Graphics card performance test?

[wip3out]

Am I correct in assuming that 3D Mark 2001 deducts points off the score if a graphics card doesn't support one of it's tests?? The test I am referring to is "Nature" which my Geforce 2 Ultra doesn't support. What otherr software can I use to Benchmark my card with. I'm trying to beat my freinds Geforce 3 Ti500 with my Geforce 2 Ultra w/ OC.

any software idea's.

[odog]

maybe a pure geometry based test. (flat shaded 3dsmax/ lightwave etc etc etc)

the geforce2u has a higher clock rate then the GF3 so it should beat it in pure T&L calculations. anything reguiring rasterizing will get smoked by the GF3... and anything that needs pixel and or vertex shading won't run very well (one of those can be emulated by the cpu the other can't don't remember which)

[odog]

yeah you are correct, 3dmark20001 deducts point for every test you can't due in hardware...(ie the bump mapping test and the nature test)

[Murman]

So if you run 3Dmark2001 and uncheck nature in selected tests it will bump the score if your card dosn't support game 4 - nature?

[wip3out]

no it will just deduct automatically.

[MartialArtist12]

Not exactly odog. If this was true, than the Intel P4 would kill an AMD Athlon XP. The GF3 Ti500 has a more advanced core with more advanced memory. Even if it should beat it in pure T&L calculations, the GF2 is without many speed and performance features as the GF3.

[Brent]

uncheck the tests the gf2 won't run, adn then also uncheck those SAME tests for the GF3 and then benchmark them that way

you will then see how they directly compare

in which the GF3 Ti 500 will win

There is no way to beat his score without actually getting a GF3 yourself, the GF3 has newer technologies that the GF2 doesn't have. They are 2 seperate generation of Cards. The GF3 is NV20 technology, the GF2 is NV15 technology.

There are lots of other benchmarks besides 3DMark though

Actually 3DMark2000 might be a better one for ya'll to compare with since it supports all features in the GF2. However his GF3 will still beat it.

Games are the best benchmark. I.E. Quake3, Dronez, Aquanox, Max Payne, RTCW, GLMark, Serious Sam 1, Serious Sam 2, and lots and lots of others.....

[odog]

MartialArtist12: i'm pretty sure that the GF2 and GF3 have identical T&L units(not exactly sure) the GF3 just brought pixel and vertex shaders along with other memory tweaks while leaving the T&L unchanged(again not exactly sure)

using that as a baseline you can extrapolate that with the GF2u clocked higher it would perform specific T&L dependent programs faster....


brent would you be so kind as to verify, or shoot down my little theory

[Brent]

Originally posted by odog
MartialArtist12: i'm pretty sure that the GF2 and GF3 have identical T&L units(not exactly sure) the GF3 just brought pixel and vertex shaders along with other memory tweaks while leaving the T&L unchanged(again not exactly sure)

using that as a baseline you can extrapolate that with the GF2u clocked higher it would perform specific T&L dependent programs faster....


brent would you be so kind as to verify, or shoot down my little theory

GF1 = original T&L unit

GF2 = Second Generation T&L unit, improved performance in vertices

GF3 = 3rd Generation T&L unit, new engine dubbed nFinite FX engine, basically it's the GF2 T&L unit re-programmed to take full advantage of DX8 and to accelerate hardware programable vertex and pixel shaders

[odog]

if the GF3 is just the GF2 torked for DX8 and shaders..... aren't the shaders a rasterization function, and completely independent of the T&L unit? if isolating the T&L units so that no rasterzation, can skew the benchmark. come to think of it..... vertexes are T&L.....


nevermind.

[Brent]

Originally posted by odog
if the GF3 is just the GF2 torked for DX8 and shaders..... aren't the shaders a rasterization function, and completely independent of the T&L unit? if isolating the T&L units so that no rasterzation, can skew the benchmark. come to think of it..... vertexes are T&L.....


nevermind.

yeah the GF3's fixed function T&L unit is run as a Vertex program for backwards compatibility with GF2 class cards, now since the GF4 has 2 vertex shader pipelines it should accelerate old GF2 T&L VERY well

[MartialArtist12]

Yeah, but T&L isn't everything. We've seen the Kyro II. Although it got killed by the GF3 and the Radeon 64MB DDR VIVO in many applications, the Kyro II did really impressive with its features and going with a smart way instead of just pure MHz and pure speed and power.

[Brent]

Originally posted by MartialArtist12
Yeah, but T&L isn't everything. We've seen the Kyro II. Although it got killed by the GF3 and the Radeon 64MB DDR VIVO in many applications, the Kyro II did really impressive with its features and going with a smart way instead of just pure MHz and pure speed and power.

yep, a true Deffered rending card using Tile Based Rending as their method

VERY Impressive

that's why i HIGHLY look forward to the Kyro 3, imagine all that but with T&L and higher core/memory clocks and DDR and a tweaked TBR engine, w0w0w0w

the Radeon 8500 and GF3/GF4 use a 'defered rending type' kind of a trick, called Occlusion Culling, it works QUITE well in not rendering what you don't see, but alas they are still immediate mode renderers at heart...

I believe NVIDIA has expressed though their interest in a deffered card, though prolly many years away, I think their next move is to use some Gigapixel/3dfx tech which will be neat

ATI i haven't heard what their take is on deffered rendering, but I would think it's something they are prolly looking into as well.

It's the future IMO, something graphics card makers really need to look into and figure out how to to it well like PowerVR!