g4 vs XP/P4 has been made, but do not know the outcome.

[CodeSmith]

3DMark is mostly meant to measure how good your PC is for playing 3D games (it's made by Remedy, a games company and ex-demo crew), so the speed of your CPU will affect scores.  Most of my current CPU knowledge comes from the PC world, unfortunately (back when I last looked at these things on the Amiga, a 68040 was the high end), so I can't tell you how this maps to PCs.  I've heard claims that an 2x MHz P3 = x MHz G4.  I suspect it's a bit less than that, but I'd be surprised if a G4 were slower than a P4 running at the same clock speed.  It's hard to compare the two because Motorola goes for lots of shallow pipelines, whereas Intel likes using just a few very deep ones.  I guess the only reliable benchmark in such a case would be to compare the same (modern) program on x86 and Mac, for example Quake3 framerates for x86 vs Mac would be a good reference (is there Quake3 for the mac?)

[strobe]

bus speed makes a difference when:

Pushing huge-ass textures to graphics memory from main memory. Bump mapping and cube mapping increase this. old-ass games like Quake1,2,3 do not.

Performing T+L with the GPU requires all the verticies to be pushed to the GPU which eats up bus bandwidth. Again, old-ass games like Quake 1,2,3 do not.

Bus limitations will be a problem only if and when you can run new games which use such features on your old-ass A1/Pegasos hardware. Bump mapping and T+L have been around for ages BTW, but there is a long game lag for some reason.

[Kronos]

@CodeSmith

Nothing uses AGPx1 any more - not even the AmigaOne (AGPx2).

Allmost correct, but just use the 1st PCI-solt on the Teron/A1 and
your down to a lovely AGPx1.

They don't tell you this, but DirectX automatically sets AGP speed to 2x for AMD processors because older athlon motherboards weren't stable with 4x.  You get the same results because your running at 2x no matter what.  Guess they still havn't bothered to fix it.

[Panthro]

Allmost correct, but just use the 1st PCI-solt on the Teron/A1 and
your down to a lovely AGPx1.

let me guess your about to say your wonderful
peg/Mos board has another hardware hack to fix this... :-?

[Kronos]

@Panthro

The Articia-chip has 2 busses, one for PCI (3 slots) and one for
AGP/PCI-66. The Teron boards have an AGP and PCI slot connected
to the 2nd bus, and whenever you use that PCI slot the AGP is down
to 1x.

The Pegasos has only 3 PCI slots on the 1st and one AGP on the
2nd bus, making that problem non-existent.

[Panthro]

LOL...... I was "Allmost correct" :-P

[KrasH]

ple3003 wrote:
So i guess the CPU had some impact on my 3dMark scores right? Does anyone know how a g3 600 Mhz / g4 800 Mhz (sdr memory) compares to XP/P4 system using sdr/ddr?
I suspect numerous mac g3/g4 vs XP/P4 has been made, but do not know the outcome.

I am a fan of 3Dmarkse. I use on every PC box and gfx card I get my hands on. Some examples of what I get.

Athlon 1.33Ghz (4x AGP) with GeForce 4 Ti4400 (128Mb DDR on GFX card), get 8600 3D marks (was only 8100 3D marks, then I got the latest detonators). The athlon uses PC133 RAM on the mainboard.
The same Athlon with GeForce 2 MX400 (32Mb on card), get only 2800 3D marks or so.
PIII 667 (2x AGP)with the same GeForce 2 MX400, get 2500 3D marks or so.
PIII 667 (2x AGP) with GeForce 256 (32 Mb on card), get only 2000 3D marks.

So basically Both CPU and Graphics card play a part in the performance. One can get bogged down by the other. In terms of AGP bus speeds, they are theoretical. eg. 8x bus cards and mainboards don't go twice as fast as a 4x agp bus card and mainboard. In some PC mag reviews its more like a 15% gain in speed on the card front. (which is why the Radeon 9700 is the fastest GFX card on the PC market atm).

I wouldn't mind having a Radeon 9700 and 3.04Ghz P4 setup (although I hate Intel and the Pentiums. Much prefer Athlons. Cheaper and better bang for bucks, and they use some RISC architecture). A recent PC User Australia mag got a 15000+ 3D mark with one of those.

[Elektro]

It's true that the Teron board is somewhat of a clunky design. The 32bit 66 MHz pci slot is mostly useless seing there's not a lot of those cards around and if you can find one it's more expensive. AMR slot too. Basically useless unless you buy the card from ole-egil. No onboard sound on the 600 version is kinda lame too.

What Eyetech should do is buy chips from Mai then go to a company like ECS to design them a board. Seing that all boards these days are basically the same the price difference really shouldn't be that big. Or call Dave Haynie.

It looks as though you can overclock it at least. Can you do that with the pegasos?

[Bodie]

I did a similar test on 3dmark with my wintel box.  

[mikeymike]

@krash

your sig... are people likely to care about the specs of all your PCs?  overkill, perhaps?

[Minion]

Kronos wrote:
@Panthro

The Articia-chip has 2 busses, one for PCI (3 slots) and one for
AGP/PCI-66. The Teron boards have an AGP and PCI slot connected
to the 2nd bus, and whenever you use that PCI slot the AGP is down
to 1x.

The Pegasos has only 3 PCI slots on the 1st and one AGP on the
2nd bus, making that problem non-existent.

Yeah I've heard that the A1 has a built in hardware fix for the AGP 1x/PCI thing.
Its called not using that PCI slot.  There - easy isn't it.