Revolution Tech Details Emerge ( Xbox1+ performance, 128 MB RAM )

[OtakingGX]

It's not limited at all. Unless there are any sneaky ways to circumvent the issue entirely, it's perfectly possible to fill the buffer to capacity, then draw what's in it, fill another buffer (possibly while drawing the first one), then draw that, etc.

I found a better article on the subject:

http://www.beyond3d.com/articles/tilebasedrendering/index1.php

From what I can tell, you need to fill the scene buffer with triangles first. While this is happening you're recording into each tile buffer which triangle is in it. Now you sort the tile, then start rendering. So the scene buffer needs to be filled, otherwise you could render a tile before the CPU has put a triangle in the scene buffer.

It's clear that the tile buffers are located on-chip. I can't tell if the scene buffer is or is not. I think this is where the notion that too many triangles would fill up an on-chip buffer came from. If that buffer is in graphics memory, then you'd have a lot more space to play with for triangles.

 

[Guden Oden]

From what I can tell, you need to fill the scene buffer with triangles first.

Well, um, that's exactly what I said...

It's clear that the tile buffers are located on-chip.

Correct, the tile buffers are, but the scene buffers (that hold the geometry) are not. These sit in framebuffer memory and were in the case of the kyro fairly arbitrarily set to like 6MB from what I recall of discussions here at the time (very long ago now). It just seemed like a sufficient figure considering software available when the chip was current. I suppose people just didn't consider that it would be possible to multipass, and hence thought there was some hard limit to the size of the scene buffer.

Then again, some people just don't plain WANT to see solutions for hardware that they don't like.

 

[OtakingGX]

Correct, the tile buffers are, but the scene buffers (that hold the geometry) are not. These sit in framebuffer memory and were in the case of the kyro fairly arbitrarily set to like 6MB from what I recall of discussions here at the time (very long ago now). It just seemed like a sufficient figure considering software available when the chip was current. I suppose people just didn't consider that it would be possible to multipass, and hence thought there was some hard limit to the size of the scene buffer.

Then again, some people just don't plain WANT to see solutions for hardware that they don't like.

Oh, don't get me wrong, I owned and loved a Kyro II until I upgraded to my Radeon 9700. I was holding out for Series 4 for the longest time, then Series 5 when that got canned. Discussing the implications that future TBDRs held, long ago, someone much wiser than me mentioned the limitations of the scene buffer and how you could fit only so much polygon data in there. I guess it was a feature unique to the Kyro.

If PowerVR announced they were returning to PCs tomorrow I'd probably jump for joy at the news. I might be bitter because of unfulfilled promises, but I'd still come back if they gave me the chance.

 

[swaaye]

http://assemblergames.com/forums/showthread.php?t=7241

 

[Teasy]

Correct, the tile buffers are, but the scene buffers (that hold the geometry) are not. These sit in framebuffer memory and were in the case of the kyro fairly arbitrarily set to like 6MB from what I recall of discussions here at the time (very long ago now).

Correct, Kyro used a 3MB pointer buffer and 3MB vertex buffer in video memory. You could actually change the size of the buffers in the driver registery settings as well. Though I never saw a game profile (in the driver) that actually used an increased size pointer/vertex buffer, 3MB each was more then enough for any game at the time.

 

[OtakingGX]

http://assemblergames.com/forums/showthread.php?t=7241

That's really quite astonishing. Where did the ~$14M cost to develop a major title come from?

Hopefully this trend can be reversed. If it doesn't, all the little developers will either get out of the business, or be bought up by big publishers, causing the industry to stagnate. Maybe I'm just being taken in by the hype, but hopefully Nintendo will play a part in driving costs down.

I'm not sure if the sample size what sufficient to give us accurate numbers, but according to Gamespot Nintendo dropped the average cost of game development from GBA to DS by 30%.

http://www.gamespot.com/news/2005/08/12/news_6130901.html

 

[swaaye]

That's really quite astonishing. Where did the ~$14M cost to develop a major title come from?

Hopefully this trend can be reversed. If it doesn't, all the little developers will either get out of the business, or be bought up by big publishers, causing the industry to stagnate. Maybe I'm just being taken in by the hype, but hopefully Nintendo will play a part in driving costs down.

I'm not sure if the sample size what sufficient to give us accurate numbers, but according to Gamespot Nintendo dropped the average cost of game development from GBA to DS by 30%.

http://www.gamespot.com/news/2005/08/12/news_6130901.html

This is why I think Nintendo is on to something. Whether it will go anywhere, who knows...

It's just that the more we demand pretty graphics, the more these games become a task to recreate an entire virtual world. It's not something that can be done with a small team anymore. The games aren't about the underlying gameplay, but are about how everything comes together in a pretty presentation because that is all that will sell. I suppose that is how it has always been, but tech is obviously a wee bit different now than it was in the SNES days.

With these games costing nearly as much as movies you are NOT going to see many companies taking risks with new franchises. You are going to see sequels. Hell look at the launch titles for 360! They are almost all sequels. Companies can't take risks in an environment like this. It would be almost stupid of them to do so because predicting that a game will appeal to gamers is damn near impossible. Assuming the game idea is friggin awesome, the suits would have to feel comfortable that technically the game would turn out fantastic. How can you do that? Hell every game that is being developed is basically uncharted technical territory. Even if they are using a middleware engine. How can it get more risky!? If the game isn't a huge success they will never recoup the devel costs.

Scary where the industry is going.

 

[Corwin_B]

Hell look at the launch titles for 360! They are almost all sequels.

Not true ! Perfect Dark Zero is a prequel.

 

[fearsomepirate]

I suppose that is how it has always been, but tech is obviously a wee bit different now than it was in the SNES days.

I remember the first time I played Jungle Hijinks (1st level of DKC), jumping to the top of the trees, and the camera slowly panned up as rows and rows of trees parallaxed in the background. I imagined a whole unexplored jungle being back there. Today, we don't want to just see something that makes us imagine the world; we want to go see the world. That takes a hell of a lot more work than parallaxing a few gifs.

I suppose content creation tools are going to advance by leaps and bounds in the next generation, to where instead of seeing 500 identical orcs on a field, we see 500 unique orcs, each being randomly modified from a set of basic orc parts.

I also think it would be an interesting tactic to base an RPG or something around a randomly generated world instead of taking tons of time building every piece yourself. Of course, look at the 360 launch line-up...the games aren't huge. They're rather short and highly detailed. I think what we're going to see are a lot of beautiful games that can be beaten in 6 to 10 hours.

Of course, it's not like it took 2 weeks to beat Mario 3. It did take that long to learn all the tricks necessary to beat it in a few hours, though. If you can't make enough content for more than 6 hours of play, you're going to have to find other ways to extend the playtime. And unlocking concept art should not be one of those ways.

 

[OtakingGX]

Of course, it's not like it took 2 weeks to beat Mario 3. It did take that long to learn all the tricks necessary to beat it in a few hours, though. If you can't make enough content for more than 6 hours of play, you're going to have to find other ways to extend the playtime. And unlocking concept art should not be one of those ways.

Well, there goes my idea for Concept Art Grab-a-rama for PS3...

 

[pc999]

Hell look at the launch titles for 360! They are almost all sequels. Scary where the industry is going.

And the "original" ones (ie new name/look), does have a really big name from the past like Too Human or Prey, or in a different maner GoW which is build in the Unreal name and yet it as a relativelly small team (but their tech is "almost free" and I wonder how much they share with the UT team in terms of parts of maps , textures, terrein etc...).

I dont want to wait 3 years for the next Rare game or Halo or UT (they even said that UT is finished they are just doing the "art" part of the game but you could play it already and I want o play it now).

IMO this will/is kill/ing the industry in a slow death even the most hardcore FPS will want one more gen of equal games, and to get different games costs most go down.

People usually forget that it is those how we dont expect that suprise us.

I really hope that Nintendo can do this even if it tradeoff is the detail on games (and really I dont think that we need (much?) more detail than in the best GC/XB games), if the others are able too do that too would be great.

 

[mckmas8808]

With these games costing nearly as much as movies you are NOT going to see many companies taking risks with new franchises. You are going to see sequels. Hell look at the launch titles for 360! They are almost all sequels. Companies can't take risks in an environment like this. It would be almost stupid of them to do so because predicting that a game will appeal to gamers is damn near impossible. Assuming the game idea is friggin awesome, the suits would have to feel comfortable that technically the game would turn out fantastic. How can you do that? Hell every game that is being developed is basically uncharted technical territory. Even if they are using a middleware engine. How can it get more risky!? If the game isn't a huge success they will never recoup the devel costs.

Scary where the industry is going.

Heavenly Sword, Lair, Motorstorm, Bladstorm, Fatal Inertia, Dark Sector, Eyedentify, and the new IP that Team Bondai is doing. And those games are for only the PS3 (well except Dark Sector). Then there's more for the Xbox 360. So where are you sounding so scared for?

 

[swaaye]

Heavenly Sword, Lair, Motorstorm, Bladstorm, Fatal Inertia, Dark Sector, Eyedentify, and the new IP that Team Bondai is doing. And those games are for only the PS3 (well except Dark Sector). Then there's more for the Xbox 360. So where are you sounding so scared for?

I'm glad there are companies planning original games. That's great. There are lots of little behind-the-scenes games that are amazingly creative. They don't get attention by most gamers because a) they aren't usually state-of-the-art and b) they don't get media attention. The little PDA game companies come to mind, among others.

But, realistically if one of those above mainstream bigtime games fail, the company that made it is going to be in trouble. How can a company eat $10+ million?

For example, I visited Raven Software a couple of years ago. They were working on Jedi Academy, X-Men, and Quake4 at the time. (They absolutely would not show me Q4, but I saw some concept art when chatting with a guy). I told them I loved their Star Wars games and that they had a great formula going with JK2 (I hadn't played JA yet). They said they were done with SW because LucasArts (the publisher) hadn't hired them for another title. I also mentioned that I really had enjoyed their Heretic/Hexen games and questioned whether we would ever see a new one. They said they'd love to but basically (paraphrasing here) that Activision (their major investor and owner of most of their hardware/software) didn't see the market. I was chatting with a game project lead and one of their top business/HR guys. Both got misty eyed at the thought of a new Heretic game. Raven is a company built upon a tradition of fantasy titles. What are they making now? These guys love games but they don't get to make what they want usually, especially today.

So you see, this is what's happening. It's not necessarily a bad thing because developers may have great ideas but may not have the business sense to see that they are signing a suicide plan with their game idea. But that control by the business majors isn't so great either, because in my experience that bunch is the least creative in the human species. It is a necessity now days though because the publishers have the know-how to get the capital. And capital is what is needed to make a modern game look perty and work right.

 

[swaaye]

Well, there goes my idea for Concept Art Grab-a-rama for PS3...

Uhhg....remember Rogue Leader and Rebel Strike. Yikes.

 

[fearsomepirate]

Hexen = GREATEST GAME OF TEH EVAR

You spent all your replay time in Rogue Squadron II & III unlocking secret upgrades, secret levels, and secret ships. Getting the Jedi Starfighter (which can take down a Star Destroyer in a few seconds) is way cooler than, say, unlocking a drawing of Samus.

 

[Megadrive1988]

I would be happy with graphics that are no more detailed than Xbox-GCN games, but with rock solid framerates, always 60fps, all the time, no matter what, and 8x anti-aliasing. Don't need ultra high resolution textures. I don't even care about pixel shaders. just some high polygon models and always solid framerates.

 

[Megadrive1988]

Is the buffer seperate from the VRAM? With such low speeds required it doesn't sound like it needs to be, why can't it just take as much VRAM as needed? Also, wasn't the buffer on the chip in the Dreamcast only like 512KB or some other value >1MB? (plus, I seem to remember dreamcast being benched upwards of 10million polys/sec, though maybe that was Naomi which had more ram)

Dreamcast's PowerVR2C graphics chip probably had less than 128K of on-chip memory. maybe 32K or 64K, I don't recall, but it was probably very small.

Dreamcast's SH-4 CPU could compute / calculate / transform at least 10M polygons, but as far as what the PowerVR2DC could render / draw and display on-screen, it was 6-7M peak, in theory.

Sega's conservative spec was 3M and Nintendo's conservative spec for GCN was 6-12M. Dreamcast could in theory do double Sega's conservative spec, and GCN could do roughly the same, double Nintendo's conservative spec.

 

[pc999]

Heavenly Sword, Lair, Motorstorm, Bladstorm, Fatal Inertia, Dark Sector, Eyedentify, and the new IP that Team Bondai is doing. And those games are for only the PS3 (well except Dark Sector). Then there's more for the Xbox 360. So where are you sounding so scared for?

They are the launch line up of PS3 (or at least they are the games that present the PS3) that alonemake them being no (big) risk (at least no more than an sequel) so they have big media covering, plus they are some of the most advanced games visually, made by big devs and they will not make radical shifts in gaming more or less they are made in already stablished genres.

If they are a small or new dev you dont even know they exist.

In all those games only Eyedentify is really innovative and made in new interfaces (and the power of PS3 I suposse too), beyond that there is no "real big" innovation.

 

[mckmas8808]

They are the launch line up of PS3 (or at least they are the games that present the PS3) that alonemake them being no (big) risk (at least no more than an sequel) so they have big media covering, plus they are some of the most advanced games visually, made by big devs and they will not make radical shifts in gaming more or less they are made in already stablished genres.

If they are a small or new dev you dont even know they exist.

In all those games only Eyedentify is really innovative and made in new interfaces (and the power of PS3 I suposse too), beyond that there is no "real big" innovation.

And the problem is?

 

[pc999]

How said there is a probem (at least by now), I said there is no risk and we cant live only by improvements, even if great games later in industry will suffer, is not in the first/second gen of games here gfx will make a big show and room enought for improvment but soon or later industry will suffer.