NGGP: NextGen Garbage Pile (aka: No one reads the topics or stays on topic) *spawn*

[anexanhume]

Apple gets 30%+ margin on their products. Want the same from console owners? They'll laugh in your face.

All of the profit in this industry is driven by userbase size. You guys are devising ways to chop that up with no explanation of how that will drive people away from your competition. You're spending more money more often to put out products. If you even think of making a profit on your box, your competition suddenly looks very attractive. This strategy would take a long time to work and appears more scorched earth than it does smart product line management to me.

 

[bkilian]

Look at that, I remember predicting months ago that neither Sony or MS would be hyping their hardware this time: Looks like Sony is on board.

 

[dobwal]

Tessellation is not a magic bullet. It has some very serious issues that have to be settled...

Modeling for tess+displacement requires a very different workflow and a very different mesh, some details have to be left out, but you'll also need to add more polygons elsewhere. Rigging will have to change, map baking will have to change, all in all I would compare the change to the introduction of normal mapping. We all know how much a step that was (something Blizzard still hasn't made in WoW or in Diablo!) and how expensive it's been to game studios.


The other problem is that it's not as scalable as you seem to believe. Not having enough geometry to manipulate means you loose geometry detail - silhouettes change, complete components disappear. The non-tessellated mesh has to be completely different, with different UVs, normal maps, to accommodate this lack of vertices to work with.

Also, a lot of stuff is not possible to model with displacement maps. Multiple independently moving pieces of armor, weapons, vehicles - how do you displace them?

It would also only work well if you introduce it at a near-micropolygon level, which would immediately kill the GPU's efficiency with the small triangles.

Tessellation without displacement also wouldn't get you too far, except for smooth objects we call hard surfaces, like cars, pipes, guns etc. And these would really require a completely different modeling paradigm to preserve sharp edges and bevels and such, increasing the polygon count astronomically.


So, in short, the way you believe it could be used just won't work, it can't provide an easy click-and-go LOD system to take advantage of faster hardware without seriously compromising the base version.

Of all the new trailers you seen of impressive next generation graphic titles based on new engines what percentage are not slated for console release? How many new gaming engines are being built without consoles in mind?

And thats a reality based in the fact that most PC games are ports of console titles. The level advancement in hardware PC is moving at faster rate than the software side of PC development. AMD and Nvidia have the ability to expand the gpu adoption outside of the gaming market through gpgpu which is why we have seen gpgpu pushed so hard. On the gaming software side the push hasn't been as significant. Consoles are encouraging that reality.

However, with a new generation of consoles MS and Sony has a chance to accelerate resolution of some of the issues that current PC software development face by injecting billions of dollars of software research and development into things like tessellation.

 

[AlphaWolf]

because the same assets are forward compatible to next durango. (and the game it's already done for the previous durango, adding textures and effects it's not a big effort)

PC devs do this all the time, for all the games, it's so incredible for you?

and devs creates games even with a 0 (zero) userbase, at launch, add that resolution, physics, tessellation, antialiasing etc are FREE with better hardware

As Laa-Yosh has explained, it is a big effort. It's not free to add things. Resolution and framerate, that's it.

 

[Xenio]

As Laa-Yosh has explained, it is a big effort. It's not free to add things. Resolution and framerate, that's it.

so PC devs did not make different resolution textures and switch ON/OFF effects depending on hardware?
in my world this happens all the times

and for PC they have to support hundreds hardware combinations, with durango, two or three.

 

[TheChefO]

As Laa-Yosh has explained, it is a big effort. It's not free to add things. Resolution and framerate, that's it.

While that may be "it", that doesn't mean it is trivial.

Many gamers swear by 60fps and anything else is a letdown. Same goes for resolution. 1080p or bust as they say.

Well, if Durango only has to focus at 720p while Orbis is pushing 1080 (difference of the hardware ROPs + bandwidth + CUs) then that leaves the door open to match Orbis' 1080p with a Durango+.

One for 720p, the other for 1080p.

For those that "care" about resolution, they will pony up the extra cash for the compute and bandwidth necessary to attain 1080p.

 

[Xenio]

One for 720p, the other for 1080p.
.

with the 16% difference in CU, it's hard to double the graphic performance
even that, someone can explain me why PC devs are able to do assets and effects for a big big number of hardware combinations and console devs are not able to do this or better, with three fixed SKU?

 

[XpiderMX]

1080P (Orbis) vs 720P (Durango)? Are you sure?

Can 12CUs vs 14CUs make this kind of differences?

 

[Laa-Yosh]

?..

Huhh? What has this got to do with my points about using tessellation?.

 

[xJumPeRJumPzZ]

1080P (Orbis) vs 720P (Durango)? Are you sure?

Can 12CUs vs 14CUs make this kind of differences?

Probably not. Console Wars are the worst every generation.

 

[joker454]

As Laa-Yosh has explained, it is a big effort. It's not free to add things. Resolution and framerate, that's it.

They may want to add variable assets/settings if that same Durango app also runs on PC's, laptops, tablets, etc. Given the range of device specs between PC's, laptops, tablets and consoles, and given how easy compatibility will be between them all (in Windows world) then it will almost become a necessity to have various sets of assets tuned to each platform anyways. Typically that may be seen as a big effort, but remember they will also be saved mountains of time by not having to throw away millions of lines of code away every few years since all those devices will be largely compatible as they will all be coding to an api. That is an absolute mountain of development time saved, which can be now spent on having various sets of assets and trying out new game ideas.

 

[Xenio]

They may want to add variable assets/settings if that same Durango app also runs on PC's, laptops, tablets, etc. Given the range of device specs between PC's, laptops, tablets and consoles, and given how easy compatibility will be between them all (in Windows world) then it will almost become a necessity to have various sets of assets tuned to each platform anyways. Typically that may be seen as a big effort, but remember they will also be saved mountains of time by not having to throw away millions of lines of code away every few years since all those devices will be largely compatible as they will all be coding to an api. That is an absolute mountain of development time saved, which can be now spent on having various sets of assets and trying out new game ideas.

exactly what I think

 

[Bagel seed]

Year 0: 1.2TF ------------------- $350
Year 2: 1.2TF, 2.4TF, 4TF ---- $250, $350, $500 ---- 4K and HQ 3D games!
Year 4: 2.4TF, 4TF, 8TF ------- $200, $350, $600 ---- Next gen begins early for the hardcore gamers (and cloud beta access)
Year 6: Let's go to the cloud! -- $300 - 600 / yr ------- Games, cable, Internet, phone, etc

I think I'll get Orbis at launch and wait for Year 2 to get Durango Elite sku if Orbis doesn't upgrade.

 

[AlphaWolf]

Year 0: 1.2TF ------------------- $350
Year 2: 1.2TF, 2.4TF, 4TF ---- $250, $350, $500 ---- 4K and HQ 3D games!
Year 4: 2.4TF, 4TF, 8TF ------- $200, $350, $600 ---- Next gen begins early for the hardcore gamers (and cloud beta access)
Year 6: Let's go to the cloud! -- $300 - 600 / yr ------- Games, cable, Internet, phone, etc

I think I'll get Orbis at launch and wait for Year 2 to get Durango Elite sku if Orbis doesn't upgrade.

Yes in 2 years they'll launch a console more than 3x as powerful. Guess everyone should just wait for that.

 

[Shifty Geezer]

So as a Dev, give them a quicky/crap port and be done with it. Sounds like a plan........

How many PS360 gamers this gen have been grumbling about the lack of IQ and framerate in their game? How many would quite happily pay $200-$300 for an updated PS3+ or 360+ that plays exactly the same games (Uncharted, Borderlands, Halo) in higher framerate and quality (3D at 60 fps, 1080p60 (or even stable 30 fps would be nice in many cases!), high AA and filtering?

There are pros and cons in all choices. Concerns about lack of hardware utilisation due to fat API's meet aggravation with hardware that can't play the best games in the desired quality. 2+ year game developments don't want a dead platform, but they don't want a start all over again every gen either. People want games specifically designed for their box, but they also want content that plays on other devices.

There is no right choice. There's only a number of choices will which upset some people and excite others. The trick is to take the business strategy and execute it most effectively such that you maximise profits, and don't care one jot about those poor gamers left with a platform/serviec they dislike. Such is the world of business. If it were making that choice, I'd be seriously considering the fat API, progressive platform. I reckon long term benefits outweigh short-term costs and competitive disadvantages, although it'd need superb market communication to pull it off.

 

[Shifty Geezer]

Upgradeable console talk here.

 

[dobwal]

Huhh? What has this got to do with my points about using tessellation?.

It has everything to do with it. Do you honestly think that the way tessellation is done and the problem it faces right now is the same way tessellation will done and will face those same problems tomorrow. Tessellation techniques will improve over time. The level and the rate of that improvement will be accelerated if the next gen consoles adopt it for common use across its software enviroment. PC game development doesn't revolve around apu hardware nor can we conclude that the GCN cores and their capabilities have been thoroughly explored. We don't know how the SRAM will be used in Durango.

Several tessellation patents have been granted to MS over the last two years including numerous patents regarding memory management, rendering and a host of other gaming related novel inventions. MS hardware doesn't look to be a standard PC configuration. MS nor Sony nor Nintendo is tied to the traditional hardware and software PC enviroment and the problems faced within those enviroments. Tessellation and other techniques' problem and issues might be resolved with consoles. Consoles have a ton of money backing them and you simply can't throw more silicon at them as a brute force method of improving performance.

Console require a more thorough and fine grain look at their hardware because its fixed and for 5-6 years at a time. No one is spending 15 million trying to make their game look as best as possible on a Radeon X1800 right now. But millions are still being poured into a 7 year old Xenos trying to coax more power out it. Console are now on the forefront when it comes to tackling problems with practical solutions on the software side while PC still maintain its leadership on the hardware side.

I not saying the tessellation will be the magic bullet as consoles might be terrible at tessellation. All I know is that MS has patented several techniques that makes me believe they are looking at tessellation as a way to make multi tiered skus based on hardware performance a more practical approach then the way it exists on PCs.

 

[tunafish]

1080P (Orbis) vs 720P (Durango)? Are you sure?

Can 12CUs vs 14CUs make this kind of differences?

No, but 16 vs 32 rops, and 12 vs 18 CUs can.

There is no %&"#& way making the 4 CUs special will make them less good than just not mucking with them. If it would, they wouldn't do it. For multiplat titles, the only thing that is easy to scale is the graphics -- if Durango isn't running physics or whatever on CUs, Orbis won't be doing that either. So when comparing multiplats 12 vs 18 is the correct number.

Also, for more pixels on the screen, rops are maybe even more important than CUs. Twice the pixels always means twice the load on the rops, but the part of CU workload that is used on geometry doesn't get any harder when res gets higher.

 

[Xenio]

No, but 16 vs 32 rops, and 12 vs 18 CUs can.

it's already stated sooooo many times, that 4 CU in orbis are not for graphic, because they can improve only slightly, this is why they are destinated to other tasks

so are 12 vs 14 CU,

the
same
thing

1080P is 2.073.600 pixels
720P is 921.600 pixels

14 CU over 12 CU = +16%
1080P over 720P = + 125%

so can finally end there the false statement "orbis is 1080p where durango is 720p" or we have to explain why a total workload of 225% can't be done with only more 16% CU units PLEASE?

and don't talk about bandwidth, because unltil contrary trial, they have both 170 GB/s, one from gddr and the other with combined/tiled ram-esram

Twice the pixels always means twice the load on the rops, but the part of CU workload that is used on geometry doesn't get any harder when res gets higher.

NO, their ROPs don't work in the same way, if I remember well orbis can't resolve in eSRAM and can't do hardware tiling on this

and durango runs physics on CPU.

 

[bkilian]

No, but 16 vs 32 rops, and 12 vs 18 CUs can.

There is no %&"#& way making the 4 CUs special will make them less good than just not mucking with them. If it would, they wouldn't do it. For multiplat titles, the only thing that is easy to scale is the graphics -- if Durango isn't running physics or whatever on CUs, Orbis won't be doing that either. So when comparing multiplats 12 vs 18 is the correct number.

Also, for more pixels on the screen, rops are maybe even more important than CUs. Twice the pixels always means twice the load on the rops, but the part of CU workload that is used on geometry doesn't get any harder when res gets higher.

Well, as I speculated on the orbis thread, what if the 4 CUs are actually _reserved_ as opposed to just "different". It's might be about the right amount of processing power to turn stereo camera images into a 3d mapped scene in real time, and probably a lot cheaper in BOM than a 3D camera like the kinect, especially over time. So 12 versus 14 might actually be a reasonable comparison.