NV30 Doom 3 alpha scores

[Cotita]

Digit-life has updated their GeforceFx review and included Doom3 alpha scores.

The Geforce Fx outperforms the Radeon9700 by 16% in 800x600 and 35%!!! at 1024x768.

Before we jump into any conclusions we have to take into account that doom3 is in alpha stage and that Carmack himself has said that the Radeon 9700 and geforceFx perform very much the same with their current build.

Still, it looks impressive.

[jAkUp]

yes it isnt really a good idea to compare an alpha game... but the link is here if anyone wants to see early early benchies
http://www.digit-life.com/articles2/gffx/gffx-ref-p9.html

[jbirney]

you really must be joking right?

Which out of the 5 rendering paths did the nV30 use? What path did the R300 take? Did you happen to really read the .plan file by JC last month?

[Nv40]

Digit-life has updated their GeforceFx review and included Doom3 alpha scores.

i cant find anything about updates ,
didnt these benchmarks are the same posted
weeks ago ?

or digitlife is using newer ATI/Nvidia drivers?

Originally posted by jbirney
you really must be joking right?

Which out of the 5 rendering paths did the nV30 use? What path did the R300 take? Did you happen to really read the .plan file by JC last month?

well it looks like Digit life is using the
standarGL COnfig.. in both the NV30 and the R300..

so can we say that both are using the maximun
precision of both cards? 96bits and 128bits?

the only card using propietary paths
according with Digitlife in that benchmarks, is the Geforce4 ,for obvious reasons. it doesnt have Fp color precision ,like any other DIrectx8 card.

i know the game test are in Doom3 alpha ,but i think
it is still good info , and a no surprising one ,
considering the Geforcefx fillrate ,and how consistent
its in raw performance , when no AA/AF is used .
something that i dont think we would see to its
maximun levels in current generation video cards
in Doom3 ,because of Performance issues.

[Bigus Dickus]

I seriously doubt both are running on the standard ARB2 path. J.C. has stated that in the latest build the R300 is roughly twice as fast as the NV30 on that path. I doubt an alpha would have that large of a relative difference between cards wrt the later version, even if the magnitude of performance for both cards improved by that factor for later builds.

I'm pretty sure the NV30 is running on an early NV30 or perhaps even the NV25 codepath. There's no telling. The R300 could be running on ARB1, or R200, or whatever as well. Those results are pretty worthless.

[Zeno]

If this build of DoomIII is from E3 2002, I'm guessing that the GeforceFX is using the NV20 codepath and the 9700 is using the R200 codepath. I.e. vendor-specific extensions for the previous generation of cards.

The reason I think this is because the OpenGL extensions for ARB_vertex_program and ARB_fragment_program weren't even approved until June 18 and September 18 2002 respectively, and they weren't exposed in the drivers for the cards until some time later.

If I'm right, this would mean that neither card is using the power of it's high-precision internal rendering, and both are doing all the extra passes that the previous cards would have needed to do.

So take these numbers with a heaping helping of salt. They may be totally different from the way the cards would stack up with a newer build of the engine that has specialized code paths for the new cards. If the old cards required n passes to draw the scene but the new cards require only 1, that would mean that a specialized codepath for the new cards would be doing 1/n times the geometry and blending work but n times as much pixel shader work on that one pass (and at a higher precision).

Another factor that could shift performance is the possibility that JC will use some of the extra power of the new cards to add new features or better image quality (which I'm hoping for).

-- Zeno

[silence]

i don't see why you insist on which path did nv30 use?

it prolly isn't ARB2, but i really don't care if r300 and nv30 use different paths if they can do their work. from standard point of view of any gamer out there there is no importance whatsoever about paths or anything.......they are only interested in HOW FAST does certain card perform (me included).

and if i remember correctlly JC said that nv30 is twice as fast as r300 when it uses it's path.so we will prolly see even better results when the game and nv30 hit the shelves.

do you honestly think that any reviewer and specially any highely hyped gamer will even bother to think about paths??.......they'll see results and as long as nv30 can produce results it will be good card.

[Kruno]

Originally posted by Cotita
Digit-life has updated their GeforceFx review and included Doom3 alpha scores.

The Geforce Fx outperforms the Radeon9700 by 16% in 800x600 and 35%!!! at 1024x768.

Before we jump into any conclusions we have to take into account that doom3 is in alpha stage and that Carmack himself has said that the Radeon 9700 and geforceFx perform very much the same with their current build.

Still, it looks impressive.

My former NV20 runs 50% better than both cards. ;)

[zakelwe]

nvidia have been showing a graph of similar results for Doom 3 for months now. I believe it was 1280 x 1024 at 2AA.

As Ut2003 also shows a big improvement at 1600x1200 for the FX over the 9700 Pro with no AA I would suggest that yes it is an indication of raw fillrate for these latest games. This sort of goes with what nvidia have been saying about the FX being for the latest or future games with this strange 8x1 / 4x2 flexible architecture.

I think people are forgetting that the FX does outperform the 9700 Pro at most non AA/AF resolutions, just not by much and about to be eclipsed by the R350.

Maybe it will be good for Doom and future games, but the nv35 will be better for those games of course. I can't see the point in the nv30 now, it missed it's window of opportunity ( to quote a terrible Dilbertesque phrase ).

[Zeno]

Originally posted by silence
i don't see why you insist on which path did nv30 use?

I realize that the average person doesn't care which code path the card takes.

I was explaining it to the people who may read this forum to support my argument that the performance of an alpha build of this engine may have no bearing on which card turns out to be better for the game once it's done.

The code that is doing the rendering on these cards in the alpha build may not even be touched when the final game is actually released. It would be like running a Quake 3 benchmark, seeing that card A is faster than card B, and then saying "See, card A is faster than card B at DoomIII !". There isn't enough correlation to support that conclusion.

-- Zeno

[silence]

okie......to clarify my post.

IMO.....without any nv30's out there and without doomIII out there we can only disscuss possibilities.....we have no clue HOW it WILL perform.

my point is that when the game and card is available no1 will ask about the paths used to give them better feeling while playing and IF nv30 can get more fps in it's own path ppl will see nv30 as better card........that's all.

btw.....i'll repeat that part about JC saying that nv30 is twice faster when using it's own path :D :angel:

[Hellbinder]

Give me a break peopple... :rolleyes:

1. There is no way the Nv30 is running enything but the Nv20 build in that alpha leak. Becuas thats the only Nvidia path in there at that point. meaning 12bit Fixed Function Vertex processing and PS 1.1 support. Vs, ATi running in FP 24 mode, and high percision.

2. I happen to know for a fact that Several of the key R300 fetures were not implimented into Doom-III until a couple months ago. Like For instance Doubble sided Stencil support.

3. I have been told that the current build of doom-III is VASTLY more complex than the E3 Alpha. In some scenes it actually has *up to* 100x the Overdraw of Quake 3. Due to various lighting effects being used.

This is another example of sub-standard Reviewers offering misleading information to the public. Neither they nor anyone else has ANY business posting benchmarks from a hacked together unfinished ALPHA verion of Doom-III.

Not that it suprises me looking at some of the other things they state in that *review*.

[silence]

i can see the future......in 5 years nv30 and r300 won't have enough power to load your desktop wallpaper:D ,so all this fuss is really pointless:cool:

[Hellbinder]

Silence..

and if i remember correctlly JC said that nv30 is twice as fast as r300 when it uses it's path.so we will prolly see even better results when the game and nv30 hit the shelves.

Then i guess you need to learn to ******* Read then eh???

The Nv30 is 1/2 the speed of the R300 when it runs ARB2 like the R300 does. It *GAINS THE SPEED BACK* when it runs its dedicated path, which i happen to know is only 12bit Fixed Function for vertex processing and 64bit floating point color. Vs 96bit floating point color, and 128bit Vertex processing on the R300.

Meaning that in the end, the Nv30 is and i quote *SLIGHTLY* faster than the R300 in most cases, with the R300 being faster in some. With the Nv30 running in a lower Precision accross the board. AND having the added benefit of 175mhz core advantage. That is BEFORE we even take AA or AF into the equation.

Where you even came up with the idea that the Nv30 is Twice as Fast is beyond me...

You guys need to do yourselves a favor and forget about these Alpha numbers now. Otherwise you are going to be really dissapointed again when the final game comes out and the GFFX is not 10-15 FPS faster. They are literally neck and neck. With any AA turned on at all the GFFX gets Trounced.

You can Quote me on that last statement if you want.

[silence]

Originally posted by Hellbinder
Silence..

Then i guess you need to learn to ******* Read then eh???

The Nv30 is 1/2 the speed of the R300 when it runs ARB2 like the R300 does. It *GAINS THE SPEED BACK* when it runs its dedicated path, which i happen to know is only 12bit Fixed Function for vertex processing and 64bit floating point color. Vs 96bit floating point color, and 128bit Vertex processing on the R300.

Meaning that in the end, the Nv30 is and i quote *SLIGHTLY* faster than the R300 in most cases, with the R300 being faster in some. With the Nv30 running in a lower Precision accross the board. AND having the added benefit of 175mhz core advantage. That is BEFORE we even take AA or AF into the equation.

Where you even came up with the idea that the Nv30 is Twice as Fast is beyond me...

You guys need to do yourselves a favor and forget about these Alpha numbers now. Otherwise you are going to be really dissapointed again when the final game comes out and the GFFX is not 10-15 FPS faster. They are literally neck and neck. With any AA turned on at all the GFFX gets Trounced.

You can Quote me on that last statement if you want.


hey........i knew about ARB2 and got that right when i read what JC wrote, but if i understood incorectlly about nv30 using it's own path i appologize and thnaks for update.
i really thought that nv30 was twice faster when using it's own path......:o :o

no hard feelings?:angel:

[zakelwe]

I can understand Hellbinders comments and agree with them, but of course nvidia have quoted these figures themselves, and they must have had JC's approval I guess.

Of course that stirs up yet more of a hornets nest again as the game is not to be released for X months from the present and nobody knows what the performance will be. I guess nv35 will be out and therefore FX performance will only be relevant to the people who bought it.

So a moot point indeed ?

[StealthHawk]

Originally posted by silence
hey........i knew about ARB2 and got that right when i read what JC wrote, but if i understood incorectlly about nv30 using it's own path i appologize and thnaks for update.
i really thought that nv30 was twice faster when using it's own path......:o :o
NV30 path makes NV30 perform twice as well as NV30 using the ARB path i think. or something along those lines.

[Chalnoth]

If the PS 2.0 results that we've seen coincide with the ARB2 results, then the performance is all over the board, and is definitely low compared to what the hardware should be capable of. I seriously doubt it's a 16-bit vs. 32-bit float issue (at least, not all of it).

And Hellbinder, the NV30 OpenGL path doesn't "just" support INT12 and half-floats. Otherwise there'd be no GL_NV_float_buffer extension.

Regardless, it still remains to be seen just how much quality is lost by going for FP16. I really don't think it will be noticeable for the vast majority of operations.

[swanlee]

This is just so hypocritical of anyone to defend these bench marks.
All the fuss about 3dmark not being accurate and yet a pre-alpha version of a game that won't be out for months is now being used as a bench? From JC's updates the code has changed so much since then these bench's are completely useless.

[Kruno]

Originally posted by silence
hey........i knew about ARB2 and got that right when i read what JC wrote, but if i understood incorectlly about nv30 using it's own path i appologize and thnaks for update.
i really thought that nv30 was twice faster when using it's own path......:o :o

no hard feelings?:angel:

He replied to you in a rude manner and you thank him? I would tell him to **** off.

If he can't accept the fact people make mistakes then he should piss off.

[randsom]

ya know, this whole thing of which card will run d3 faster, is fairly pointless, if the game was coming in may this would be a diiferent story, but the games been pushed back to late oct/early nov(which means a christmas release), there are much faster cards on the horizon.i dont how many of you are hardware junkies like me, but ill bet that by oct/nov neither of these cards will be in my box.

[kyleb]

Originally posted by K.I.L.E.R
He replied to you in a rude manner and you thank him? I would tell him to **** off.

If he can't accept the fact people make mistakes then he should piss off.


well being rude is obviously not the best way about things but spreading false information isn't to cool either; i imagine silence understands this and that is why he simply thanked Hellbinder and didn't tell him to **** off. at least thats how i would have played it were i in his shoes. :angel2:

[John Reynolds]

Originally posted by Chalnoth
Regardless, it still remains to be seen just how much quality is lost by going for FP16. I really don't think it will be noticeable for the vast majority of operations.

True, but such scores should never be compared against each other. Completely unfair benchmarking.

[Chalnoth]

Originally posted by John Reynolds
True, but such scores should never be compared against each other. Completely unfair benchmarking.
No more unfair than benchmarking 24-bit vs. 32-bit.

[kyleb]

actually it is ~8% more unfair in a strictly bit-to-bit analytical sence if you want to get into it Chalnoth. also the visual differences between the 3 levels of precision is something that deserves investigation when such claims are made. unfortunately none of us "regular" people has been given the chance to see then nv30 do fp32 and fp16 and we don't know how much int12 is going on either so then whole discussion is purely academic at this point.