a question of Poly's

Just a argument with Cause, but does VF1 have more poly's than the rest of the vf series? I do not think so (in that the hardware has gotten even more powerful)...we are just talking about character models.

"Victory can be anticipated, but not assured" Sun-Tzu

 

Hell no!

VF1: 100,000
VF2: 300,000
VF3: 1,000,000
VF4: ??,???,??? I've heard 10,000,000...?

Keep in mind that polys became progressively better quality, with texturing, lighting, etc. Also, all Model hardware use square polys. Not sure about Naomi 2, but I think they are triangle, but the N2 uses deferred rendering, meaning they get more with fewer polys calculated.

 

i heard vf2 used less polys than vf1, cause texture mapping helps. correct me if im wrong..



<font color=red>nAndato</font color=red>

 

this is actually very interesting... I'm 99% certain that in terms of polys per character VF4 > VF3 > VF2 > VF1, and I was going to reply saying just that, but then I realized that I don't actually have any data to back that up.

I seem to remember hearing that (on the Saturn) VF2 characters were around 2,000 polys each. I also seem to remember that VF1 characters were in the low hundreds. Both of these figures sound right to me -- I am in no doubt that VF2 characters are higher detail (meaning more polys) than VF1 characters.

Of course the textures help, but not _that_ much (remember that VF Remix had textured and gouraud shaded VF1 characters, and its characters didn't look as detailed as VF2 IMHO). I'm going to weigh in and say that VF2 characters _definitely_ have more polys than VF1 chars.

But does anyone have the actual figures?

 

vf1 had more polygons in the bodies of the characters than in the bodies in vf2, and probably even for some characters in vf3. the faces in vf2/3, however, outnumber vf1 by a ridiculous amount.

--
"What we got here is a failure to communicate..."

 

Well the floors in VF1 had to be made up of many polygons whereas the ones in VF2(flat, square arena) were made up of exactly... one.

I still reckon VF2 has more though.

 

I think what you're thinking of is that VF1 Saturn used more polys than VF2 Saturn, but that's certainly not the case for the arcade versions.

 

Naomi 2 use triangles. It's doubtful VF4 use 10 mpps but it shows how amazing is the hardware.

 

*creed screams "fuck 'im up, rich, fuck 'im up!!@!!" over my shoulder....*

i'll repeat..

vf2 arcade had less polygons in the bodies of the characters in vf1. texture mapping/lighting/shading took care of this. compare jeffry's vf1 torso to his vf2 torso. his vf2 legs are practically square boxes. have you ever seen a glitchy vf2, where the lighting effects have been disabled? the models are frighteningly simple. it's really amazing how much the shading made up for it.

some body parts are very simple in vf3, as well.

--
"What we got here is a failure to communicate..."

 

Do you have actual figures? I'm not doubting what you're saying -- this whole thread has just made me curious as to how detailed the models actually are in each version.

grib

 

i've only got figures for vf1...

--

Polygons - number of polys composing each character:

Akira: approximately 2300
Jeffry: approximately 2000
Pai: approximately 2000
Sarah: approximately 1900
Lau: approximately 1900
Wolf: approximately 1800
Kage: approximately 1500
Jacky: approximately 1500


--
"What we got here is a failure to communicate..."

 

cool... that's interesting. That's actually a lot _more_ than I would have thought per character for VF1 (considering a Quake 2 model has what... 400 or 500 polys?).

Where'd you get these figures from?

grib

 

If those figures are correct, than VF1 used more polygons on the characters than Tekken 3. Tekken 3 characters are modeled using approximately 1500 polys.

I think that VF2 used more polys overall than VF1 but you could be right about the body counts. Since I never had the opportunity to see VF2 and VF1 side by side in the arcade I cannot tell.

One thing you should remember though is that Model2 did not have the ability to use gauraud shading(the saturn had this ability) so because of this I would say that the bodies of VF2 have more geometry than those of VF1.

The reason why the characters in VF2 sat appear more rounded is because the arcade lacked gauraud shading not because of more polys. So IMO I think that VF2 bodies had more polys than those in VF1 but I could be wrong. Regarding VF3.

There is no way that the characters in VF1 had more polys than those in VF3. VF3 characters had about 5000polygons each and another 5000 for the background.(I read this in an interview).

CIN

 

Yep...Mr. Bungle and Piccolo are right. I also read somewhere that VF1 (arcade) characters' bodies were made up of more polys than VF2 (arcade) characters' bodies. Model 2s texture mapping ability allowed them to work with fewer polys on character bodies. Just can't remember where i read it tho....

 

They were in a Japanese VF book.

The thing about this is that because there were no lighting-and-texture tricks in VF1 (as opposed to what was used in 2 and up) the designers had to 'fake' some of the detail in their characters by adding more polys. Now they can get amazing effects with just really well done scanned texture maps. Imagine trying to do lau's wrinkles with no texture maps. That'd add another 600 polys right there.


/versus/images/icons/mad.gif<font color=red>~~~ Don't make me rape you with a sharp stick ~~~/versus/images/icons/mad.gif<font color=red>

 

What're you spazzing about? I read you the first time. And my figures, though somewhat misapplied, are correct and apply for overall polys per second for the game (not just the character models).

And I do hearing from someone that Saturn VF1 had more polys overall than Saturn VF2.

 

On average VF2 used more polygons that VF1 but in the caracter bodies VF1 might have a slight advantage.(The bodies excluding the head) Remember we are not talking polygons per second but polygons used per character body.

One thing that you should also remember is that VF1 ran at 30fps and not at 60. This would mean that if VF2 used 3 times as much polygons per second it would not be equal to 3 times as much polygons per frame!!!

Overall polygons per second and per frame I think that it should be VF1>VF2>VF3>VF4

PS. IIRC Model 1-3 rendered in quads while N@omi1-2 render in tris and also in quads. Also deferred rendering does not save polygon calculations but saves on bandwith and texturing(that is why up till now it is called deferred texturing). ie. The polygons are still calculated like Immediate mode renderers but textures are only rendered after the renderer(PVR)establishes which are the visible polygons so that there is no wasted rendering. Also opaque overdraw for a deferred renderer is 0 while for a traditional it can be extremly high 5+ in some game situations.

The only problem for tilers up till now has been that they were aimed to the middle range(PVR2 came out late so it was aime to the middle range) and so they have low RAW fillrate(for high res gaming). So when there is a game with lots of transparencies tilers tend to suffer more(at least until PVR4 with its 1GPPS RAW fillrate).

Regards

CIN

 

Slightly off topic, but Sega should have developed VF4 on a GeForce 3 platform instead /versus/images/icons/laugh.gif

That way they would have gotten incredible 3D performance, programmable shaders, a (theoretically) infinite number of light sources (depending on how you set up the shaders... check out the MacWorld DOOM3 demo, for e.g.) ... AND it would be a lot cheaper to produce the arcade machines, using stock chipsets, plus minimal effort required to port to PC and Xbox.

The days of arcade hardware outstripping PC components have long gone... (probably because arcade manufacturers wasted too much time creating new Bemani games and not pushing the technology envelope).

 

At first I thought you were kinda dumb and missing the point by arguing # of polys vs. frames per second when we were just talking about total polys per char. But then you mentioned the opaque overdraw for a deferred renderer being as high as 5 and I was like "Holy shit! Thatz some technical bizzidness meign!" ...

Now I think we're still missing the point. Even the Ultimate Archiver (mr bungle) only has stats for one game, and that's VF1. So I think any speculations we have about which game has more polys are just pissing in the wind. You can't say for sure which has more when we only know how many 1 of them has.

But common sense would seem to tell us at first that it's VF2.
But then the argument seems to be that VF1 had more polys because it needed to cheat and use extra polygons to accomplish the same thing that more advanced rendering engines do with textures.
But anyway. What was my point? Oh yeah, without official statistics who can say?


/versus/images/icons/mad.gif<font color=red>~~~ Don't make me rape you with a sharp stick ~~~/versus/images/icons/mad.gif<font color=red>

 

When you consider the specs of the Arcade Boards

Model 1

Main CPU : 32bits RISC NEC v60 uPD-70616, 16Mhz (2.5 MIPS)
Co-Processor : Fujitsu TGP FPU 32bits 16M flops (Floating Point, Stem Rotate, 3D Matrix).
Sound CPU : 16bits 68000 @ 10Mhz
Sound chip : 2 x Custom 28 channel PCM chips (can access up to 8meg sample rom *per chip*) + YM3834 (only used for it timers)
Video resoution : 496x384 in 65536 colors.
Geometry : 180,000 polygons/sec, 540,000 vectors/sec
Rendering : 1,200,000 pixels/sec
Video : ShadingFlat Shading, Diffuse Reflrection, Speculer Reflection, , 2 Layers of Background Scrolling, Alpha Channel





Model 2A

Main CPU : Intel i960 @ 25 MHz 32bits RISC
Co-Processor : 4xFujitsu TGP FPU 32bits 16M flops (Floating Point, Stem Rotate, 3D Matrix)
Sound CPU : 16bits 68000 11.3Mhz
Sound chip : SCSP/YMF292-F (315-5687)/"LAKE" @ 11.3MHz, 32 PCM (Pulse Code Modulation) Channels, 44.1 khz Sampling Rate
Audio RAM : 540 Kilobytes (4 megabits)
Video resoution : 496x384 in 65536 colors
Geometry : 300,000 polygons/s. 900,000 vectors/s
Rendering : 1,200,000 pixels/s
Video : ShadingFlat Shading, Perspective Texture, Micro Texture, Multi Window, Diffuse Reflrection Mode, Speculer Reflection Mode



Model 3 step1

Main CPU : 32bits RISC PowerPC 603 66Mhz
Sound CPU : 16bits 68EC000 11.3Mhz
Sound chip : Yamaha SCSP/YMF-292F/"LAKE" FH1 128-step DSP x 2, MIDI interface, 16 bits 64 voices 4 channel, maximum of 16.5 Mbytes ROM, 64 PCM channels
Audio RAM : 1meg (8 megabits, 540 bytes per SCSP chip)
Main Memory : 8 Mbytes 66mhz Ram, graphic ROM maximum of 64 Mbytes, backup RAM 64 Kbytes
Video resolution : 24KHz 496(H)x384(V) one or two plane or 31KHz 640x480 one plane in 16bit colors
Scroll Window : two plane (24KHz/two plane mode), 16 colours/32,768 1024 palette x 2 bank, 256/32,768 64 palette x 2 bank
Geometarizer : 1,000,100 polygons/s for square polys, 2,000,200 for Triangle polys
Renderer : 60,000,000 pixels/s
Video : Full Color Texture Mapping, Tri-Linear Interporation Micro Texture, ShadingHigh-Specular Gouraud Shading , Fix Shading, Flat Shading, Texture & Edge Multi Layered Anti-Allasing, Lighting EffectParallel Light, 4 Spot Light, Pin Spot Light, Special EffectZoning-Fog, 32 Levels of Translucenncy
Board composition : CPU + VIDEO + ROM boards
Others : 10mbs Connection, calender IC



N@omi2

CPU : SH-4 128-bit RISC CPU (200 MHz 360 MIPS / 1.4 GFLOPS)
Graphic Engine : 2 x PowerVR 2 (PVR2DC-CLX2) GPU's - (under the fans)
Geometry Processor : Custom Videologic T+L chip "Elan" (100mhz) - (Under Heatsink)
Sound Engine : ARM7 Yamaha AICA 45 MHZ (with internal 32-bit RISC CPU, 64 channel ADPCM)
Main Memory : 32 MByte 100Mhz SDRAM
Graphic Memory : 32 MByte
Model Data Memory : 32MByte
Sound Memory : 8 MByte
Media : ROM Board / GD-Rom
Simultaneous Number of Colors : Approx. 16,770,000 (24bits)
Polygons : 10 Million polys/sec with 6 light sources
Rendering Speed : 2000 Mpixels/sec (unrealistic max, assumes overdraw of 10x which nothing uses)
Additional Features : Bump Mapping, Multiple Fog Modes, 8-bit Alpha Blending (256 levels of transparency), Mip Mapping (polygon-texture auto switch), Tri-Linear Filtering, Super Sampling for Full Scene Anti-Aliasing, Environment Mapping, and Specular Effect.

Fully backwards compatible with all Naomi and GD-Rom games



As you can see from the specs, it is very difficult to believe that VF used more polys per second than VF2 but regarding the body and total counts it is entirely believable that VF used more polys per character or per scene(frame) than VF2 especially since VF1 rendered at 30fps and there where no polygonal backgrounds.

Still IMO VF1>VF2>VF3>VF4 altough VF1>VF2 is really arguable until someone has the official nos(probably noone)

Regarding the use of polygons instead of textures. I would say that when using textures u are actually cheating because you are trying to show a smooth model with a blocky one. When using polygons with no shading(just flat) you cannot cheat since every poly you add will be seen on the character as increased detail. This can be seen in SC and in TTT. SC characters use less polygons but when viewed from playing camera the fighters appear little or no less detailed that those in TTT which use much more polygons.

PS. From the specs above one can see that Model 3 was a really powerful hardware.(maybe the hardware that was the most advanced for the longest time.) It can do 4 spot lights when using 2Million polygons and that is still impressive to this day, especially since even GFORCE cards on PC strugle to get that kind of performance.

PPS. All the information above came from www.system16.com while the VF4 pic came from www.computerandvideogames.com .

CIN