You're kidding right? MGS4 has some of the most impressive character models this generation...and ingame (microwave)...you're telling me that's only 10,000 polys. Not sure if that's impressive or not..lol
"Yes, but how many polygons?" An artist blog entry with interesting numbers
- Thread starter Farid
- Start date
You're kidding right? MGS4 has some of the most impressive character models this generation...and ingame (microwave)...you're telling me that's only 10,000 polys. Not sure if that's impressive or not..lol
He was truly excited with the potential of the hardware at the time. If I'm not mistaken (and someone please correct me if I am), there were 2 "downgrades" until the final version, the first being the introduction of the actual RSX (pre-E3 2006 I believe) and somewhat closer to US/JP release, 50MHz downgrade in both the GPU core and memory clock frequency.
I'd like to know how strong an impact the reduction in clock frequency was... Only now I noticed this is the "how many polygons thread" and it seems I'm veering into the off-topic though...
Well lets hear from the mods what they think. Perhaps they can make another thread and move the related posts in there.
I was wondering the same thing about the clock speed. I think some other titles were also affected by the hardware changes. I think Heavenly Sword too. Colors in the original trailer were more vivid, AF was better, and detail was somewhat more visible. Enemy models probably had a small reduction in polygons as well I think.
Come to think of it, how many polygons where the Heavenly Sword models?
I was wondering the same thing about the clock speed. I think some other titles were also affected by the hardware changes. I think Heavenly Sword too. Colors in the original trailer were more vivid, AF was better, and detail was somewhat more visible. Enemy models probably had a small reduction in polygons as well I think.
Come to think of it, how many polygons where the Heavenly Sword models?
jandlecack
Banned
But the picture on the left was from the 2006 demo, at the time, the PS3 definitely had the RSX. Do you think the further downgrades so to speak were that significant? I doubt the clock speed makes that much of a difference and were there PS3 prototypes with two Cells at all?
In any case, I highly doubt Old Snake - at least in realtime cutscenes - is less than 30000 polygons.
By the way, what do we think the chances are Sony ups the GPU speeds in a later firmware update optionally for developers to use? Sorta like with the PSP. I think that's quite likely. Again, 50MHZ really aren't much, and the way the PS3 is built, I'm actually surprised they didn't keep that speed.
Why are you using a cutscene model vs a gameplay model. In the final version there is quite a difference in both lighting, mapping and ploy count between cutscenes and gameplay models.
Here's a more accurate comparison which shows that they are more similar than some believe:
edit: here's a new one i just made with my camera
Last edited by a moderator:
No, the pic on the left is from the TGS 2005 real-time trailer. I'm not sure about the framerate but from what I read it was 60 fps, besides having quite the poly count, and amazing effects (environmental like dust, debris, and smoke from explosions).
jandlecack
Banned
Well, no doubt the prototype PS3s were more powerful. Kojima has said that they had a lot of trial and error because the hardware wasn't finalized. But being so early in the development, I really, REALLY doubt the downgrades were that astonishing to look at. I mean, the product they had at the time can't have been that advanced, given the time period.
Essentially, the hardware was more powerful than it is now, but the software at the time was nowhere near as optimized as it is now. Meaning it's very much what you should have expected.
And to be frank, from my recollection the cutscene model of Snake looks a helluva lot sharper & better than in the 2005 TGS tech-demo. I do recall they said they render every stream of hair individually, but are we sure that has changed? Looking at the later posted comparison.
Essentially, the hardware was more powerful than it is now, but the software at the time was nowhere near as optimized as it is now. Meaning it's very much what you should have expected.
And to be frank, from my recollection the cutscene model of Snake looks a helluva lot sharper & better than in the 2005 TGS tech-demo. I do recall they said they render every stream of hair individually, but are we sure that has changed? Looking at the later posted comparison.
Why are you using a cutscene model vs a gameplay model. In the final version there is quite a difference in both lighting, mapping and ploy count between cutscenes and gameplay models.
Here's a more accurate comparison which shows that they are more similar than some believe:
edit: here's a new one i just made with my camera
I wonder if the in game model are different than the cut scene models. Sometimes the cutscenes shifted to gameplay smoothly while retaining the detail.
I also tried in some sections to zoom very close to the character in narrow spaces. In act 5 I begun blowing up snake to make him take different poses and rotated the camera near objects to get a good close up on him. The detail was astonishingly similar.
I wonder if it is the cinematic camera angles that sometimes make the model look more detailed nin terms of polygon detail
Edit: Personally the biggest difference I see between the original demo and the final game is the lighting. THAT made a LOT of difference. In that very same environment when Snake ran with MKII into the battlefield, the marble on the floor was shiny and the sun's light reflected on it. On the final built this was absent. Self shadowing was toned down and the shiny parts were also removed. I think there were also a few other objects in the room but I am not sure
Last edited by a moderator:
Why are you using a cutscene model vs a gameplay model. In the final version there is quite a difference in both lighting, mapping and ploy count between cutscenes and gameplay models.
Here's a more accurate comparison which shows that they are more similar than some believe:
http://www.ziua-nuntii.ro/fun/Before-After-Hair.jpg
edit: here's a new one i just made with my camera
http://www.ziua-nuntii.ro/fun/Snake2.jpg
You are completly missing the point, the image was to show the hair difference only, not character model. Its obvious that the older model had volumetric hair made out of straws while the newer has maybe 10-20 hair texture blocks bent over the head. :smile:
Why would I be kidding? There are lots of obvious polygon edges on all MGS4 characters even in the cutscenes. Actual gameplay models are even more low res, particularly the enemies.
I reserve the right to decide which game has the most impressive character models, by the way... Personally, I'm more impressed with stuff in Uncharted, and on a simple detail level, UT3/Gears is still quite beyond the rest (even though I could easily list some artistic weaknesses).
And don't forget that 10K-20K is an awful lot of geometry. Quake3 models were on average only 800 triangles...
Laa Yosh. I doubt very much they are 10,000...and that you "guessed" that when you yourself said "no, Snake IS only 10-15,000" just adds that you're trying to impress upon your opinion that MGS4 models aren't that impressive. When I look at Snake's model ingame i don't see polygon edges - I see a very well rendered model.
jandlecack
Banned
Whatever the case, it's all speculation at this point, let's not argue about it.
@Laa-Yosh: Uncharted has incredible textures but the models themselves are a bit blurry for my taste, and don't look like a high poly count at all. Same for Gears. They just seem too soft/unsharp. Don't think any of those can match MGS4's models at the moment.
@Laa-Yosh: Uncharted has incredible textures but the models themselves are a bit blurry for my taste, and don't look like a high poly count at all. Same for Gears. They just seem too soft/unsharp. Don't think any of those can match MGS4's models at the moment.
man, i agree as well, i even posted a reply a few posts back presenting some tehnical arguments but i edited it as this topic isn't about X vs Y, so let's focus on the actual topic, which is presenting confirmed poly numbers (which we don't have yet for MGS sadly).
Honestly Deepbrown, maybe you are misstaking the mapping for actual true geometry? In this picture you can see lots of hard edges on his arm, shoulder, rear, legs etc for example, you see it?
Example image
Of course mapping is there to lure the brain to believe things are made out of more polygons than they actually are but it doesnt change the fact that its not true geometry.
Drake was something around 25-30k polygons acording to a ND dev. But as said mapping does a lot to fool the brain that its real geometry when its not. And by doing it means its good mapping, good design placing the polygon budget in the most important/visible parts and artwork! :smile:
In a thread on poly counts, it makes sense to hear what poly counts are being used and comparing poly budgets with what we see on screen. It's also great to listen to a seasoned expert in the field with a wealth of experience behind them, who can chime in when we don't have real numbers with best-guesses based on previous experience. Until real evidence to the contrary, I'd happily take Laa-Yosh's ball-park figures, educated guesses, as indicative of final in game amounts. And from that, if the models look great, we can then question why such-and-such a low/high polygon model looks good.
How on earth can models be blurry? Models are defined by a series of triangles, mathematically defined with perfect edges and surfaces. Any 'blur' on the models has to be produced in the pixel space, such as adding a blur filter, upscaling, bloom, or other post-process. A model looking sharp or not has nothing to do with polygon density. Polygon density will only affect silhouettes and presence of unnatural angles in curves, assuming normal mapping to add virtual geometry detail.
Yes, well, those are gameplay screens with the gameplay model. The cutscene model for Snake is at leats twice as complex in geometry, lighting and mapping.
Is the model in the pic you posted even close to this one ?
i don't think so, look at the hands, look at the shape of the face, it's not only the improvements in mapping and lighting but in the actual raw geometry. Not to mention the scheletal superiority of the model in the cutscenes which allows complex facial animations, body movement and interaction.
jandlecack
Banned
What I meant to express was that even if the poly count is high, the models still have that blurry-muddy look to them. You can call it counterbalancing the poly advantage or whatever.
Take a look at this picture: http://pub.tv2.no/multimedia/TV2/archive/00447/Uncharted_Drake_s_F_447515g.jpg
Although it's most likely doctored, the entire gameplay, at least for me, has that muddy colour-mash on models, especially their faces. In a nutshell, Uncharted's models aren't as sharp or detailed as they should be with that poly count. They just look halfway SDTV to be honest.
@Nebula: Those shots, look ridiculous. Honestly. What's with the grainy, dithered picture? That doesn't do the graphics justice in the slightest.
Take a look at this picture: http://pub.tv2.no/multimedia/TV2/archive/00447/Uncharted_Drake_s_F_447515g.jpg
Although it's most likely doctored, the entire gameplay, at least for me, has that muddy colour-mash on models, especially their faces. In a nutshell, Uncharted's models aren't as sharp or detailed as they should be with that poly count. They just look halfway SDTV to be honest.
@Nebula: Those shots, look ridiculous. Honestly. What's with the grainy, dithered picture? That doesn't do the graphics justice in the slightest.