Quantcast

Page 31 of 32 FirstFirst ... 21272829303132 LastLast
Results 451 to 465 of 480

Thread: Comparison of 6th generation game console hardware

  1. #451
    Hero of Algol TrekkiesUnite118's Avatar
    Join Date
    May 2010
    Age
    32
    Posts
    8,217
    Rep Power
    128

    Default

    Did Carmack even program anything for the Saturn? All of id Software's Saturn ports were outsourced last I checked. I know he programmed for the 32X, is it possible in that interview he mixed the two up?

    EDIT:

    Missed these earlier posts:

    Quote Originally Posted by sheath View Post
    Yep, it is just too bad you are too busy being "right" that you cannot see the flaws in your own argument. What exactly is a polygon or vector according to these emulators?
    The number in PCSX2 was labeled as ppf in earlier versions, or Polygons Per Frame. Primitive is another term for Polygon, and that's what Dolphin uses. NullDC uses the term Vetices, which could be a loose term for polygons or could mean the points for the polygons. If it's the later, the polygon count is even lower.

    The numbers may not be perfect, but it's far more than you've given.

    Quote Originally Posted by sheath View Post
    Can we compare one emulator's output to another?
    Yes, Barone actually did it earlier. The counts were similar.

    Quote Originally Posted by sheath View Post
    Since they are all undoubtedly high level emulators I already know they cannot produce absolutely infallible numbers for the original hardware. Does anybody know how far off they are? Does anybody know whether any version of any 3D system emulator actually has accomplished the feat of showing the world what each game is showing polygon wise, and texture wise, per second or per frame?
    Dolphin's giving a pretty good read out. Yeah, it's a high level emulator, that doesn't mean it magically puts more polygons in the models or tells the CPU and GPU to draw more Polygons than what the compiled code tells it to.

    Quote Originally Posted by sheath View Post
    Their target audience will_not care whether the source material is accurately represented for the original SD output in this case.
    There is an option to run in Native Resolution with no enhancements on all these emulators. Which is what I was running in when I took those shots.

    Quote Originally Posted by sheath View Post
    Claiming that loading emulators and looking at the outputs for an absolute figure on what the systems, or games, were actually producing is the worst disservice to this group and this thread that I have seen yet.
    So you'd rather we pull numbers out of our ass like you? It's not like these emulators magically increase draw distance or make the models stored on the discs have more polygons. The same number of polygons are being calculated by the CPU, and are being sent to the GPU. And as I've stated below, I'm pretty sure they are being culled close to how they are on the real thing since the numbers we are getting are very close to the numbers the developers themselves gave for these games. All that's going to improve here is rendering resolution, and frame rate. Both of those numbers we know what the real systems do for these games already, which is why Barone and I are using games that we know the real framerates for.

    Quote Originally Posted by rusty View Post
    I think in terms of polys being sent to the GPU, an emulator can be pretty spot on. Since were MOSTLY dealing with triangles, it's a good measurement of what a game is pushing through the graphics pipeline. However, that's where it ends. In determining the actual number of triangles that are drawn, all an emulator can do is provide us with a baseline figure to do our estimates, which is why there's this disagreement of the numbers being actual polys rendered or number of polys submitted before culling and rendering happens.

    After looking at the emulators source, I'm pretty certain that it's the numbers of polys submitted before any guard band and back-face culling occurs, so the number of polys actually drawn is around half of that number, perhaps less.

    If you want anything more fine grain than that then an emulator isn't going to help. I think one of the issues here is that Trekkie wants a certain result, and won't look or lacks the knowledge to be able to look any deeper to verify it. Which is ok, because somebody else can always do that digging and propose a counter-argument.
    I would be inclined to believe you if it weren't for the fact that some of the shots Barone posted matched up with what the developers were saying about those games polygon performance. The Rogue Squadron 2 shot has a very high count, it's also a shot where there's not much in the way of those high detail models to cause a lot of culling. If I got right up infront of the X-Wing the polygon count dropped substantially. If I looked at the ships from different angles, the polygon counts changed dramatically as well. If the polygons weren't being culled, I would imagine the numbers wouldn't be fluctuating that much correct?

    The fact is, the Rogue Squadron 2 numbers match up quite nicely with the numbers given by Factor 5 when the game came out:
    http://www.gamespot.com/articles/sta.../1100-2800309/

    From the Article:
    Average Polygons Per Frame: 150,000
    Rebel Ships: 15,000 Polygons
    Enemy Ships: 3,000 to 130,000 Polygons.

    So I think it's safe to say those numbers are pretty damn accurate. In game I was getting 80k to 130k polygons per frame, which is within the number Factor 5 gave.

    If these numbers are not being culled then that means that some of these games are getting into unbelievably low polygon counts. Using the estimate you gave of half or less, that would mean Daytona USA on the Dreamcast would be in the 100,000-600,000 polygons per second range in that case. That's 1,000-10,000 polygons per frame. That's Saturn and PS1 level polygon counts. Metal Gear Solid 2 would be in the 900,000 Polygons per second range, which you claimed was pumping out tons of polygons. Saying these counts aren't accurate and are most likely half or less to disprove one of the most impressive polygon pushing games of that generation falls apart when you try to apply that logic to the more average games.
    Last edited by TrekkiesUnite118; 04-04-2014 at 12:01 AM.

  2. #452
    ESWAT Veteran Team Andromeda's Avatar
    Join Date
    Jul 2010
    Location
    Wales, UK
    Posts
    5,776
    Rep Power
    64

    Default

    Quote Originally Posted by stu View Post
    sigh.

    Well back to normal programming.
    Well C'Mon . To make out that SEGA Japan went behind SEGA Europe back and secretly shipped to CORE wit out SOE knowing is pushing it . If they were 'such' secrecy . 1) CORE would have never talked about in an Interview with one of the biggest selling UK mag's at the time, never mind they was also an SEGA PR man as part of the Interview.

    Did Carmack even program anything for the Saturn
    No he never, that doesn't mean comments are without merit though . Come to think of it , I don't think he wrote anything for the PS either . That said I remember Peter Molyneux saying all 16 bit consoles were utter rubbish and he be glad to see the back of them. So we can all cherry pick the odd interview I guess
    Panzer Dragoon Zwei is
    one of the best 3D shooting games available
    Presented for your pleasure

  3. #453

    Join Date
    Jan 2014
    Location
    Frankfurt, Germany
    Posts
    181
    Rep Power
    0

    Default

    Quote Originally Posted by TrekkiesUnite118 View Post
    I would be inclined to believe you if it weren't for the fact that some of the shots Barone posted matched up with what the developers were saying about those games polygon performance. The Rogue Squadron 2 shot has a very high count, it's also a shot where there's not much in the way of those high detail models to cause a lot of culling. If I got right up infront of the X-Wing the polygon count dropped substantially. If I looked at the ships from different angles, the polygon counts changed dramatically as well. If the polygons weren't being culled, I would imagine the numbers wouldn't be fluctuating that much correct?
    All I can say is to look at the source yourself. Unless the emulator is going through the triangle list, transforming them into clip space and counting the front-facing triangles before submitting to the GPU (which I can't see), I can't think of any other way it would be counting the tris.

    How exactly are you looking at the ships from different angles? By moving around the ship with the camera? Because there could be a number of reason for the fluctuations.


    Quote Originally Posted by TrekkiesUnite118 View Post
    The fact is, the Rogue Squadron 2 numbers match up quite nicely with the numbers given by Factor 5 when the game came out:
    http://www.gamespot.com/articles/sta.../1100-2800309/

    From the Article:
    Average Polygons Per Frame: 150,000
    Rebel Ships: 15,000 Polygons
    Enemy Ships: 3,000 to 130,000 Polygons.

    So I think it's safe to say those numbers are pretty damn accurate. In game I was getting 80k to 130k polygons per frame, which is within the number Factor 5 gave.
    Yup, they do. However, by "triangles rendered" I think they really meant "triangles submitted" because at the 150,000 triangle average, your average triangle size is pretty small, assuming that they are un-textured and also un-lit, using one TEV stage, and you're running the fill-rate at it's maximum theoretical limit.


    Quote Originally Posted by TrekkiesUnite118 View Post
    If these numbers are not being culled then that means that some of these games are getting into unbelievably low polygon counts. Using the estimate you gave of half or less, that would mean Daytona USA on the Dreamcast would be in the 100,000-600,000 polygons per second range in that case. That's 1,000-10,000 polygons per frame. That's Saturn and PS1 level polygon counts. Metal Gear Solid 2 would be in the 900,000 Polygons per second range, which you claimed was pumping out tons of polygons. Saying these counts aren't accurate and are most likely half or less to disprove one of the most impressive polygon pushing games of that generation falls apart when you try to apply that logic to the more average games.
    So for things like the PS2, I hit on this on the other thread. Where on the GC, you relied on Flipper to do the culling for you, it worked differently on the PS2. You did that at the vertex transform stage on VU1. Sony from the very beginning recommended that you didn't bother doing the calculation (and so we didn't) because the triangles would be rendered in both contexts (rendering two triangles) in less time than it would take to perform the calculation. And that calculation isn't difficult, it's just a 2d cross product. Given a triangle in clip-space with vertices v0, v1, v2 you would calculate;

    d0 = v1 - v0;
    d1 = v2 - v0;

    direction = (d0.x - d1.y) * (d0.y - d1.x);

    The sign of direction would tell you if the vertices are clock-wise or counter clock-wise. And on VU1, that's only a few clock-cycles. What emulator have you been using? I wouldn't mind looking into that and mucking around with the rendering pipe-line on MGS2

  4. #454
    Hero of Algol TrekkiesUnite118's Avatar
    Join Date
    May 2010
    Age
    32
    Posts
    8,217
    Rep Power
    128

    Default

    As I mentioned before the PS2 Emulator is PCSX2.

    As for Rogue Squadron 2, the ships are using very small triangles. You can see it in the Developer video on the game disc. Here is a very realistic situation that can happen in that game:

    Your ship with 15,000 polygons (let's say 7,000 for an estimate of culled polygons) is sitting between two star destroyers that are very close together. Each Star Destroyer model is 130,000 polygons. So if the count get's dropped in half for culling, that's still 130,000 polygons between the two of them. So we're already at close to 140,000 culled polygons and that's not even including the oodles of enemy ships and allied fighters flying around you as well as the lasers, shadows, lights, and background objects going on.

    This is a very realistic scenario that happens in the Battle of Endor stage. And it's right in line with that 150,000 number Factor 5 gave. After playing it for hours I have no problem believing the numbers Factor 5 gave or that I got from Dolphin. The game is using tons of polygons for ship detail. I have to ask, have you ever played this game or seen it in action?

    You keep saying this is the max theoretical limit, but considering the official specs of the system are listed as realistic specs and give polygon performance as up to 12 million with all effects, I think you need to go back and rethink things. I'm not doubting your PS2 knowledge, I'm doubting your Gamecube knowledge. From what you've posted it seems like you really did try to push the PS2 very hard in the games you made for it, but it seems that all you did on Gamecube was quick PS2 ports. It doesn't seem like you really tried to push that system like you did with the PS2. So sorry but I'm going to take the word of developers like Factor 5 who really pushed that system before I take your word that they were going above some theoretical limit.

  5. #455

    Join Date
    Jan 2014
    Location
    Frankfurt, Germany
    Posts
    181
    Rep Power
    0

    Default

    Quote Originally Posted by TrekkiesUnite118 View Post
    As I mentioned before the PS2 Emulator is PCSX2.
    I thought that it was. I just wanted to check.

    Quote Originally Posted by TrekkiesUnite118 View Post
    As for Rogue Squadron 2, the ships are using very small triangles. You can see it in the Developer video on the game disc. Here is a very realistic situation that can happen in that game:

    Your ship with 15,000 polygons (let's say 7,000 for an estimate of culled polygons) is sitting between two star destroyers that are very close together. Each Star Destroyer model is 130,000 polygons. So if the count get's dropped in half for culling, that's still 130,000 polygons between the two of them. So we're already at close to 140,000 culled polygons and that's not even including the oodles of enemy ships and allied fighters flying around you as well as the lasers, shadows, lights, and background objects going on.

    This is a very realistic scenario that happens in the Battle of Endor stage. And it's right in line with that 150,000 number Factor 5 gave. After playing it for hours I have no problem believing the numbers Factor 5 gave or that I got from Dolphin. The game is using tons of polygons for ship detail. I have to ask, have you ever played this game or seen it in action?
    Yeah, I have played them. I just didn't think they were anything special. I do know that there is some quite aggressive level of detailing going on (where lower res models are used for objects) and you often won't see the full-res model until it's quite close.


    Quote Originally Posted by TrekkiesUnite118 View Post
    You keep saying this is the max theoretical limit, but considering the official specs of the system are listed as realistic specs and give polygon performance as up to 12 million with all effects, I think you need to go back and rethink things.
    Ok, so you make a good point. However, that "realistic" specs thing is a bit muddy - it has more to do with how Flipper process vertices than anything else and was hard for the ArtFX guys to really quantify; partially because I think (well...after talking to certain people, I KNOW this) they lacked real world data and went for a conservative estimate. The fill-rate is the finite barrier here. I used in my own calculation (Flipper speed * number of pixel pipes), and the theoretical spec is merely raw pixel bandwidth. It doesn't take into account for things like the cost of triangle setup, or texture fetch latencies.

    Quote Originally Posted by TrekkiesUnite118 View Post
    I'm not doubting your PS2 knowledge, I'm doubting your Gamecube knowledge. From what you've posted it seems like you really did try to push the PS2 very hard in the games you made for it, but it seems that all you did on Gamecube was quick PS2 ports. It doesn't seem like you really tried to push that system like you did with the PS2. So sorry but I'm going to take the word of developers like Factor 5 who really pushed that system before I take your word that they were going above some theoretical limit.
    I can see where you're coming from from on this, and I'm not going to contest this too much because it sort of becomes about me, rather than the hardware. I'd like to avoid that. I will say that I never did a quick PS2 port on the Gamecube, though. And I never, ever settle for sub-standard work.

  6. #456
    Hero of Algol TrekkiesUnite118's Avatar
    Join Date
    May 2010
    Age
    32
    Posts
    8,217
    Rep Power
    128

    Default

    Quote Originally Posted by rusty View Post
    Yeah, I have played them. I just didn't think they were anything special. I do know that there is some quite aggressive level of detailing going on (where lower res models are used for objects) and you often won't see the full-res model until it's quite close.
    Yeah, it does do that on the fighters. On the larger ships like the Star Destroyers it doesn't seem to be nearly as aggressive. In Battle of Endor it's very possible to have both Star Destoryers in view with rather high detailed models being used.



    Quote Originally Posted by rusty View Post
    Ok, so you make a good point. However, that "realistic" specs thing is a bit muddy
    Yes, it is a bit muddy. It's because the figures are on a conservative low end. There were quite a few developers at the time saying 20-30 million polygons was reasonable for a realistic figure. Factor 5 was one of those developers.

    Quote Originally Posted by rusty View Post
    The fill-rate is the finite barrier here. I used in my own calculation (Flipper speed * number of pixel pipes), and the theoretical spec is merely raw pixel bandwidth. It doesn't take into account for things like the cost of triangle setup, or texture fetch latencies.
    I get what you're saying about the fillrate, but what I'm asking is are you sure that those 150,000 polygons per frame is actually getting to close to the pixel fill rate boundary? What calculation are you using to determine that?

    Quote Originally Posted by rusty View Post
    I can see where you're coming from from on this, and I'm not going to contest this too much because it sort of becomes about me, rather than the hardware. I'd like to avoid that. I will say that I never did a quick PS2 port on the Gamecube, though. And I never, ever settle for sub-standard work.
    Sorry, I'm not trying to make it personal, I'm just saying from what you are posting that you come of as much more experienced with the PS2 than you do with the Gamecube. Out of curiosity what games did you work on for the Gamecube?

  7. #457

    Join Date
    Jan 2014
    Location
    Frankfurt, Germany
    Posts
    181
    Rep Power
    0

    Default

    Quote Originally Posted by TrekkiesUnite118 View Post
    Yes, it is a bit muddy. It's because the figures are on a conservative low end. There were quite a few developers at the time saying 20-30 million polygons was reasonable for a realistic figure. Factor 5 was one of those developers.
    Yeah, given the vertex throughout on the GC that wouldn't surprise me. However, I think it's important to state that it would probably be more accurate to say that Flipper can transform the vertices for 20-30 million triangles. And this is much much higher than Nintendo's figures. They gave something like 32 million VERTICES with only a single colour, and 20 million VERTICES for 1 light source and 1 texture. Which would only be ~10 and 7 million triangles per second as peek performance.


    Quote Originally Posted by TrekkiesUnite118 View Post
    I get what you're saying about the fillrate, but what I'm asking is are you sure that those 150,000 polygons per frame is actually getting to close to the pixel fill rate boundary? What calculation are you using to determine that?
    I gave it above. I took the speed of Flipper in Mhz and multiplied it by the number of pixel pipelines. So that's 162Mhz multiplied by 4 which is 648 megapixels/second. Divide that by 60fps and 150,000 triangles, that gives us 72 pixels for each triangle or a 16 by 9 pixel triangle. However, all of that assume absolute optimal throughput without any stalls.

    You could do more than 150,000 triangles per frame, if you rendered smaller triangles. But at that point, you're also spending more time in triangle clipping, culling and setup. I've seen a few developers really get killed on performance (not just on GC, but modern platforms too) because they assumed that just because a triangle wasn't being rendered, that it didn't cost anything to submit.

    It's worth point out that in terms of pixel/texel throughout, all of Nintendo's numbers assumed a 200Mhz GPU speed. So 1 texture was 800MegaPixels a second. Given that there were 4 pipelines, it's obvious that the number was calculated by multiplying the clock speed by the number of pixel pipes.


    Quote Originally Posted by TrekkiesUnite118 View Post
    Sorry, I'm not trying to make it personal, I'm just saying from what you are posting that you come of as much more experienced with the PS2 than you do with the Gamecube. Out of curiosity what games did you work on for the Gamecube?
    No need to be sorry. So this is the bit where I have to say, I'm not willing to share. I sort of value what little anonymity that I have. I did a fair bit of GC stuff, but all of it wasn't released. Only two PS2 titles (out of six) that I've worked on were never released. One employer didn't have a very good relationship with Nintendo, and another, well, we know the story with the last time that I looked at the GC. I think at this point, you'd be quite right to be skeptical but you're right in saying that my GC experience is nowhere near as in-depth as PS2.

    I've also done a lot of Wii stuff, but again, nothing released since my employer canned the stuff we were doing.
    Last edited by rusty; 04-04-2014 at 03:53 PM.

  8. #458
    Hero of Algol TrekkiesUnite118's Avatar
    Join Date
    May 2010
    Age
    32
    Posts
    8,217
    Rep Power
    128

    Default

    Quote Originally Posted by rusty View Post
    I gave it above. I took the speed of Flipper in Mhz and multiplied it by the number of pixel pipelines. So that's 162Mhz multiplied by 4 which is 648 megapixels/second.
    I got that, I was asking for the one below.

    Quote Originally Posted by rusty View Post
    Divide that by 60fps and 150,000 triangles, that gives us 72 pixels for each triangle or a 16 by 9 pixel triangle. However, all of that assume absolute optimal throughput without any stalls.
    And why would that be a bad thing? That's only half what you'd have for a game running at around 75,000 polygons per frame. The resolution here is 640x480, so a polygon that has an area of 72 pixels wouldn't be that much of an issue in terms of how big it appeared on screen I'd assume. And what if the sizes were mixed, such as some bigger than 72 pixels and some smaller than 72 pixels. When you look at the models for Rogue Squadron the polygons get really tiny in the more detailed areas of the models, so I wouldn't be surprised that's where those high numbers are coming from. Not to mention RS2 isn't doing that many advanced effects, or using lots of different models. It's just pumping out a lot of polygons at a smooth framerate and reusing a lot of models.

    To bring it back to the camera issue, in the hangar if I stood right infront of the X-Wing, the polygon number dropped to below 20,000 per frame. So that would be the scenery, one piece of the X-Wing, and the pilot. That's not unbelievable. If the polygons weren't being culled wouldn't that number be much higher as it would be including all the polygons for the X-Wing as well as all the objects behind it? Behind it we would have another X-Wing and at least 2 Y-Wings with a lot more scenery. If they polygons weren't being culled wouldn't those high detail models increase that number?

  9. #459
    Master of Shinobi
    Join Date
    Sep 2012
    Posts
    1,284
    Rep Power
    38

    Default

    Quote Originally Posted by stu View Post
    Saturn was nuts
    But that's what made it great!

  10. #460
    Hero of Algol
    Join Date
    Aug 2010
    Posts
    7,816
    Rep Power
    177

    Default

    Quote Originally Posted by sheath View Post
    Claiming that loading emulators and looking at the outputs for an absolute figure on what the systems, or games, were actually producing is the worst disservice to this group and this thread that I have seen yet.
    High-level emulated and possibly inaccurate numbers are still more relevant than numbers created out of your guesswork; but of course you claim otherwise.
    Yeah, right, providing non-"sheath certified" info is always a disservice...

  11. #461

    Join Date
    Oct 2014
    Posts
    5
    Rep Power
    0

    Default

    The best way to look for polycounts , is to rip meshes/models directly from the games, imho. There are plenty of tools to do so...

    Looking at some models most would be surprised.
    Last edited by Esppiral; 02-09-2015 at 04:25 PM.

  12. #462
    Hero of Algol
    Join Date
    Aug 2010
    Posts
    7,816
    Rep Power
    177

    Default


  13. #463
    Hero of Algol TrekkiesUnite118's Avatar
    Join Date
    May 2010
    Age
    32
    Posts
    8,217
    Rep Power
    128

    Default

    I stumbled across this recently. It's a video of F-Zero GX drawing only wireframes:



    Once again, it think it's safe to assume this game is pushing a buttload of polygons.

  14. #464
    Raging in the Streets Yharnamresident's Avatar
    Join Date
    May 2013
    Location
    British Columbia
    Posts
    4,080
    Rep Power
    65

    Default

    Quote Originally Posted by Barone View Post
    *Dreamcast dev video*
    Looking at his other videos, pretty impressive what hes got going on in this:



    The rendering is kinda screwy in his engine. But, looks like a small step towards Skyrim on the Dreamcast.

    Quote Originally Posted by TrekkiesUnite118 View Post
    I stumbled across this recently. It's a video of F-Zero GX drawing only wireframes:



    Once again, it think it's safe to assume this game is pushing a buttload of polygons.
    Its mind blowing they also got it running at 480p and 60 FPS.

    Like I said in the other thread, it sucks they put so much effort into it and it only sold 500,000.
    Certified F-Zero GX fanboy

  15. #465
    I remain nonsequitur Shining Hero sheath's Avatar
    Join Date
    Jul 2010
    Location
    Texas
    Age
    42
    Posts
    13,313
    Rep Power
    131

    Default

    Wow, this thread has been dead for a while.






    "... If Sony reduced the price of the Playstation, Sega would have to follow suit in order to stay competitive, but Saturn's high manufacturing cost would then translate into huge losses for the company." p170 Revolutionaries at Sony.

    "We ... put Sega out of the hardware business ..." Peter Dille senior vice president of marketing at Sony Computer Entertainment

Thread Information

Users Browsing this Thread

There are currently 1 users browsing this thread. (0 members and 1 guests)

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •