How fast do current generation consoles go?

Quote:

Original post by PeteCow Say if we take 100mph as the fastest speed a car actually needs to go, given other no performance related considerations (price, safety, economy, ecology etc ). And we assume that there is an analogous maximum performance level for consoles. What is the current console generation’s top speed on this scale? I'm not talking physical limitations of console performance here. I mean the limitations caused by there being only so many pixels on a home entertainment system, and no need to renew them past a certain refresh rate. Also that developers can only utilise so much processor power without huge dev times, unaffordable platforms and ridiculous overengineering ( eg. giving 50,000 spectators in a racing game an AI each ) I asked this question of some console development colleagues, and got facetious answers; entertaining, but un-enlightening. Perhaps someone here might give me an intelligent opinion...

So lets remove the "car velocity" analogy because it just leads to me wanting to drop an XBox from a plane to measure it's terminal velocity. Instead, what you really seem to be asking is "In percentage, how fast are current consoles compared to the maximum that would still be practical to develop for".

0.000000001%, give or take a percent.

In other words, there are too many variables to answer that question.

First think about the visual resolution: there are only so many pixels on current home entertainment systems, but resolutions continue to rise as technology becomes cheaper. We're at 1080p right now, but that doesn't even come close to the human visual resolution. Television resolutions will continue to rise, driven by games and interactive media.

Also, just because you're using more power doesn't mean it's immediately a higher budget game. Your example of spectators in a racing game was flawed, but what about a 50,000-man army on a battlefield, each run by their own AI? That could be practical in a game settting.

Finally, as tools get more advanced and engines/libraries become more complex, code and assets can be developed more quickly and become much more powerful. While it might cost 20-30 mil to make a Gran Turismo game right now, what about 10 years from now when there's art tools designed specifically for creating vehicles quickly, there's a licensable asset library for outdoor environments, and a fully-realistic vehicle physics model is part of the PlayStation 4 SDK? It would cost a fraction of what it does now, and therefore games could be much more complex at the same budget.

In the end, this question is basically "how close are we to creating The Matrix". The answer: not very.

Don't confuse speed with complexity. Some problems are so slow to process that current hardware is not even remotely fast enough; these problems can often be expressed in a few dozen lines of code and/or are simpler to program than the algorithms we currently use.

As a point of reference there are plenty of algorithms that cannot run on current computer hardware because it is too slow:

factoring of large prime numbers
a raytracing graphics engine (still doesn't run fast enough to be used in AAA games)
various AI algorithms have to be timesliced because the hardware is too slow.

-me

Quote:

Original post by Palidine a raytracing graphics engine (still doesn't run fast enough to be used in AAA games)

An excellent example. We're a long way from having real-time lighting and shading on par with a Pixar movie.

Let's first hit your analogy. How fast does a car "actually need to go"? A century ago the first US speed limit was implemented: 8 mph (13 km/h) in Connecticut. At the time, that was deemed the fastest that people actually needed to go.

The speed limits for today haven't changed much in 50 years, mostly because people generally can't react to road conditions at faster speeds. Sure and alert person could possibly handle a bit more speed, and a carefully trained alert driver could handle triple the current speed on a closed course. But since we must share the road with sleepy drivers, drivers who are eating or talking on phones, or even drunk, the higher speeds are considered unsafe.

Now, I would *want* a vehicle to travel several hundred miles per hour or even more. I would *want* a vehicle that I don't have to drive, it would drive itself. When I must travel home, I *want* to be there instantly -- I don't want any travel time. Realistically (given the lack of price, economy, and ecology constraints) I'd be happy with smart vehicles that travel at supersonic speeds and get me to and from work without any action on my part.

So on the 0-100 scale, I'd even rate todays vehicles at a 0, since the fastest transportation speed is infinitely less than my desired transportation time, or put them at 5 of 100 using the more realistic supersonic travel notions.





I'll mirror this with computer performance. I don't want to have budgets for processing time or memory space. I don't want to even think about swapping out textures or limiting my model space.

What you said is ridiculous I believe would be wonderful.

I have been working on an group AI system for a few months. In order to stay below the (very tiny) computation budget I've been allotted, each AI node is only updated a few times per second, and only a small number are updated every frame. Distant nodes are all treated with the same element, so at a distance they are all doing exactly the same thing. Now imagine what would happen if you could lift the budget requirements. We could just have artists blanket the stadium with seat templates. A simple image of the track then gives the population of the model and use that to establish if a person is sitting there. Next, a simple tool actual DOES implement 50,000 spectators with their own AI. You would only need to write a small set of tunable values and suddenly the stands come alive.

We don't *need* to have it, but that would provide a level of convenience that would cut the development time by quite a bit.


Another one: saved data. I have worked a few times with implementing replay systems to both replay earlier games and to help train to improve your score. In each case, we struggle with recording the best data we can over all the levels, but this is in direct opposition to the storage requirements on a console. Our "ideal" save game structure for the PS2 title would have been around 80MB of data if we could save everything we wanted. We joked about shipping the game with a collection of memory cards that users would need to swap when playing different courses.

We don't *need* boundless storage space, but having it would be very convenient, would provide a better player experience, and would mean we don't have to spend the time trying to reduce the size of structures a few bytes at a time to fit on the memory cards or budgeted number of blocks.



If I could swap out C++ for a language that is actually designed around multiprocessing (I don't believe there is such a thing) that gracefully handles problems like automatic and intelligent mutex locks when working on objects and other issues, move to a language along the lines of Java and C# in terms of memory strategies and cleanup, and has better error handling than any existing procedural or functional language that I have ever encountered. Such a language would obviously have a high processing overhead when run, so the CPU would need to compensate for that.


If I could get a console that runs "as fast as it actually needs to go", it would be countless orders of magnitude faster and bigger than today's fastest machines.

So just as I *want* to travel instantly, I similarly *want* to have systems that are effortless to develop, and consequently give current consoles a 0 of 100.

Quote:

Original post by JBourrie Quote: Original post by Palidine a raytracing graphics engine (still doesn't run fast enough to be used in AAA games) An excellent example. We're a long way from having real-time lighting and shading on par with a Pixar movie.

Actually.....

At a conference three years back, we discussed this issue. I can't remember who was giving the talk for the session, but he was one of the high level Pixar execs. In 2001 with the GeForce 3 we passed the level to render in real time their earliest productions up through the 90's. With the dual GeForce 7000 series and a dual core processor (and boundless storage) they could theoretically render Toy Story in real time, and much of A Bug's Life.

The trends he presented were that we were catching up at about 1.5 years of offline rendering to every 1 year of realtime rendering. There was a spike with programmable video cards, but both curves followed the same growth patterns merging together in the not-so-distant future.

I imagine that given the trends he presented, a very high-end consumer machine of today could -- with difficulty -- render in real time what they were producing five years ago. I can easily imagine the GeForce 9 series easily handling nearly all the scenes in Finding Nemo, although the 8800's could also handle most of it.

Well, which has more graphics power, Wii or PS2?

Wii by a long shot. Even the Gamecube had more. The PS2 is a very weak machine.

And yet there are some visually gorgeous games for PS2.

Anybody remember Black?



(Actually, I detest drab gray color schemes, but I liked the visual fx on this)

So it's too bad there haven't been any titles of that visual quality announced for Wii, because if they could show those off, it could help to address doubts about Wii's graphics power.

Although, I have to admit, seeing the same FPS gun poking from the bottom of the screen in ever higher detail has become boring to me. I'd rather see the Wii focusing more on fully exploiting the richness of its control interface.
ie. I want to lash a whip like Indiana Jones or Pitfall Harry; I want to toss a grenade using the motion of my hand; I want to handle a sword and shield using the Wiimote and nunchuk; I want to punch my way through a pack of thugs Wii-boxing style; and, of course, I want lots lightgun shooting action.

Well, I'm not a technician, but I think the progression is not a line but a curve and in the next years the jumps in technology will be more near one to the next than what we think.

Hi,

Can anyone please suggest the tech tools and languages i should learn to go into console games development(playstation, xbox, wii , etc.) Right now im in to pc/online and mobile multimedia games development.


Thanks in advance ,

-Basker D.

Quote:

Original post by d basker Can anyone please suggest the tech tools and languages i should learn to go into console games development(playstation, xbox, wii , etc.) Right now im in to pc/online and mobile multimedia games development.

It's a bit off-topic, but I'll answer anyway.

So-called devkits (development kits) are very expensive and generally are only provided to established game companies and publishers. I've also heard of some schools/universities with some devkits.

However there's one alternative: Microsoft XNA. XNA is a game platform for both Windows and Xbox 360. For the latter you'll need a $ 100 yearly subscription, which allows you to copy your game to your Xbox 360 and play it. You can also give it to other XNA subscribers, but you can't run those games without XNA.

I think XNA is the way to go.