Xbox's NV2A, which of these features does support?

so, we can assume that almost the entire feature set of the nv25 is indeed present in nv2a? right? But turning back to antialiasing, which was, the the most commonly used type in xbox games? i beleived it was the quincunx, since the 2x isn't worth the performance loss (on regular crt's the benefit would be barely noticeable), and the 4x and 4xS were way too heavy, to be used in a relatively complex engine.
 
btw: prior to GF3, GeForce FSAA options were limited to supersampling only. It always felt to me like the options were mainly a reaction to Voodoo5's highly-publicized FSAA because the GF1/2 modes were buggy and very slow.

so, we can assume that almost the entire feature set of the nv25 is indeed present in nv2a? right?
Yeah I think you can be sure that "almost the entire feature set of NV25 is present in NV2A". GF3 also has "almost the entire feature set of NV25". NV2A is certainly more advanced than NV15/17, and it's less advanced than NV30.
 
Last edited by a moderator:
btw: prior to GF3, GeForce FSAA options were limited to supersampling only. It always felt to me like the options were mainly a reaction to Voodoo5's highly-publicized FSAA because the GF1/2 modes were buggy and very slow.


Yeah I think you can be sure that "almost the entire feature set of NV25 is present in NV2A". GF3 also has "almost the entire feature set of NV25". NV2A is certainly more advanced than NV15/17, and it's less advanced than NV30.

so in which way the nv25 is an evolution of nv2a? if nv2a has LMA2 (maybe minus the auto pre charge feature), and the accuview antialiasing, and has even the bonus of some functions not exposed in dx 8 (like signed stencil buffers), i just can't see any real sign evolution, just more raw power thanks to increased clockspeeds and quantity of ram.
 
NV2A is certainly more advanced than NV15/17, and it's less advanced than NV30.


Btw, there were some indications (at least from Dave Baumann) that the NV2A had double pumped Z rates and an ALU config similar to the NV30.
 
Btw, there were some indications (at least from Dave Baumann) that the NV2A had double pumped Z rates and an ALU config similar to the NV30.
Hmm......

so in which way the nv25 is an evolution of nv2a? if nv2a has LMA2 (maybe minus the auto pre charge feature), and the accuview antialiasing, and has even the bonus of some functions not exposed in dx 8 (like signed stencil buffers), i just can't see any real sign evolution, just more raw power thanks to increased clockspeeds and quantity of ram.
I think you should ask yourself if those marketing terms are worth judging the hardware by in the first place. Look at the games as they are all that matter.

Going by the games, I think a few things are fairly obvious.
-NV2A capabilities are practically very similar to GF4. (shader flexibility)
-Xbox didn't have enough RAM to leverage its capabilities fully (Doom3/HL2/FarCry conversions and DXIW)
-Xbox didn't have enough memory bandwidth and/or memory capacity to use AA without serious compromises
 
Last edited by a moderator:
Btw, there were some indications (at least from Dave Baumann) that the NV2A had double pumped Z rates and an ALU config similar to the NV30.

yeah, i also read that nv2a has verious features similar to nv30, other than these 2 , and the support to signed stencil buffers, do you know what are the others?


@ swayee

behind those marketing terms, there are actual technical features, so they are important in judging the hardware imho. What i wanted to understand, is how the geforce 4 evolved from nv2a, and since the better antialiasing (accuview) and better memory management (LMA 2), wich are the most important changes in nv25 cores, are supported in both chips, i see geforce 4 as nothing more than a overclocked nv2a.
 
I suppose it is interesting from a "dot the I's and cross the T's" sort of documenting history thing. :) The games are what matter in the end though. Xbox is effectively a RAM & CPU starved 2001 PC with an underclocked, RAM bandwidth and capacity-starved GeForce 3/4.

This is a topic that has been hit on a lot. I remember reading them too, but this stuff drains from my brain after years. More threads like yours: (lots to digest in these)
http://forum.beyond3d.com/showthread.php?t=2540&highlight=nv2a
http://forum.beyond3d.com/showthread.php?t=2168&highlight=nv2a
http://forum.beyond3d.com/showthread.php?t=2169&highlight=nv2a
 
Last edited by a moderator:
I suppose it is interesting from a "dot the I's and cross the T's" sort of documenting history thing. :) The games are what matter in the end though. Xbox is effectively a RAM & CPU starved 2001 PC with an underclocked, RAM bandwidth and capacity-starved GeForce 3/4.

This is a topic that has been hit on a lot. I remember reading them too, but this stuff drains from my brain after years. More threads like yours: (lots to digest in these)
http://forum.beyond3d.com/showthread.php?t=2540&highlight=nv2a
http://forum.beyond3d.com/showthread.php?t=2168&highlight=nv2a
http://forum.beyond3d.com/showthread.php?t=2169&highlight=nv2a

don't understand that part sorry :LOL:, i don't see it like that, the cpu hadn't to compute much of the floating point calculations so it was ok, for what it had to do. The gpu, was ram and bandwidth starved, as any other consolle gpu more or less, ram is just never enough for 3d gaming. Consolles compensate with higher level of optimization possible, thanks to the closed hardware, and more efficient design (no drivers, no overhead, no slow busses between the chips). And sincerely, i don't think that a normal geforce 3, would pull off graphics like the best i saw on the xbox, nor the geforce 4 would vastly outperform the best games i have seen on xbox. Or at least the benchmarks i saw online inuge me to think so.
 
Keep in mind that devs did have lower level access to NV2A than the DirectX spec at the time. "The best" on Xbox might have exploited some of those whereas the API on PC would not have had those features exposed at all.
 
don't understand that part sorry :LOL:,
sorry it's a dumb saying we have :). It basically means to be very thorough and leave no loose ends.

i don't see it like that, the cpu hadn't to compute much of the floating point calculations so it was ok, for what it had to do.
You can see Xbox's serious CPU limitations in some PC ports. It's most obvious in ported games because you can easily compare the two games cross platform.

Take Morrowind, for example. It runs on Xbox like it does on a PC with a similar CPU. It's a very CPU limited game and that doesn't change on Xbox. The result is significantly reduced draw distance and still a low framerate.

Xbox is what it is and most games are well tailored to its capabilities, but you can see the performance disparity when you can directly compare it to a PC with a port.

And sincerely, i don't think that a normal geforce 3, would pull off graphics like the best i saw on the xbox, nor the geforce 4 would vastly outperform the best games i have seen on xbox. Or at least the benchmarks i saw online inuge me to think so.
Remember that Xbox only has to deal with 480p and that most of the games are designed for its limitations. GeForce3/4 had to deal with 1024x768 or better most of the time, and games that were more generally programmed. But you can run Doom3, Half Life2, or FarCry on a GF3/4 and pretty well at that. Doom3 was first demoed on GF3.

You can't ignore that the GF3/4 had more exclusive RAM bandwidth and capacity in every variant than the UMA design of Xbox either. That will let the GPU be much more efficient. The added capacity on PC let games look better, with better textures (see Xbox Doom3). It has been said on here that Xbox's peak fillrate was a pipe dream due to memory bandwidth limits.
 
Last edited by a moderator:
And sincerely, i don't think that a normal geforce 3, would pull off graphics like the best i saw on the xbox, nor the geforce 4 would vastly outperform the best games i have seen on xbox. Or at least the benchmarks i saw online inuge me to think so.

Oh, it absolutely did, my Geforce 4 pumped out games like UT 2K3 and Half Life 2 in much higher quality than their Xbox counterparts, with better framerates and at 1024x768 with 2xaa and some af.
 
And sincerely, i don't think that a normal geforce 3, would pull off graphics like the best i saw on the xbox, nor the geforce 4 would vastly outperform the best games i have seen on xbox. Or at least the benchmarks i saw online inuge me to think so.

I owned a PC with Geforce3 ti200, Geforce4ti200 and Xbox. Games ranging from DOOMIII, battlefield, UT series, KOTOR series, HALO 1, HL2, all did run much better and looked much better. No problems doing 800x600-1024x768 and Quincunx AA for those that supported it. The ti4200 ran Far Cry demo in DX8 mode at better framerate and vastly better visuals than the Xbox Far Cry offspring.

another game was Comanche 4 which was no problem at 1024x768 with Qunicunx AA on the Geforce 4. Well cant remember all games now. :p
 
i don't remember the geforce 4 to perform like it was a completely new generation of hardware (like for example the geforce fx or even better radeon 9700), it performed better tough i certaily wasn't debating that, maybe you got me wrong on that ;). However The disparity of performance between consolles and pc will always be there, because of the ever in evolution hardware of pc's, i don't want to discuss more on that in this thread, since it is a pretty well known fact :D. What i want to discuss is what i do not know, for example which are the nv30 features inside the nv2a?
 
here's something. from one of the threads I linked you. This functionality sounds like it would have an end result similar to how NV30 can double z rate in some cases. I'm pretty sure that GF3/4 can't do this, but I could be wrong.
http://forum.beyond3d.com/showthread.php?t=2540&highlight=nv2a
ERP said:
It has an extra set of 4 Z units/pipe up front in the pipeline, so if you draw front to back, it can reject occluded pixels at 4x it's maximum fill rate. When coupled with the ZCompression it actually has a chance of getting near this rate.

Alstrong mentioned this on the previous page too.

More fun info from the threads I posted.
http://forum.beyond3d.com/showthread.php?t=2168&highlight=nv2a
ERP said:
If NV2A is mostly fillrate limited, then why don't we see Antialiasing in more Xbox games?
I've covered this before here, it's still not free.
Given 2x multisampling there is some additional cost incurred because the Z compression doesn't function as effectively. In addition the filter/copy forwards operation is expensive.
The cost isn't necessarilly prohibitive, but it does need to be planned for.

ERP said that when he benchmarked NV2A at some point (probably from his work on Boss Games' unreleased racer) that NV2A is usually more fillrate limited than bandwidth limited. I imagine that you can thank 480p for that.

Although he also said that Xbox has inadequate bandwidth to allow NV2A to ever reach it's theoretical texture fillrate limit, even in his synthetic tests.
 
Last edited by a moderator:
so in which way the nv25 is an evolution of nv2a? if nv2a has LMA2 (maybe minus the auto pre charge feature), and the accuview antialiasing
Why don't you forget about all that marketing buzzword crap? There's no such thing as a "lightning memory architecture" or "accuview" AA, it's just fancy bullshit words thought up by some PR suits.

Anyway, practically ZERO xbox games used AA so it doesn't even matter by what name you want to call it. There's actually more games on the PS2 using AA, even though that console lacked dedicated hardware support for it... ;)
 
here's something. from one of the threads I linked you. This functionality sounds like it would have an end result similar to how NV30 can double z rate in some cases. I'm pretty sure that GF3/4 can't do this, but I could be wrong.
http://forum.beyond3d.com/showthread.php?t=2540&highlight=nv2a


Alstrong mentioned this on the previous page too.

More fun info from the threads I posted.
http://forum.beyond3d.com/showthread.php?t=2168&highlight=nv2a


ERP said that when he benchmarked NV2A at some point (probably from his work on Boss Games' unreleased racer) that NV2A is usually more fillrate limited than bandwidth limited. I imagine that you can thank 480p for that.

Although he also said that Xbox has inadequate bandwidth to allow NV2A to ever reach it's theoretical texture fillrate limit, even in his synthetic tests.

very interesting, i have to read better those discussions. I'm getting even more curious about the nv30 features in nv2a, if you or anybody can add any other informations about that, i'll be more than eager tho learn.

@ grall
as i told before behind that names, are real tehnology features, so they are indeed an important factor in understanding the nv2a . That no games on xbox used AA, I Definitely don't think so, because at least 4 games supported it for sure (pgr2, uc, shen mue 2, bloodwake), and i think thera are even more. I also highly doubt, that ps2 had more games with antialiasing, than xbox, but obviously, if you can prove this, i'll have no reason to doubt anymore ;) .
 
if i have understood in the right way, antialiasing was pretty feasable on xbox, it only had to be taken in account very early in developing process, in order to make specific choiches to reduce bandwidth requirements in other areas (geometric compexity, number and quality of effects etc.). It is just difficult to me to understand if a game uses or not the antialiasing and at which rate, since i play xbox on an hdtv (with component, in 480 p), the image may be incurring, scaling-induced aliasing. Some games however seems very smooth, even on hdtv, so i presume, they use some kind of AA. Dead or alive ultimate and max payne 2 for example has a very clean look, i wonder if the actualy use AA.
 
Back
Top