Welcome, Guest!

Here are some links you may find helpful

Virtua Fighter 4 Dreamcast....It would have been possible with enough I+D?

accel99

Well-known member
Registered
Jun 3, 2019
48
44
18
AGName
Accel99
AG Join Date
March 28 2008
Probably the thing with Fighting Vipers 2 and others Model 3 ports is exactly that, those are games built around the complex tech of Model 3, so at the time they were not capable to convert the game 100% faithfully, bc DC has an entire different architecture. And about the DOA 2 numbers, yes, they absolutely prove VF4 could be somewhat handled by DC. Could DOA 2 Engine be modified to create a....DC VF 4?

While what you say might be true about being super different ,he posted more examples( Virtua fighter 3 rooftop stage, Virtua striker 2 stadium) that genki had more detailed models for their model 3 ports despite how looked down upon they were. Overall it seems they out performed am2 in porting from the same machine, again what happened to that famed prowess that an outside dev could out do them on their own machine and tools? It's just a very curious and mysterious case. At its best judging from downgraded Dreamcast examples model 3 probably runs no better/ higher than doa2 on the DC tops, shouldnt first party devs done better at these ports? Especially genki getting much closer ( by Virtua striker 2 it's super close to the model 3 except for a few things) compared to am2?
 
  • Like
Reactions: Xaeroxcore

Xaeroxcore

Well-known member
Original poster
Registered
Aug 22, 2019
51
40
18
AGName
xaeroxcore
AG Join Date
Oct 17, 2012
While what you say might be true about being super different ,he posted more examples( Virtua fighter 3 rooftop stage, Virtua striker 2 stadium) that genki had more detailed models for their model 3 ports despite how looked down upon they were. Overall it seems they out performed am2 in porting from the same machine, again what happened to that famed prowess that an outside dev could out do them on their own machine and tools? It's just a very curious and mysterious case. At its best judging from downgraded Dreamcast examples model 3 probably runs no better/ higher than doa2 on the DC tops, shouldnt first party devs done better at these ports? Especially genki getting much closer ( by Virtua striker 2 it's super close to the model 3 except for a few things) compared to am2?
So, in this case can we say AM2 was just lazy? or maybe they were busy working on Shenmue 1 and 2 for DC and VF 4 for Naomi 2, so they decided to put little effort on the Model 3 ports.....
 

dark

Well-known member
Registered
Jun 19, 2019
91
72
18
AGName
dark
AG Join Date
September 2, 2011
According to descriptions on the internet, Genki was basically founded by ex-Sega AM2 members. I suppose that would motivate them to do a good job whe doing work like porting later AM2 properties.
 

accel99

Well-known member
Registered
Jun 3, 2019
48
44
18
AGName
Accel99
AG Join Date
March 28 2008
More comparison. He has converted Lau from vf3 DC but it's incomplete. Even incomplete it's higher than bahn from fv2. Interesting. To compare to another game from the same years, scorpion from mortal Kombat gold is 4,200 triangles. That's crazy , might actually be higher than Lau completed if you look at it. Then when you think about it mkgold can display up to 3 characters, I guess it was a really good conversion. Makes me wonder if the model 3 is overhyped perhaps? Despite how much shit genki got the model looks fairly intact. Well IAM going off topic, but it seems than a similar/ slightly less ps2 spec of vf4 might have been possible if they probably thought it economically viable.

dclau1.jpg

dclau2.jpg


Scorpion mkgold:
MKGOLDSCORP.jpg
 

Xaeroxcore

Well-known member
Original poster
Registered
Aug 22, 2019
51
40
18
AGName
xaeroxcore
AG Join Date
Oct 17, 2012
More comparison. He has converted Lau from vf3 DC but it's incomplete. Even incomplete it's higher than bahn from fv2. Interesting. To compare to another game from the same years, scorpion from mortal Kombat gold is 4,200 triangles. That's crazy , might actually be higher than Lau completed if you look at it. Then when you think about it mkgold can display up to 3 characters, I guess it was a really good conversion. Makes me wonder if the model 3 is overhyped perhaps? Despite how much shit genki got the model looks fairly intact. Well IAM going off topic, but it seems than a similar/ slightly less ps2 spec of vf4 might have been possible if they probably thought it economically viable.

dclau1.jpg

dclau2.jpg


Scorpion mkgold:
MKGOLDSCORP.jpg
Amazing job bro! It would be possible to have the geometry numbers of Force Five? I´m 100% sure Force Five Polycounts per characters are less than the ones on VF 3 and MK Gold, despite looking better.
 

MetalliC

Well-known member
Registered
Jun 28, 2019
75
71
18
AGName
MetalliC
AG Join Date
23.04.2014
Amazing job bro! It would be possible to have the geometry numbers of Force Five? I´m 100% sure Force Five Polycounts per characters are less than the ones on VF 3 and MK Gold, despite looking better.
Force Five is like 16K vertices per scene during fight. nowhere near VF4 - 50-70K, at some arenas with water - 120K vertices per scene
 
  • Like
Reactions: Xaeroxcore

accel99

Well-known member
Registered
Jun 3, 2019
48
44
18
AGName
Accel99
AG Join Date
March 28 2008
Force Five is like 16K vertices per scene during fight. nowhere near VF4 - 50-70K, at some arenas with water - 120K vertices per scene

Do you think you can tell me the vert per scene count of the peak intro of virtua striker 2 and gameplay of soul surfer Naomi 2 , please?

I remember in the old assembler forum you admitted wacky races peaked at around over 70k vertices per frame(which is the mid-high load of vf4,no?) at 30 frames per second. Sure it's probably 1 light and half the fps of Virtua fighter 4 but no one's arguing that the Dreamcast can match the Naomi 2 under the same conditions. Just that given that information it can process enough polygons/ detail to make a good approximation of most of that Naomi 2 game( most stages in the game don't have water) even if running half the frame rate and 1/6 of the amount of light source.
 
  • Like
Reactions: Xaeroxcore

MetalliC

Well-known member
Registered
Jun 28, 2019
75
71
18
AGName
MetalliC
AG Join Date
23.04.2014
Do you think you can tell me the vert per scene count of the peak intro of virtua striker 2 and gameplay of soul surfer Naomi 2 , please?
Naomi1 VS2 is up to 28K in intro, during gameplay average ~16K with some ~20K peaks
Noami2 Soul Surfer is like 20K-55K during gameplay with peaks up to 100K at high score scene

Sure it's probably 1 light
hehe, yes, it is easy way to determine people who have nearly no any idea what they talking about lol - they compare different hardwares or games only counting triangles/vertices, without taking in account lighting complexity/quality, which easilly may increase required T&L computing power several times.
 
  • Like
Reactions: accel99

accel99

Well-known member
Registered
Jun 3, 2019
48
44
18
AGName
Accel99
AG Join Date
March 28 2008
Naomi1 VS2 is up to 28K in intro, during gameplay average ~16K with some ~20K peaks
Noami2 Soul Surfer is like 20K-55K during gameplay with peaks up to 100K at high score scene


hehe, yes, it is easy way to determine people who have nearly no any idea what they talking about lol - they compare different hardwares or games only counting triangles/vertices, without taking in account lighting complexity/quality, which easilly may increase required T&L computing power several times.

Last request , hikaru is always listed as 2mpps machine. I understand completely different architecture from the rest but still curious . Do you happen to have vertex per scene figures for air Trix and planet harriers, please?

About the quality per scene yes I understand lighting is a big burden on the sh4 as other things than polygons like physics , game logic and shuffling files take a toll and so on.

To be honest it always looked grim in that regard, the head of warp mentioned that about d2 in an interview and even gave figures , which is rare. He said he had the game running at 60k polygons per frame with 1 light but was dissatisfied with the look so preferred to lower the peak figures to 30k polygons per frame but with 3 lights for a better look. And that was an early game.
 

MetalliC

Well-known member
Registered
Jun 28, 2019
75
71
18
AGName
MetalliC
AG Join Date
23.04.2014
hikaru is always listed as 2mpps machine. I understand completely different architecture from the rest but still curious . Do you happen to have vertex per scene figures for air Trix and planet harriers, please?
Air Trix is like 40-50K, most of time in the middle
Planet Harriers ~50K+ - ~60K+ during gameplay, peaks up to 80K
but, Hikaru does per-pixel lighting, which means it has many times more computing power than lets say Naomi2.
 
  • Like
Reactions: Xaeroxcore

Xaeroxcore

Well-known member
Original poster
Registered
Aug 22, 2019
51
40
18
AGName
xaeroxcore
AG Join Date
Oct 17, 2012
Naomi1 VS2 is up to 28K in intro, during gameplay average ~16K with some ~20K peaks
Noami2 Soul Surfer is like 20K-55K during gameplay with peaks up to 100K at high score scene


hehe, yes, it is easy way to determine people who have nearly no any idea what they talking about lol - they compare different hardwares or games only counting triangles/vertices, without taking in account lighting complexity/quality, which easilly may increase required T&L computing power several times.
So, according to your knowledge, how could an hypothetic VF 4 port could be achieved by DC? or that Naomi 2 game is straight impossible to port directly to DC, and the only way would be with a version built from zero around DC limitations that probably would look on par with DOA2 in the best case scenario?
 

Xaeroxcore

Well-known member
Original poster
Registered
Aug 22, 2019
51
40
18
AGName
xaeroxcore
AG Join Date
Oct 17, 2012
Force Five is like 16K vertices per scene during fight. nowhere near VF4 - 50-70K, at some arenas with water - 120K vertices per scene
How it compares in vertices count, lighting, etc with DOA 2, VF 3 or Soul Calibur?
 

MetalliC

Well-known member
Registered
Jun 28, 2019
75
71
18
AGName
MetalliC
AG Join Date
23.04.2014
So, according to your knowledge, how could an hypothetic VF 4 port could be achieved by DC?
there needed skilled&talented&enthusiastic person(s) with a ton of spare time to work on this task. the rest is not much important, especially some DC or whatever else target platform tech limitations.

How it compares in vertices count, lighting, etc with DOA 2, VF 3 or Soul Calibur?
was not cared to compare
 

la-li-lu-le-lo

Well-known member
Registered
Jun 13, 2019
319
115
43
AGName
la-li-lu-le-lo
AG Join Date
2-8-2006
People seem to forget that were major differences between the Model 3 and the DC/Naomi hardware. Just talking about poly-counts doesn't tell the whole story. The DC and the Model 3 each had their own advantages, but I would argue the Model 3 was the more advanced system overall.

The Model 3 had multiple revisions, some of which had a CPU that had similar power to the DC's CPU. The Model 3 had dual GPUs instead of the DC's one. Those GPUs in the Model 3 were designed by Lockheed Martin, based on their multimillion dollar military flight simulators. It supported more advanced types of lighting effects. A really major difference between the two is that the Model 3 had a massively larger amount of memory: ~250MB total, compared to 8MB for the DC. Just the texture RAM alone for the Model 3 was larger than all of the DC's memory. All of those things have effects that go beyond just the number of polygons.

Another thing people often ignore when talking about VF3 is that it was the very first Model 3 game, and as such it was one of the least graphically complex and demanding games for the system. Later games for the later revisions were much more complex, as were even games like Scud Race that came out only months after VF3. The Model 3 was a high-end system, and the DC was a low-cost consumer system. The DC was a great system for sure, but it couldn't really compare to the Model 3 in some pretty important areas. I own a Model 3 VF3 board - I've played it quite a bit. I play my DC all the time too. I could say more about this, but I'll leave it at that for now.
 

accel99

Well-known member
Registered
Jun 3, 2019
48
44
18
AGName
Accel99
AG Join Date
March 28 2008
Naomi1 VS2 is up to 28K in intro, during gameplay average ~16K with some ~20K peaks
Noami2 Soul Surfer is like 20K-55K during gameplay with peaks up to 100K at high score scene


hehe, yes, it is easy way to determine people who have nearly no any idea what they talking about lol - they compare different hardwares or games only counting triangles/vertices, without taking in account lighting complexity/quality, which easilly may increase required T&L computing power several times.

There's a tech demo released recently , Simpsons bug squad, it doesnt seem to use any lighting and nulldc seems to indicate it pumps out alot of verts around 20 ish fps. I was wondering if you have a more accurate vertex per frame count on this demo and frame rate? Just to see how it is unhindered by lighting.
 
Last edited:

accel99

Well-known member
Registered
Jun 3, 2019
48
44
18
AGName
Accel99
AG Join Date
March 28 2008
People seem to forget that were major differences between the Model 3 and the DC/Naomi hardware. Just talking about poly-counts doesn't tell the whole story. The DC and the Model 3 each had their own advantages, but I would argue the Model 3 was the more advanced system overall.

The Model 3 had multiple revisions, some of which had a CPU that had similar power to the DC's CPU. The Model 3 had dual GPUs instead of the DC's one. Those GPUs in the Model 3 were designed by Lockheed Martin, based on their multimillion dollar military flight simulators. It supported more advanced types of lighting effects. A really major difference between the two is that the Model 3 had a massively larger amount of memory: ~250MB total, compared to 8MB for the DC. Just the texture RAM alone for the Model 3 was larger than all of the DC's memory. All of those things have effects that go beyond just the number of polygons.

Another thing people often ignore when talking about VF3 is that it was the very first Model 3 game, and as such it was one of the least graphically complex and demanding games for the system. Later games for the later revisions were much more complex, as were even games like Scud Race that came out only months after VF3. The Model 3 was a high-end system, and the DC was a low-cost consumer system. The DC was a great system for sure, but it couldn't really compare to the Model 3 in some pretty important areas. I own a Model 3 VF3 board - I've played it quite a bit. I play my DC all the time too. I could say more about this, but I'll leave it at that for now.

No one saying that it would be 1 to 1 conversion because it's widely known model 3 had the edge especially laters revisions. I was just pointing that am2 own conversion of it's step 2 game fighting vipers 2 was super lack luster as they seemingly just decided to chop chunks off the characters and the stages all together rather than reworking things. Look at genki effort, an outsider company, of a probably more demanding step 2 game Virtua striker 2.
While not exactly the same it's actually super close to it's model 3 source material. It has more aggressive lod and seems to cull players close to the border of the screen more aggressively ( they might just end up suddenly disappearing) and there might be something different in the color depth or lighting but if you put the last model 3 version and the DC version next to each other you can tell it's probably the best model 3 conversion. And it's from step 2 to boot.
 
  • Like
Reactions: Xaeroxcore

MetalliC

Well-known member
Registered
Jun 28, 2019
75
71
18
AGName
MetalliC
AG Join Date
23.04.2014
A really major difference between the two is that the Model 3 had a massively larger amount of memory: ~250MB total, compared to 8MB for the DC. Just the texture RAM alone for the Model 3 was larger than all of the DC's memory. All of those things have effects that go beyond just the number of polygons.
cool story bro! are you traveling to parallel realities or something ? because this reality's Sega Model 3 have such RAMs amount (in MB):
  • texture RAM: 4*2
  • display list RAM: 1
  • culling RAM: 4
  • polygon RAM: 4
  • main RAM: 8

M3's texture RAM is same size as whole DC video RAM, nowhere near "larger than all of the DC's memory" as you say. yes, M3 had more video RAM in total than DC, about twice of it = same as NAOMI1video RAM size.
 

accel99

Well-known member
Registered
Jun 3, 2019
48
44
18
AGName
Accel99
AG Join Date
March 28 2008
cool story bro! are you traveling to parallel realities or something ? because this reality's Sega Model 3 have such RAMs amount (in MB):
  • texture RAM: 4*2
  • display list RAM: 1
  • culling RAM: 4
  • polygon RAM: 4
  • main RAM: 8

M3's texture RAM is same size as whole DC video RAM, nowhere near "larger than all of the DC's memory" as you say. yes, M3 had more video RAM in total than DC, about twice of it = same as NAOMI1video RAM size.

Maybe he meant the vrom? I remember I read somewhere that vrom has all the graphics data( polygons and texture) and ship it off straight the GPU or something like that. I might be remembering wrong though. Isn't that just storage though?
 

Druid II

Well-known member
Registered
Jun 5, 2019
105
68
28
Maybe he meant the vrom? I remember I read somewhere that vrom has all the graphics data( polygons and texture) and ship it off straight the GPU or something like that. I might be remembering wrong though. Isn't that just storage though?
Model 3 used normal ROMs so it's not impossible that a game had 250mbyte (did you not mean mbit?) of graphics in total. But I imagine the graphics chip could only draw what was in its VRAM, not everything at once.

I did not look too hard into either system but I would guess that Model 3 had a small advantage in raw power, but DC chipset was by far more advanced and able to match the Model 3 with far less resources (due to having tiled deferred rendering, texture compression, etc). Also DC draws in higher resolution.
 

accel99

Well-known member
Registered
Jun 3, 2019
48
44
18
AGName
Accel99
AG Join Date
March 28 2008
Model 3 used normal ROMs so it's not impossible that a game had 250mbyte (did you not mean mbit?) of graphics in total. But I imagine the graphics chip could only draw what was in its VRAM, not everything at once.

I did not look too hard into either system but I would guess that Model 3 had a small advantage in raw power, but DC chipset was by far more advanced and able to match the Model 3 with far less resources (due to having tiled deferred rendering, texture compression, etc). Also DC draws in higher resolution.

I was going off of something said by the author of supermodel. Maybe I misunderstood but it almost seems theres " vrom" memory where things already stored in a format to be directly read by gpu? If it does something like this I can see why it would be cumbersome for the Dreamcast to cope with it. But I agree with you in that regard, look at Naomi 2 ,give dedicated tnl hw and it completely outpaces model3.

1. I talked about meshes being uploaded but in fact, Model 3 had very limited memory for this. Only about a megabyte or so. Very few meshes are uploaded this way. The Scud Race and Daytona vehicles (but only the player's) are uploaded because they can be dynamically deformed. But nearly everything else is actually stored in VROM and the VROM is directly connected as a memory region to the Real3D. The CPU can't even see the VROM -- it just passes VROM addresses of meshes and textures to the Real3D, which then can fetch them directly when it needs them. [ /Quote]
 

Make a donation