PC visuals in games will change by leaps and bounds when graphics cards die off.

I have been reading up on things like programmable shaders, and basically even directx 10 with it's shader model 4 features does not even come close to software rendering capabilities. Remember playing half-life or quake for the first time, and not even needing a graphics card?

The possibilites in theory were unlimited, yet in reality cpu's were just too slow at the time.

But welcome to 2007, the mass produced and affordable intel core2duo quad processor (just one example) could do so much with software rendering. Dedicate a whole 2 cores to graphics computations with smart and highly efficient voxel based rendering techniques and you can achieve crysis like scenes on a cpu at a playable framerate. (OK, I'm exaggerating ;) )

Unlike programmable gpu's, software rendering is unlimited in it's possibilities, pixar and other studios use software rendering to produce their highly detailed 3d films, graphics card don't have the flexibility for techniques such as ray-tracing etc in real time.

So, here's what I'm getting at....about 20 years from now, gpu's are basically going to become so flexible, so general purpose with their shader implementations, that they will in fact have all the capabilities of cpu's. Then the logical step is to combine any differences and create a hybrid cpu/gpu design that will become so powerful, so efficient, that specialised graphics processing boards will become obsolete.

Why bother optimizing your game for a particular gpu instruction set like directx10 or opengl2.0, when you can just write all your shaders in software and achieve any effect you wish on any scale, which would be possible, given the power of the cpu you're dealing with.

To further simplify my point....we will only get true to life graphics and leave the uncanny valley, when cpu's become our graphics and physics processors as well. From the beginning it has been about integration. We are headed towards some sort of processing singularity, when the core features of a whole PC will fit on a single nail sized chip, I guarantee it, and it's sooner than you think.

And with it will come games that will be so realistic, they will be games no more. Convincing AI will have the chance to flourish when such supercomputing power is in the homes of mainstream consumers. We are headed for a fusion of hardware. When that fusion occurs, developers will only need to worry about one thing, software. They will be limited no more.

What does this mean for the future of specialised games consoles?

Nothing sinister. They will simply become simplified entertainment extensions/appendages of the greater processing hub of the home of 2027 that will be the PC.

Couple all this with new peripherals, that will allow user interaction on an entirely new level (direct brain-to-pc interfaces) and we are living in ancient history people.

Remember the games of 1987? We live in an entirely new era. Imagine what 2027 will look like...

I think we should look ahead, instead of focusing on directx11 or the geforce 9800, the gpu makers and cpu makers should experiment, work together and dvelve into entirely new and radical technologies. Let's skip this time consuming sequential evolution of computer graphics, and move by leaps and bounds. To do this the cpu and gpu manufacturers should pool their resources and technology.

Gpu's are starting to share so many aspect of the cpu, it is becoming wasteful, us gamers are ready for union NOW. No separate physics cards, no more graphics cards, just one single processing unit smack bang in the heart of our machines. No need for specialisation, for fancy visual tricks, if the cpu is fast enough.

If nano/quantum computing ever takes off, and mainstream pc's hit the terahertz range, specialised hardware will become obsolete, and the possibilities endless. I look forward to this day.

~Gangans~

nice post man.

From reading that, I think my school failed me. But I'm understanding what your saying, graphics made from a GPU and a processor would be far more forgiving on the system than if you just had the processor doing it all.Platearmor_6

Yeah you got the gist of it, basically what I'm saying is, integrate cpu's, gpus and just about anything else on a single multicore cpu, and you greatly simplify the process for games developers. You have one set of tools needed to tap into all this power, not a whole bunch of different compilers and tools for gpu specific things like direct3d and opengl etc and then having to worry about physics calculations on the cpu or even a separate psu (physics processing unit) and having to use yet another set of tools to define those calculations etc..

I share your enthusiasm and we can only hope that technology will advance enough in our lifetime to the point that gene-therapy, nano-technology are words in common use and there is even slang out there for them. anyway i saw a video today of a guy controlling objects in what looked to be a hl2 environment with his mind. so i guess the days we're almost at THAT point, where we can control machines with only our thought. villians will have a field day with this.but back to the point, i agree witht he poster, a single card/processor should meet all our computational needs, but the trend seems to be, if you take a&eia into account, more and more cards that cater to very specific areas of gaming...whether this is good or bad and whether we need to be cutting down on peripherals instead of incrementing their numbers, who knows? let's hope they find that simplicity is the key to progress in gaming .if not i can't wait until 2027 when i need to buy a ageia emotion card or whatever just to run a game...wait, no; that'll suck.:PtrenchermanX

Functional simplicity is always desirable.

And lol at ageia emotion cards :lol:

nice post man.

druglord6

But welcome to 2007, the mass produced and affordable intel core2duo quad processor (just one example) could do so much with software rendering. Dedicate a whole 2 cores to graphics computations with smart and highly efficient voxel based rendering techniques and you can achieve crysis like scenes on a cpu at a playable framerate.

Gangans

While as a whole I agree with your post, and in 20 or so years, I'd love to have an integrated CPU & GPU, that statement that I've quoted above isn't true.

The reason that graphics cards are so fast at what they do, is becuase they are extremely specialised at what they do.

While I can't be 100% sure on what I'm about to say, and I'm completely fine if anyone decides to completely own me but, 2 cores of a quad-core COULD NOTproduce crysis like graphics at a PLAYABLE frame rate. It just isn't possible.

Maybe in 20 years, but not right now.

[QUOTE="Gangans"]

But welcome to 2007, the mass produced and affordable intel core2duo quad processor (just one example) could do so much with software rendering. Dedicate a whole 2 cores to graphics computations with smart and highly efficient voxel based rendering techniques and you can achieve crysis like scenes on a cpu at a playable framerate.

G013M

While as a whole I agree with your post, and in 20 or so years, I'd love to have an integrated CPU & GPU, that statement that I've quoted above isn't true.

The reason that graphics cards are so fast at what they do, is becuase they are extremely specialised at what they do.

While I can't be 100% sure on what I'm about to say, and I'm completely fine if anyone decides to completely own me but, 2 cores of a quad-core COULD NOTproduce crysis like graphics at a PLAYABLE frame rate. It just isn't possible.

Maybe in 20 years, but not right now.

I'm talking highly optimized, efficient code running in the voxel(not polygon) rendering regime. THIS game: http://en.wikipedia.org/wiki/Outcast_%28game%29 had BUMPMAPPING, FSAA and DEPTH OF FIELD EFFECTS in 1999before gpu's even used pixel shaders.....because it didn't use a gpu! Sure it doesn't look like much now but it outdid the gpu's of it's time with software rendering.

[QUOTE="G013M"][QUOTE="Gangans"]

But welcome to 2007, the mass produced and affordable intel core2duo quad processor (just one example) could do so much with software rendering. Dedicate a whole 2 cores to graphics computations with smart and highly efficient voxel based rendering techniques and you can achieve crysis like scenes on a cpu at a playable framerate.

Greyhound222

While as a whole I agree with your post, and in 20 or so years, I'd love to have an integrated CPU & GPU, that statement that I've quoted above isn't true.

The reason that graphics cards are so fast at what they do, is becuase they are extremely specialised at what they do.

While I can't be 100% sure on what I'm about to say, and I'm completely fine if anyone decides to completely own me but, 2 cores of a quad-core COULD NOTproduce crysis like graphics at a PLAYABLE frame rate. It just isn't possible.

Maybe in 20 years, but not right now.

Once again I refer you to the game called outcast.

Also, I think most of you underestimate the core2duo line of processors.

I have a low end core2duo 5300 in my laptop, and a p4 3.4 ghz in my desktop. I did some benchmarks, lo and behold, my little laptop is essentialy 4x faster in real world applications! It's as if my p4 was running at 13.6ghz, the slow bastard.

It's at this time that I'd like to highlight a relevant point about Crysis in particular, it will use voxel based software rendering for it's highly detailed terrain! Got this from a recent pc hardware magazine. I think crysis wiki also mentions it.

The reason I believe they did this is because traditional rendering of tesselated near-photo realistic terrain would simply consume too much bandwidth. So they went with a cpu assisted method that is much faster in execution.

Even if you do away with GPUs you still need an API.

GPUs are ridiculously fast when it comes to certain types of calculations, and you probably wont see them go away for quite some time. It's not like pixar doesn't use GPUs for their real time apps where they make everything. Throwing hundres of CPUs at something thinking you'll get unlimited processing powers is not a viable solution. The design problems associated with it are monstrous, and having that many CPUs creates power problems. We're in the green house age after all.

Voxels are great for some things, yes, but they won't replace polygons.

Quantum computers are something radically different so don't bring them into the discussion.

AMD is already planning an architecture that combines the CPU and GPU as one chip, sounds like your prediction is becoming true.hamidious

These sound like baby steps to me, but ATI and AMD may make the winning combination.

The thing is, they are already working on CPU/GPU unification. This is precisely why AMD bought ATI, because both mindsets are needed in order to research the area. Nvidia is also closely working with Intel to achieve the same thing. However, a unified CPU/GPU architecture would be more like a CPU and GPU combined, rather then a super fast CPU. Also, this unification process is going to take a long time to perfect.

Even then, they arn't simply going to get rid of graphical APIs like DX - they are needed to simplify development. DX10 in itself is a beast and took Microsoft three years to create their first release. Development houses arn't going to want to write their own APIs, considering how much overhead would be involved.

While I agree software rendering allows people to use more graphical techniques, it is not practical for the real world in the least. That is why hardware is created, in order to speed up graphical techniques. Using Quake 2 as an example; many people played that game in software rendering mode, but at a much lower framerate and quality then someone who bought a 3dfx with OpenGL support. This is because the hardware sped up the processing involved.

In fact, all the DX10 techniques we see today were once done in software. Many of it's new techniques are very old. The only difference is, research brought us the ability to break down the software into circuitry.

I agree, one day we will see techniques like ray tracing become possible in realtime (in games). In fact, ray tracing is actually very efficient, however it has a very big 'constant' to it's algorithm, which means we have to wait for a gpu that can handle it.

The bottom line is though, hardware will always be used to replace software techniques (in gaming). Software is plainfully slow, because the logic being implemented then has to be interpreted by a vast number of components until it is finally executed. However with hardware, once the instruction is fired, it is almost instantly completed. The only time you will see software implementation is with non-time critical rendering, ie: rendering a 3D scene for a movie. Even then, most "rendering farms" have been using GPUs to handle much of their processing.

Even if you do away with GPUs you still need an API.

GPUs are ridiculously fast when it comes to certain types of calculations, and you probably wont see them go away for quite some time. It's not like pixar doesn't use GPUs for their real time apps where they make everything. Throwing hundres of CPUs at something thinking you'll get unlimited processing powers is not a viable solution. The design problems associated with it are monstrous, and having that many CPUs creates power problems. We're in the green house age after all.

Quantum computers are something radically different so don't bring them into the discussion.

kyrieee

I am aware that a general purpose cpu would experience problems with the massively parallel calculations needed to fulfill gpu tasks, that is why I've been peddling the hugely efficient yet horribily neglected 'voxel' rendering regime, but more importantly in my fusion of all the components, I stressed a multicore design, suggesting certain cores specifically designed around the massively parallel process of the gpu, to maximize their graphical output, and others optimized for things like traditional logic computation and yet others designed with physics in mind. Combine all these cores on a single die and you get what I want.

Once again I refer you to the game called outcast.

Also, I think most of you underestimate the core2duo line of processors.

I have a low end core2duo 5300 in my laptop, and a p4 3.4 ghz in my desktop. I did some benchmarks, lo and behold, my little laptop is essentialy 4x faster in real world applications! It's as if my p4 was running at 13.6ghz, the slow bastard.

It's at this time that I'd like to highlight a relevant point about Crysis in particular, it will use voxel based software rendering for it's highly detailed terrain! Got this from a recent pc hardware magazine. I think crysis wiki also mentions it.

The reason I believe they did this is because traditional rendering of tesselated near-photo realistic terrain would simply consume too much bandwidth. So they went with a cpu assisted method that is much faster in execution.

Gangans

Well according to wikipedia they used voxels becuase :

"Voxel Object: Allows creating geometry a heightmap system wouldn't support to create cliffs, caves, canyons and alien shapes. Voxel editing is as easy as heightmap editing and fast in rendering."

Becuase thenormal heightmaps don't support the ideas that they want to do. Not becuase GPU's don't have enough "bandwidth".

Altough I'll admit thatthey do say, "fast in rendering".

I think at a certain point the limitations in games will not be because of the hardware but because of the software. Most pc games now really don't benefit much from having the top of the line screaming fast processors. I think it's more up to the developers to actually take advantage of the hardware available now. My computer is mid-range and I have yet to playany game that actually taxes it even when I have two or three other programs running in the background. I'm more interested in seeing where in game AI goes rather than graphics. It's nice if the enemies in my FPS have individually rendered eyelashes but who cares if they're too stupid to duck and cover when I shoot at them? Still, I agree that at some point the cores of a cpu will be assigned to specific tasks like graphics, sound, etc ... Very good thread.

The thing is, they are already working on CPU/GPU unification. This is precisely why AMD bought ATI, because both mindsets are needed in order to research the area. Nvidia is also closely working with Intel to achieve the same thing. However, a unified CPU/GPU architecture would be more like a CPU and GPU combined, rather then a super fast CPU. Also, this unification process is going to take a long time to perfect.

Even then, they arn't simply going to get rid of graphical APIs like DX - they are needed to simplify development. DX10 in itself is a beast and took Microsoft three years to create their first release. Development houses arn't going to want to write their own APIs, considering how much overhead would be involved.

While I agree software rendering allows people to use more graphical techniques, it is not practical for the real world in the least. That is why hardware is created, in order to speed up graphical techniques. Using Quake 2 as an example; many people played that game in software rendering mode, but at a much lower framerate and quality then someone who bought a 3dfx with OpenGL support. This is because the hardware sped up the processing involved.

In fact, all the DX10 techniques we see today were once done in software. Many of it's new techniques are very old. The only difference is, research brought us the ability to break down the software into circuitry.

I agree, one day we will see techniques like ray tracing become possible in realtime (in games). In fact, ray tracing is actually very efficient, however it has a very big 'constant' to it's algorithm, which means we have to wait for a gpu that can handle it.

The bottom line is though, hardware will always be used to replace software techniques (in gaming). Software is plainfully slow, because the logic being implemented then has to be interpreted by a vast number of components until it is finally executed. However with hardware, once the instruction is fired, it is almost instantly completed. The only time you will see software implementation is with non-time critical rendering, ie: rendering a 3D scene for a movie. Even then, most "rendering farms" have been using GPUs to handle much of their processing.rimnet00

I agree with everything you say. However in my initial post I was speaking of 20 years into the future. One can't predict what will happen in 5 years when it comes to things like gpu's and cpu's, let alone 20. I was just postulating a hypothetical based on the constant trend of unification/intergration of various components into ever smaller spaces when it comes to cpu and gpu design.

I don't believe we will still have separate graphics cards in 20 years according to such trends. The humble audio accelerator card has basically died, with the ability of cpu's to emulate all such hardware functions in software easily. I think this is where gpu's are headed, not soon mind you, but in 20 or more years, which is an era when it comes to computer technology.

[QUOTE="rimnet00"]

The thing is, they are already working on CPU/GPU unification. This is precisely why AMD bought ATI, because both mindsets are needed in order to research the area. Nvidia is also closely working with Intel to achieve the same thing. However, a unified CPU/GPU architecture would be more like a CPU and GPU combined, rather then a super fast CPU. Also, this unification process is going to take a long time to perfect.

Even then, they arn't simply going to get rid of graphical APIs like DX - they are needed to simplify development. DX10 in itself is a beast and took Microsoft three years to create their first release. Development houses arn't going to want to write their own APIs, considering how much overhead would be involved.

While I agree software rendering allows people to use more graphical techniques, it is not practical for the real world in the least. That is why hardware is created, in order to speed up graphical techniques. Using Quake 2 as an example; many people played that game in software rendering mode, but at a much lower framerate and quality then someone who bought a 3dfx with OpenGL support. This is because the hardware sped up the processing involved.

In fact, all the DX10 techniques we see today were once done in software. Many of it's new techniques are very old. The only difference is, research brought us the ability to break down the software into circuitry.

I agree, one day we will see techniques like ray tracing become possible in realtime (in games). In fact, ray tracing is actually very efficient, however it has a very big 'constant' to it's algorithm, which means we have to wait for a gpu that can handle it.

The bottom line is though, hardware will always be used to replace software techniques (in gaming). Software is plainfully slow, because the logic being implemented then has to be interpreted by a vast number of components until it is finally executed. However with hardware, once the instruction is fired, it is almost instantly completed. The only time you will see software implementation is with non-time critical rendering, ie: rendering a 3D scene for a movie. Even then, most "rendering farms" have been using GPUs to handle much of their processing.Gangans

I agree with everything you say. However in my initial post I was speaking of 20 years into the future. One can't predict what will happen in 5 years when it comes to things like gpu's and cpu's, let alone 20. I was just postulating a hypothetical based on the constant trend of unification/intergration of various components into ever smaller spaces when it comes to cpu and gpu design.

I don't believe we will still have separate graphics cards in 20 years according to such trends. The humble audio accelerator card has basically died, with the ability of cpu's to emulate all such hardware functions in software easily. I think this is where gpu's are headed, not soon mind you, but in 20 or more years, which is an era when it comes to computer technology.

My main issue is though, your hypothesis is talking about devolving the entire field of computer graphics. Like I mentioned before, CPUs and GPUs are in the process of being unified. In other words, a multiple CPU and GPU diodes on one chip. This does not translate to cpu's in the future emulating grapical techniques. Ray tracing as you mention in your first post is actually a 20 some odd creation, and it still isn't possible to do efficiently in real time. Imagining CPU taking over graphical processing, even in 20 years is far fetched, and once again devolving. This isn't like audio techniques being emulated on the CPU, the audio emuation example you refer to is much less complicated.

It is nice to see that there are people who are looking forward. I am just letting you know, as someone who has studied in this field, I know it's not a feasible notion.

The thing is, they are already working on CPU/GPU unification. This is precisely why AMD bought ATI, because both mindsets are needed in order to research the area. Nvidia is also closely working with Intel to achieve the same thing. However, a unified CPU/GPU architecture would be more like a CPU and GPU combined, rather then a super fast CPU. Also, this unification process is going to take a long time to perfect.

Even then, they arn't simply going to get rid of graphical APIs like DX - they are needed to simplify development. DX10 in itself is a beast and took Microsoft three years to create their first release. Development houses arn't going to want to write their own APIs, considering how much overhead would be involved.

While I agree software rendering allows people to use more graphical techniques, it is not practical for the real world in the least. That is why hardware is created, in order to speed up graphical techniques. Using Quake 2 as an example; many people played that game in software rendering mode, but at a much lower framerate and quality then someone who bought a 3dfx with OpenGL support. This is because the hardware sped up the processing involved.

In fact, all the DX10 techniques we see today were once done in software. Many of it's new techniques are very old. The only difference is, research brought us the ability to break down the software into circuitry.

I agree, one day we will see techniques like ray tracing become possible in realtime (in games). In fact, ray tracing is actually very efficient, however it has a very big 'constant' to it's algorithm, which means we have to wait for a gpu that can handle it.

The bottom line is though, hardware will always be used to replace software techniques (in gaming). Software is plainfully slow, because the logic being implemented then has to be interpreted by a vast number of components until it is finally executed. However with hardware, once the instruction is fired, it is almost instantly completed. The only time you will see software implementation is with non-time critical rendering, ie: rendering a 3D scene for a movie. Even then, most "rendering farms" have been using GPUs to handle much of their processing.

rimnet00

Actually,people have integrated ray-tracing into old games such as Quake 2,so it may be in the near future.Greyhound222

Very good post. I felt like I was reading a Pop Sci article or something. What's your line of work?

Let's skip this time consuming sequential evolution of computer graphics, and move by leaps and bounds.

Gangans

In reference to this, it makes sense that when companies make a leap in the advancement of their technology, they release it when it's profitable. It's the same across all markets, what something can offer matters less than how much money it can make. They'll want to squeeze their cash cows till the milk runs dry, then advance in increments, and not introduce the latest technology (at least at affordable prices) till they stop making money on everything in-between.

AMD is already planning an architecture that combines the CPU and GPU as one chip, sounds like your prediction is becoming true.hamidious

From reading that, I think my school failed mePlatearmor_6

[QUOTE="Greyhound222"]Actually,people have integrated ray-tracing into old games such as Quake 2,so it may be in the near future.rimnet00

hmm, you do amke a every good poitn because CPU frequencies are MUCH higher than GPU frequencies and correct me if im wrong but i dont think there is a gpu out there that goes ghz. one thing that might be aproblem is that a gfx card is liek its own computer, not just a processor. It ahs its own mobo, cpu(Gpu) and memory. excellent post though, a truly revolutionary idea.Deamon321

Graphics cards also have very specific designs that work better for processing 3d graphics.

I always have the tendency to think that marketing and hungre for maximum profit always blocks the evolution of technology, especially computer technology. I have the feeling that the biggest computer science companies always are 4 or 5 steps ahead of what we call reality. This in order to make their research fully profitable.

What I see in such posts is that the human mind can push the limits very far...if you put 20 like minded people together it goes even further and faster. Yet in our reality,it evolves relativly slow, babysteps, one after another.

[QUOTE="rimnet00"][QUOTE="Greyhound222"]Actually,people have integrated ray-tracing into old games such as Quake 2,so it may be in the near future.Greyhound222

After reading all the utter crap in the E3 forum, this post WOKE UP MY MIND!

Seriously though, you have a great writing style.

Onto the point, at the moment, I do not think acore2duo quad could render crysis-like graphics. Although I am sure your theory is correct, and in the not-too-distant future, we will have combined GPU's and CPU's into one powerful processing unit. This might actually come sooner than expected, as just last month or so, Nvidia released a super computer of sorts to be used medical research, and it seems that they are gaining experience in fields other than just GPUs.

rimnet00 wrote:
My main issue is though, your hypothesis is talking about devolving the entire field of computer graphics. Like I mentioned before, CPUs and GPUs are in the process of being unified. In other words, a multiple CPU and GPU diodes on one chip. This does not translate to cpu's in the future emulating grapical techniques. Ray tracing as you mention in your first post is actually a 20 some odd creation, and it still isn't possible to do efficiently in real time. Imagining CPU taking over graphical processing, even in 20 years is far fetched, and once again devolving. This isn't like audio techniques being emulated on the CPU, the audio emuation example you refer to is much less complicated.

I think we should bring up radical new ideas if we want to instigate revolutionary change. If we allow market trends and baby steps to reign, we will simply take longer to get to point B. I guess I'm just a visionary and all-too-eager to see the future. ;)

rimnet00 wrote:
It is nice to see that there are people who are looking forward. I am just letting you know, as someone who has studied in this field, I know it's not a feasible notion.

elmertheowl wrote:
Very good post. I felt like I was reading a Pop Sci article or something. What's your line of work?

elmertheowl wrote:
In reference to this, it makes sense that when companies make a leap in the advancement of their technology, they release it when it's profitable. It's the same across all markets, what something can offer matters less than how much money it can make. They'll want to squeeze their cash cows till the milk runs dry, then advance in increments, and not introduce the latest technology (at least at affordable prices) till they stop making money on everything in-between.

FragMonkey09 wrote:
After reading all the utter crap in the E3 forum, this post WOKE UP MY MIND!

Seriously though, you have a great writing ****

Onto the point, at the moment, I do not think acore2duo quad could render crysis-like graphics. Although I am sure your theory is correct, and in the not-too-distant future, we will have combined GPU's and CPU's into one powerful processing unit. This might actually come sooner than expected, as just last month or so, Nvidia released a super computer of sorts to be used medical research, and it seems that they are gaining experience in fields other than just GPUs.

As to your other point, AMD/NVIDIA are actively tinkering with cpu/gpu fusion. I refer you to this article which is only 4 hours old (since this post was made): http://www.dailytech.com/AMD+Talks+GPGPU+and+Fusion/article8066c.htm

This idea would seem even more feasible once manufacturers go to man-made daimond and phases out silicone.Hondo189

I think nano-tubes in place of copper wires and photons in place of electrons would be even more interesting to try. A light run nano-processor that doesn't even require electrical input and barely even generates any heat, yet is millions of times faster.

Ah thanks for that link. My prediction is that this will start out slowly. Once AMD releases this new GPGPU, it will be something like the Aegia PhysX processor, that nobody is currently buying because games do not support it. Given enough time, probably not as much as the 20 years you predict, this could become mainstream.

Another thing is, the GPGPU AMD is working on is not really just a CPU that handles graphics too, its seems like its basically a GPU "fused" into a CPU, hence the project "Fusion." If this does become a reality though, how does one upgrade the GPU part of the CPU? Buy a whole CPU/GPU Fusion all over again?

EDIT: It seems to have censored the word "style" when you quoted me... :/

Ah thanks for that link. My prediction is that this will start out slowly. Once AMD releases this new GPGPU, it will be something like the Aegia PhysX processor, that nobody is currently buying because games do not support it. Given enough time, probably not as much as the 20 years you predict, this could become mainstream.

Another thing is, the GPGPU AMD is working on is not really just a CPU that handles graphics too, its seems like its basically a GPU "fused" into a CPU, hence the project "Fusion." If this does become a reality though, how does one upgrade the GPU part of the CPU? Buy a whole CPU/GPU Fusion all over again?

EDIT: It seems to have censored the word "style" when you quoted me... :/

FragMonkey09

One can always get a separate gpu and stick it into their pci-e slot as they do now if the gpu accompanying their cpu becomes obsolete. The two could even be made to work in unison. But I agree it wouldn't really be a great change.

What amd/ati are doing seems to be more like an onboard intergrated video solution, in other words, more baby steps.

But it would be good if one could get moderate 3d acceleration with just a cheap cpu.

This whole post is very interesting, here's my noobish input.

First off, stop trying to use very big words in gigantic paragraphs to try and scare each other off. Not working, and it makes the entire thread harder to grasp. I got the point, being a moderate computer person. But wehn you start railing off and your blah blah blah in your debates...settle down, you're all smart, let it fall.

Now, as for fast processors...twenty years is a long time. Did anybody in the year 1987 think of 3D games? Heck no! We were all too busy playing Zelda, wondering how they could make the world bigger. Graphics smaphics. We were not interested in hardware, we were interested in ideas. A new age, of whatever Nintendo could bring to our TV and such. I don't think anyone concieved the whole notion of 3D graphics until everything got more powerful, and Wolfenstein 3D decided to take it further.

But that's the thing I'm thinking: we have no idea what's coming next. Most Mario gamers didn't understand 3D anyway, much less guess that their favorite plumber would be bunching off of polygonal walls a decade later. Who are we to guess what will be coming aroudn the mountain? Holograms? VR?

As for the hardware, that's the last of my worries. Graphics will be increasing, but the increase will always be an exponentially decreasing slope. Every gap between the PS6 and Xbox 2160 will be miniscule, because the developers will be wanting that little increase in the graphics. Every drop of sweat will cost a new processor by the time we're finished, and the whole thing will just suddenly end. And where will the industry be when there aren't any graphics, just one life-perfect engine that is the only thing that needs to exist any more?

That's where the industry is going. Perhaps people will not find a need for graphics cards when CPU's can run everything with style, but I doubt it. As processors go up, the games and proigrams follow suit. Do you think that Halo 2 needs the system that it should to run it? Heck no, I might be able to run the damn thing on my crappy box if they did a good job with it. But they're taking initiative with the possiblities, and so we get...well, shinier suits and a demand for the next faster unit.

Besdies, I'm not the only one on this forum who would defend a gameplay-awesome game that looks like crap. I play StarCraft all the time, and I love my DS, even though it isn't very good for graphics. I'm a sucker for pretty faces, sure, but for StarCraft II, they could do it in a slightly better 2D engine and I'd be happy. Make the AI awesome and behaive like real people, and I don't care how nice it looks.

Onme last thing: why is it that we are so amazed at computer graphics over real life? Crysis footage makes me gasp, but the bluffs around my area bore me. Huh.

So what exactly is ray tracing? From reading about the description on wikipedia, it seems like it's just an accurate representation of light being casted based on light being seen from your eyes.onemic