Working backwards: Overcoming technical challenges for the next generation.

Squilliam

Beyond3d isn't defined yet
Veteran
Supporter
Instead of working forward and attempting to predict what type of technology might be used in consoles for the next generation, I thought it would make an interesting thread if we figure out what the technical challenges the next generation of consoles will face and propose how best the console companies might overcome them.

What is the best way to increase bandwidth or reduce the bandwidth required on a 128bit bus cheaply?

Is 60gb/Sec bandwith with GDDR 5 on a 128bit bus adequate to render at 1080P?

Thats my starting question, feel free to add your own.
 
I'm not sure "adequately render at 1080p" will be a major design goal for the next generation, especially if it's a two-Xbox360s-ducked-taped-together, $250 at launch type of generation.

I would design around the following goals:

- $299 at launch
- backwards compatible, or at the very least reuse full toolset from current generation
- HDD-less SKU

The main technical challenges thus would be how to produce very cheaply something which is a natural extension of the current architecture.
 
Ok, but if you are looking at 299 at launch, how much are you willing to eat ($ loss per unit sold) to reach that price?

What kind of process shrinks are going to be available over the lifetime of the console? What target launch date? (Do you define it by when a process is mature or shoot for a target date that could give you a leg up on your competitors.) There has been plenty of discussion about the intervals between process shrinks become available increasing.

I admit to being one of those who do not think that 1080p is as important as what developers refer to as "eye candy." So 1080p may not be as much a consideration as what you can do at 720p.

(Is there a discussion on the MS announcement about the software DX emulation? - That could be pointing towards their ideas going forwards.)

I have a basic question. What is going to be done about load times? If the RAM doubles 3x for each generation (Xbox from 64 - 128,256, 512 for the 360), what happens to load times without a monster increase in physical media speed. (hdd, disc or Flash) There is going to have to be a real increase in this speed if we are talking about 4GB of RAM (even if we are talking about 2).

(Just out of hospital, drugged to gills and not in shape to do much simple math.)
 
Ok, but if you are looking at 299 at launch, how much are you willing to eat ($ loss per unit sold) to reach that price?

I admit to being one of those who do not think that 1080p is as important as what developers refer to as "eye candy." So 1080p may not be as much a consideration as what you can do at 720p.

(Is there a discussion on the MS announcement about the software DX emulation? - That could be pointing towards their ideas going forwards.)

I have a basic question. What is going to be done about load times? If the RAM doubles 3x for each generation (Xbox from 64 - 128,256, 512 for the 360), what happens to load times without a monster increase in physical media speed. (hdd, disc or Flash) There is going to have to be a real increase in this speed if we are talking about 4GB of RAM (even if we are talking about 2).

(Just out of hospital, drugged to gills and not in shape to do much simple math.)

As for the Microsoft Direct X renderer I believe that had something to do with allowing every PC to run the "Aero" effects and so avoid the Vista Premium/Basic ready debacle.

Its really a question of price really and what they are targetting. I guess questions such as using a small bus, say 64 bits with some blisteringly fast ram targeting 720p would be quite cheap if they used Ed-ram or an equivelent.

As for physical media, IIRC the PS3 on a 2* Blu ray drive gets about 3/4ers of the throughput of the Xbox 360s 12* DVD Rom drive. Therefore if you use either a 4,6,8 speed drive you should get 50% more, 125% more or 200% more performance from the drive at a consistant read speed with an accompanying drop in latency as well.
 
As for physical media, IIRC the PS3 on a 2* Blu ray drive gets about 3/4ers of the throughput of the Xbox 360s 12* DVD Rom drive. Therefore if you use either a 4,6,8 speed drive you should get 50% more, 125% more or 200% more performance from the drive at a consistant read speed with an accompanying drop in latency as well.
Considering the texture pop-in and terrible loading times (or so I hear) of many Xbox 360 games, that still might not be enough. Doesn't Halo 3 spend a minute or so loading levels? Assuming a bit less than 10 MB/s average transfer rate and having to read some 400 MB, that sounds about right.

Completely unacceptable, IMO. I really think next-gen consoles should all feature a HDD or fairly large solid state storage. Gameplay, convenience and overall functionality first, visual quality second.

(edit)
As for bandwith and resolution, you should take stereoscopy into consideration. S3D will most likely be a big thing in most gamers' minds by 2011/2012.
 
Last edited by a moderator:
Halo 3 spends a minute loading levels, but Call of Duty 4 maxes out at 15 seconds.

Maybe we should be looking at Jasper here as the model - perhaps the HDD cache could be integrated into the system as flash RAM. Surely by 2012, that would be viable cost-wise and it needn't be *that* high in terms of capacity.

I really don't want to see mandatory installs on any console, ever. To me, it's against the entire plug and play gaming ethos.
 
Considering the texture pop-in and terrible loading times (or so I hear) of many Xbox 360 games, that still might not be enough. Doesn't Halo 3 spend a minute or so loading levels? Assuming a bit less than 10 MB/s average transfer rate and having to read some 400 MB, that sounds about right.

Completely unacceptable, IMO. I really think next-gen consoles should all feature a HDD or fairly large solid state storage. Gameplay, convenience and overall functionality first, visual quality second.

(edit)
As for bandwith and resolution, you should take stereoscopy into consideration. S3D will most likely be a big thing in most gamers' minds by 2011/2012.

Well you could do two consoles that way, and guess which one would win?

In the hardcore market no doubt in my mind it's the second.

I mean people are not gushing over how fast Killzone 2 will load..

Also to the second post, 299 would allow quiet a lot of pop in next gen I think. By that time Blu Ray should be near as cheap as DVD so thats not a factor. Consider 360 debuted at 399/299 already, and it was fully "next gen" and certainly not two previous consoles duck taped together..
 
Also to the second post, 299 would allow quiet a lot of pop in next gen I think. By that time Blu Ray should be near as cheap as DVD so thats not a factor. Consider 360 debuted at 399/299 already, and it was fully "next gen" and certainly not two previous consoles duck taped together..

The next round of consoles will however face technological challenges that don't affect the current consoles as much. The manufacturers could crank up lot's of silicon into the current gen, because it was pretty much quaranteed that over their lifetime they shrink to 32nm, or maybe even abit further, that means that the average unit cost over the entire lifetime of let's say 360 is quite low, meaning that the initial loss is not that big of a deal, and even know it's not certain that X360 will turn profit, when it's all said and done.

The next round of consoles will likely start at 32nm, which is a nice "upgrade" or generational leap from the last round, but they most likely don't have as many process shrinks ahead of them, that means if you crank them up with silicon and make a big loss at the beginning, it would be even more difficult to break even or turn profit. The ability or inability to cut costs during the lifetime of a product is very important, when you decide what you can put there in the first place.
 
I think that the challenges that manufacturers will face are more on software front.
Intel, Nvidia, AMD, IBM all will provide better technology more powerfull more flexible.
Bandwidth constrains are the same for everyone, manufacturers already provide chips that made better use of available bandwidth.

What improvement will be noticed by the average consumer?
Imho what the average costumer will perceived as improved IQ is not AFxx AAxx ubber resolution but smoothness/softness and coherence. Some kind of AA a proper level of AF + smoothness add by up scaling will do the trick. I don't feel like the average costumer is about ubber sharp/clinical rendering.
Overall improved IQ will be noticed, but it's not a critical factor especially the next generation systems are likely to do well enough in this regard. No headache here.
I think better lighning and shadows are more noticeable to anyone but once again next generation (even if somewhat conservative) should deliver good enough improvement.
I would say the same for particles and thing as tissues for example.

What I think average costumer is after is "consistence". On the graphic side having things smooth/soft enough (as opposed to sharp as in the PC space) shouldn't be difficult. to make the overall experience more consistent we need a huge improvement in the way your character or AI interact with the world.
Bungies spoke of this lately and they are spot on proper animation is the next big thing.

All this points need a their share of computing power and memory but I still think that the huge effort is on the software side.
I mean even granted ubber resources on the hardware side, the cost of content creation can't afford a big jump.
Artists, animators just can't deliver to extend even granted enough processing power, media space and RAM.
Think about ID Megatexture, or their idea about infinite geometry it's great technically (as BRD or its successor will grant enough space) but without proper content creation chain tool the cost for content creation will explode.
Procedural content creation tools has to improve. In this regard regard the interview of Dominic Guay (Far Cry 2 technical director) is really insightful (see the state of procedural environment creation thread).

The same thing is true for AI and the way the story adapt to your action, last gabe newell interview in regard to left4dead and procedural narrative is super interesting too (I posted the interview in the game section // lef4dead topic).

For animation things like Euphoria are a good start and these kind of effort need to be pushed forward.

Devs need proper tools in regard to these points, how to deal with unpredictability, what to do offline/real time, etc.

For me the next huge effort is not graphics but how makes what happens on screen consistant with a pretty acceptable graphics.
 
Last edited by a moderator:
As for physical media, IIRC the PS3 on a 2* Blu ray drive gets about 3/4ers of the throughput of the Xbox 360s 12* DVD Rom drive. Therefore if you use either a 4,6,8 speed drive you should get 50% more, 125% more or 200% more performance from the drive at a consistant read speed with an accompanying drop in latency as well.

The drop in latency when raising the rotation speed isn't linear, if there ever is one - the optical pickup is still the same weight, and it still has to be repositioned around.

Also, when you raise the rotation speed, you raise the noise - the Xbox 360 isn't noisy because it uses a DVD, but rather because it uses a 12x[/v] DVD.
 
Maybe we should be looking at Jasper here as the model - perhaps the HDD cache could be integrated into the system as flash RAM. Surely by 2012, that would be viable cost-wise and it needn't be *that* high in terms of capacity.
Yes, that's what I was getting at.

I really don't want to see mandatory installs on any console, ever. To me, it's against the entire plug and play gaming ethos.
I will readily accept a 25 minute install (from a single Blu-ray); I can do something else in the meantime. But if I see 50 second load times, I'll ditch the game and the developer won't see any more money from me.


Also, when you raise the rotation speed, you raise the noise - the Xbox 360 isn't noisy because it uses a DVD, but rather because it uses a 12x DVD.
Yes, that's also something to consider. Noise is the main reason I haven't bought a 360.
 
IMHO the real performance benefit from SSD technology (say for the OS and data caching) would come from super high end SSD. I think of the like as the last Intel SSD (super expansive) with super fast seek time and impressive bandwidth but that could be very sill too expansive by the time the next generation systems are launched.

But I don't denie that more "standard" SSd would provide quiet some benefit in power/noise/form factor :)
 
Artists, animators just can't deliver to extend even granted enough processing power, media space and RAM.
Think about ID Megatexture, or their idea about infinite geometry it's great technically (as BRD or its successor will grant enough space) but without proper content creation chain tool the cost for content creation will explode.
Actually, Megatexture should, if anything, make content creation less costly, because artists can add new detail at any time without worrying about memory budget. The ability to put unique texture everywhere doesn't mean you have to.
 
I will readily accept a 25 minute install (from a single Blu-ray); I can do something else in the meantime. But if I see 50 second load times, I'll ditch the game and the developer won't see any more money from me.

A 50 second load time if you have 10-15 minutes of gameplay isn't really a big deal. It's all going to be relative. Basically not being able to play a game for 25 minutes because you deleted the installation to make way for another game is certainly not user-friendly and would most likely especially annoy the casual gaming audience both Sony and MS need to address in the next gen.

Bottom line is that optional installation on all games as per NXE is the way to go - you've got the console plug 'n' play ethos, but if you want faster loading times or a quieter console, and you're willing to put up with the annoyance of memory management, you can have it.
 
A 50 second load time if you have 10-15 minutes of gameplay isn't really a big deal. It's all going to be relative. Basically not being able to play a game for 25 minutes because you deleted the installation to make way for another game is certainly not user-friendly and would most likely especially annoy the casual gaming audience both Sony and MS need to address in the next gen.
This depends on how often you switch games and need to make installs. How much do people jump betwen games on average?

Bottom line is that optional installation on all games as per NXE is the way to go
That goes without saying ;) Especially with PS3 style open HDD replacements. Let me chose me experience for my budget.
 
A load time of 50 seconds on an FPS doesn't bother me as much. They tend to be rather intense games and a break of a minute between levels is often a chance to refill drinks and take a breath around here. The load time on RPG's though, that is often something that brings curses out when one sees the quest is going to require 3 or 4 load screens and there is little involved in the quest itself.

As to the use of BR next gen (I don't think there is much argument for anything else at this point if we are assuming the existence of an optical drive) - "According to the Blu-ray Disc specification, 1x speed is defined as 36Mbps." A 10x drive would be 360Mbps. I'm writing this out so someone can check my oxy filled brains math. So 45 MB/sec. At 2 GB of RAM that is 44.4 seconds to fill it. For 4 GB double that to 89 seconds. A minute and a half. I don't know that the RAM always needs to be completely filled for a game to begin playing its level/area (meaning I don't know what currently is done but with the prevalence of streaming it seems they use a large portion of the available RAM for what is immediately in view and the rest is streamed), but 90 seconds seems like a nightmare for for almost anything to me but FPS's.

If content generation is going to be a major topic for this ( i.e - prodedural generation) then I dearly hope it gets a lot better than what I have seen so far (limited to what is in actual use and not an artists page.)
 
As to the use of BR next gen (I don't think there is much argument for anything else at this point if we are assuming the existence of an optical drive) - "According to the Blu-ray Disc specification, 1x speed is defined as 36Mbps." A 10x drive would be 360Mbps. I'm writing this out so someone can check my oxy filled brains math. So 45 MB/sec. At 2 GB of RAM that is 44.4 seconds to fill it. For 4 GB double that to 89 seconds. A minute and a half. I don't know that the RAM always needs to be completely filled for a game to begin playing its level/area (meaning I don't know what currently is done but with the prevalence of streaming it seems they use a large portion of the available RAM for what is immediately in view and the rest is streamed), but 90 seconds seems like a nightmare for for almost anything to me but FPS's.

There's no guarantee that Blu-Ray drives will see the same 'X' factor as DVD drives did. Just as DVD drives have never seen the 48X and 52X factors that CD drives achieved. The baseline rotational speed is already so much higher to begin with.

I think the real trick for the next generation will be to take advantage of all the work done this generation. Everyone will have to either do installs or re-use the kind of sophisticated streaming systems that games are incorporating this generation.

At some point, all of the proposed next-gen memory systems (spintronic, magnetic RAM, etc.) will come to fruition, but it's not at all clear that they will be ready for the next generation of consoles. Cheap fast flash should be, though.
 
From what I understand, a magnetic hdd or flash ram are not enough faster to make a serious dent in this. Seek times may be substantially better, but looking at the time differences (minus Halo 3) for 360 installs, that is not enough to turn 90 seconds into something tolerable.
 
What I'd like to see is better support for compression at the hardware level, or flexibility that allows it.

Having desktop style image compression for textures would be very interesting.
I could see microsoft retrofitting the HD Photo (JPEG XR) format for texture use. Given it already supports mipmaps, HDR and as many channels as you want. It also apparently decodes with just adds/muls and can efficiently decode subrects.

Do something similar with geometry compression and you are on your way to [strike]solving[/strike] easing a lot of problems.
 
What I'd like to see is better support for compression at the hardware level, or flexibility that allows it.

How does hardware support for such features get into the spec? Do the voices of developers count for much for low-level aspects of design? (And to everyone else thinking of the 256MB/512MB RAM issue with Epic, that's not what I mean by low-level).
 
Back
Top