Um VF4 NOT to be best looking fighter?????

doesnt the nv20 have any onboard ram?
or has it just not been announced yet.
wtf?????????????????????????????????????????????????????
????????????????????????????????????????????????????????
????????????????????????????????????????????????????????
are you saying its going to take everything from main memory? are you sure?
that is like purposefully taking a step backwards. all pc vid cards have onboard memory.
plus the thing has T&L so it should have a lot of onboard mem.
????????????????????????????????????????????????????????
????????????????????????????????????????????????????????
????????????????????????????????????????????????????????
thats crazy

 

Dude anyone knows that to use texture compression, it takes a hit on the CPU. S3TC the form of texture compression used by the N2 is not a standard. In other words most games do not take advantage of S3TC because of the peformance hit.

The 10GB HD the X-Box has is being used to store game data as well as OS, and as a swapping drive.

CrewNYC

 

NAh it just means that it renders only what the eye can see. So a scene that takes 10 million polygons to create it can fake with say 5 million yet still give the same visual resukts, since only so many polygons are being shown at any given time. Traditional methods render everything regardless of whether or not you can see it.

Efficency is a matter of coding, I think if you have the power in the machine you dont need tricks to produce the same results. You just use the traditional methods of creating things. Games designed for multiple systems for instance probably wont use this feature.

CrewNYC

 

"does xbox have some kind of insane bandwidth to stream textures in (like what could have saved ps2.)
thats what agp is supposed to do isnt it. "

Yeah its called a dedicated bus. The X-Box wont have any of the legacy technologies associated with PC's. The dedicated bus will give the X-Box a huge amount of memory bandwidth.

On a quick note the X-Box is capable of 30-40 Million polygons per second with more than 8 simulatenous light sources, and several effects (NV20 T&L Chipset). Regardless of how many lights the N2 uses it is limited to 12 million polygons per second max.

CrewNYC

 

Looks like im playing teacher today:

"how would doing T&L calculations ever run in (anything resembling) sli? how would you do it so it wouldnt be a duplication of work.
there must be one gpu for n2 and the pvrs just do rendering. or i dunno and theyre geniuses"

The answer is basically even and odd lines. Basically to render the image half the work is done by one GPU the other half by the second GPU. The CPU recobmines the image and shoots it to the display. Basically allowing you to render twice as much. This was the same as Voodoo 2 SLI. Voodoo 2 SLI rendered one line in the image (even) while the other rendered the another line in the image (odd).

CrewNYC

 

lol. youre really wrong there, buddy.
you better admit this one, lol.
i know what scanline interleave is. but it has to do with _rendering_. both units rendering each other line. they are not 2 gpus doing all polygon calculations on a every other line basis. because you just need to know where the verts are and draw the lines. that woudlnt even work (or just be totally insane.) we are talking about tranform and lighting, which the GPU (or T&L) unit does. it has nothing to do with rendering at all.
calculating verts or lighting on an every other line basis is an unnecisary duplication of work and extremely crazy.
its so crazy i cant even explain it.
thats why i said there must be 1 gpu and 2 pvrs for rendering. not 2 gpus.
there is no way in hell you are going to design a gpu for anything resembling sli. maybe something along the line of vector units for ps2 but those arent gpus.
admit this one, or ill put you on my list of people who cant admit. lol. thx. jtx

 

thats cool. good info, thx. cuz i know fast agp ports are *supposed* to be able to stream in textures per frame. that was the original plan atleast. but i never kept up with it. the intel graphics card which was meant to show off (i740, actually i *think* it was codesigned by L&M, but not 100% sure, the guys who did the m2 and m3) the original agp spec was designed to take advantage of this (intel thought the agp bus would be able to get passed mem on the card constraints.) anyway, yadayada. atleast theres a precedent for them trying it out on pcs. and this was a while ago. hopefully this means they can stream in textures per frame, what they hoped ps2 would do. if they can, this will make xbox even more of a monster.

if it is using a pc cpu it has tons of legacy tech on that alone, but i wont be anal. the reason pc cpus have to be so fast compared to others is cuz they have to support all the old legacy tech. a lot of people would be laughing on this one 'there wont be any old legacy tech.' is that from ms marketing?

do you know how fast the bus is? is it some new tech and called something special? it has shared mem and i dont know much about the design of those systems, there are regular cheapo pcs with shared mem for vid, maybe it is based on those designs. has all the specs been finalized for nv20?

 

crewnyc: "Dude anyone knows that to use texture compression, it takes a hit on the CPU. S3TC the form of texture compression used by the N2 is not a standard. In other words most games do not take advantage of S3TC because of the peformance hit."

there is such thing as in hardware, free texture compression/decompression on the fly. it doesnt need cpu at all. s3tc is a compression standard for hw texture decompression, and it can be licesned (you need hw that can support it though.) its made by the vid card company s3. s3tc. wooop!
lots of really false information here. im getting less willing to correct it all so my replies from now on will be shorter.


i dunno. this place is whack. whats up with here. this is like charlie and the chocolate facotry for misiniformation
thx. peace. out. jtx

 

i know what it is but didnt think it was to that level of efficiancy where it litterally doesnt waste any rendering at all. are you sure? i thought it just takes little tiles of the screen and loads the textures in for that tile, so there willl still be waste, it still does it the normal method of rendering. i think if a building is half obstructed, id probably render the whole thing. it just saves if it doesnt have to load anything for that particular tile. but if it does, it will render it the normal way. if it does what you say, and there is litterally no waste it is truly a marvel of engineering. somehow im owrried to believe you cuz you were wrong on the other stuff.

 

general thread!!!!!

ok, last time i play mod. ^_^


fuck spelling!

 

Jtx

Forst of all the PVR2 is the GPU there is no GPU then 2 Power VR setup. It is not scan line interleave I was referring too. Basically a game engine determines, there are x number of polygons on the screen, a single GPU setup would render and typically light (with T&L bulit into the hardware), and assist in mapping the textures. A SLI GPU setup as the N2 has does not duplicate the work, basically its the same thing. The OS supports the SLI function so it says to the N2, Hey we have a second GPU, render every other line of RESOLUTION, an image whether polygons or not is bacially a few lines of resolution being displayed on your screen.

This is the exact same concept as a filter render in photoshop with a dual processor setup. SLI and all dual processor setups follow the same basic logic. Peace.

CrewNYC

 

Nah man this has been known for a while, S3TC was created by s3 the reason for developes not using it was because of the performance hit. It is built into the hardware. The claim of "realtime decompression and compression on the fly" was not so real time. This is why though S3TC was great and shown off nearly 3 to 4 years ago no one adopted it.

CrewNYC

 

um no, I was not wrong on the other stuff jtx. X-box shit is pc based which is my strong suit. NV20 specs are done. The geforece 3 is coming to market as we speak, so the card and drivers are complete.

About lighting I should probably describe this too since I didnt earlier

Geforece 1 was capable of 2 lights in game easily
Geforce 2 was capable of 4 lights easily in a game
Geforce 3 is capable of 8 lights easily in a game

(this does not reflect the maximum amount of lights, since even geforce 2 could have more than 4 lights)

CrewNYC

 

actualy let me clarify before you get the chance to say something esle about this. The type of SLI is important. In the case of a V2 it was doing two things, It was rendering every other line in a polyginal image, but the SLI of V2 was designed to provide higher resolutions. So the drivers which powered the cards, did not take full advantage of what SLI was designed for. Voodoo 2 SLI which I believe is what you are referring to above is not the same as N2. The N2 has native support within the drivers to split T&L calculations and rendering. It boils down to how the drivers were designed to work with the games.

CrewNYC

 

read complete reply.

Heres a basic definition of SLI. There are different implentations of SLI, read below.

SLI mode means two chips share the chore of painting frames in a 3D application; they each simultaneously work on different areas of each frame.

 

Guys calm down!!!

First of all you should delete all the threads on pg 3 since nearly all the info on the N2 tech is wrong. You should search a little bit on N2 before you try and guess how it works.

N2 has 1 SH4(CPU)+1 Elan(T&L)+2 PVR2DC(rasterizer). It has 32+32+32megs of RAM.

Now the PVR2DC does not render in every other line like the V2 sli but in tiles. This mean that the screen is split in 32*32 tiles and the 2 PVR2s just render half of the tiles that they would render if there was only one rasterizer. Basically they are 100% efficient.

On the polygon argument. The T&L has to calculate all the polygons as a normal(non tile based) renderer. The advantage of a tiler is that it has differed texturing amongst other advantages such as bandwith saving. The ELAN can calculate and light 12mpps with up to 6 LOCAL lights with no performance hit. The RAW rendering is greater but since the PV2s are limited to 6.7mpps each there is no point in saying otherwise. If more lights are used the polygon performace is reduced slightly. XBOX with a 300mhz GPU can calculate and light 8mpps with 8 local lights. Since the GPU was reduced in speed to 250mhz the figure is reduced considerably.

Regarding the compression method. The N2 does not use S3TC but Power VQ texture compression as the DC. Power VQ is able to achive an average of 8:1(maybe even more with the latest tweaks) compression with NO performance hit. S3TC gets 6:1 compression no matter the size of the texture. So when using large textures Power VQ has a big advantage. Both those methods should not produce a performance hit but they should actually improve performance since the bandwith needed to transfer the textures is greatly reduced by 8:1 on N2 and 6:1 on XBOX.

The only advantage that the XBOX has on the N2 when using 6 LOCAL lights is the Pixel shaders. This is something which will give a big rendering advantage to the XBOX since the shading is done on a per pixel basis and not on a per vertex basis. N2 can still do pixel shading but in limited form. An examble of this is the bump mapping on the clothes of the VF4 characters. If less lights are used the XBOX will get more polygon performance than the N2 simply because the N2 is limited by the PVR2s.

N2 is incredible if you consider that everything is running at 100Mhz exept for the CPU which is at 200Mhz, and that the tech used is from 1998(PVR2DCs and SH4) and 1999(ELAN T&L). This is competing favourably with 2001-2002 tech which runs with 250Mhz and higher components.

CIN

 

CIN, thanks for your post. If you are right then there really is order in the world. I was wondering how the Xbox could possibly so much more powerful than the N2 and still be so cost-effective; the gap between DOA3 and VF4 helps to cast doubt to that notion.

Obviously a big reason Sega chose to use N2 is that N2 is backwards compatible with Naomi, but I didn't think Sega would go with a platform to showcase their premiere franchise that is technologically inferior.

 

Exactly.

In fact N2 is on the same level as XBOX with the advantage of having tile based rendering(imagine shenmue type graphics) and the disadvantage of not having pixel shaders.

N2 could also be more cost effective since all the components are super cheap.

 

basically youre missing the whole thing.
#1. pvr doesnt use sli. or anything resembling it.
#2. sli does in fact mean scan line internleave, and youre saying youre not refering to scan line interleave, then you start using the sli abreviation. doesnt matter anyone since it doesnt use it!
#3. tile rendering and sli are totally different.
#4. my orig question was simply that does the pvr somehow use 2 gpus in sli, since that is what you implied. i said in anything resembling sli since i know it wasnt using sli but thought you still knew what you were talking about and just used the term to describe the concept. this whole argument is my fault cuz i believed you.

you said: you said "The last element the GPU is about 2X as powerful as the N2 PVR2 Sli setup."

how can a gpu be more powerful than 2 renderers in sli (or anything resembling)
the n2 has 1 gpu and uses the pvrs for renderers. you implied it somehow had 2 gpus, i believed you, asked about that, thats all. then you started getting your terms mixed up and i replied pointing that out.
whole point is now moot because youve been wrong a couple a bunch now. no hard feelings, honestly.
peace. out. jtx.

 

Heheheh..

Listen to "CIN" Andy, he knows his shit!

Thanks for clarifying shit, CIN-dude.

This is gonna be HOT, mo-fos!!


<font color=red>Jyunen hayaindayo!</font color=red>

<font color=white>adam</font color=white><font color=red>YUKI</font color=red>