Dual Gpu console revisited

[Bill]

The X360 has driven home for me that consoles have lost that wow factor.

Maybe PS3 can come through..that remains to be seen.

Because the incremental improvement of PC's.

The way to achieve a true leap is to include not one, but TWO high end GPU's.

Imagine if Ps3 had come out in Spring 2004. Had it included two 6800 Ultra's, in internal SLI, it would still stand up today to a 7800 GTX. We'd be talking 32 pipes, much better utilized than a SLI rig.

Price it at 499. I'll pay. People lined up for 399 Xbox.

The pricing breakdown of the 360 show the GPU cost at $100-$140..I think that's doable.

I realize these threads haven't gone well in the past, but it's really my opinion.

How much bandwidth would that take? What kind of designs would make that work? Etc.

 

[Deepak]

This is a futile exercise. Next gen consoles will/already have some of the most powerful and cutting-edge hardware in the market and anyway those SLIed GPUs will never be efficiently used like console GPUs.

 

[AlphaWolf]

The X360 has driven home for me that consoles have lost that wow factor.

Maybe PS3 can come through..that remains to be seen.

Because the incremental improvement of PC's.

The way to achieve a true leap is to include not one, but TWO high end GPU's.

Imagine if Ps3 had come out in Spring 2004. Had it included two 6800 Ultra's, in internal SLI, it would still stand up today to a 7800 GTX. We'd be talking 32 pipes, much better utilized than a SLI rig.

Price it at 499. I'll pay. People lined up for 399 Xbox.

The pricing breakdown of the 360 show the GPU cost at $100-$140..I think that's doable.

I realize these threads haven't gone well in the past, but it's really my opinion.

How much bandwidth would that take? What kind of designs would make that work? Etc.

Its not just the cost of the 2nd gpu. It snowballs into a whole bunch of other requirements.

 

[The GameMaster]

The X360 has driven home for me that consoles have lost that wow factor.

Maybe PS3 can come through..that remains to be seen.

Because the incremental improvement of PC's.

The way to achieve a true leap is to include not one, but TWO high end GPU's.

Imagine if Ps3 had come out in Spring 2004. Had it included two 6800 Ultra's, in internal SLI, it would still stand up today to a 7800 GTX. We'd be talking 32 pipes, much better utilized than a SLI rig.

Price it at 499. I'll pay. People lined up for 399 Xbox.

The pricing breakdown of the 360 show the GPU cost at $100-$140..I think that's doable.

I realize these threads haven't gone well in the past, but it's really my opinion.

How much bandwidth would that take? What kind of designs would make that work? Etc.

Well... the first problem would be the most obivious problem. Heat and power... as it stands a GPU that consumes as much power and puts out as much heat as the Geforce 6800 Ultra did is not workable in a console enviroment... to say the least two. Two Geforce 6800 Ultras at 90nm might be doable in a console, but it still be difficult to make it work at an acceptable level. Console enviroments are limited in the cooling they can have as well as the power supply that can be used with it... cost is of course another issue as you need to manufacture the processors as well as the memory banks for both GPUs. Currently the closest we are going to get to dual processors in a console is XENOS in the XBox360 as it IS two individual cores, but it is more accurate to say it is a split processor where one core has certain tasks and the other has certain other tasks. Multicore GPUs as well as MultiGPU chipsets are the future as it is looking like right now, but that also means the costs of graphics chipsets are likely to balloon as a result even more so than it has now. I do not see a SLI or other such MultiGPU setup in any future console in the next decade for reasons of cost and other mitigating problems from heat and power.

On a unrelated note... I really don't see how you can't be impressed with the XBox360 has accomplished ALREADY in the graphics and performance of those games on the system... especially when it equals or exceeds that of even a SLI Geforce 7800/Radeon x1800 $3000+ dollar gaming rig. Call of Duty 2 on the XBox360 is an extremely good port of the PC version and I can't say I have seen anything on the PC (currently) that would come close to the likes of Kameo or Project Gotham Racing 3 in terms of graphics. I would say what it accomplished as a first generation game on a console that is pretty much unknown at this point and radically different from the previous console architecturally it was pretty impressive. The TRUE test is still to come for the XBox360 though (and the PS3 for that matter)... it is generally not a good idea to judge the system by it's first generation games. Those games coming out a couple of years from now WILL be extremely impressive and may as well exceed the curve that games on the PC have.

I kept my expectations realistic for the XBox360 and they was met overall as the XBox360 exceeds even my high end gaming computer in terms of graphics and performance, but the best is yet to come... and it will only get better.

 

[Bill]

Its not just the cost of the 2nd gpu. It snowballs into a whole bunch of other requirements.

Yeah I know. That's the kicker. So I was wondering how those can be worked around.

I dont see a point THIS GEN as both competitors are using one GPU. But I feel it's fairly inevitable, almost.

If you REALLY want to push the technology envelope, that's how you'll have to do it.

It would be cheaper than doing one chip, for yield reasons.

Also, realize in a console the programming is fine tuned..you will get max performance, not like a buggy SLI PC rig.

Gamemaster, in some ways I'm impressed, but in others I'm not. All in all it's just like a high end PC graphically. There's no wow factor. I really miss it.

Gears of War could have wow factor..but it's slated for July.

 

[Urian]

The problem is:

-RSX 300 milion transistors.
-NV40 220 milion transistors.
-NV40x2= 440 milion transistors.

 

[wireframe]

Imagine if Ps3 had come out in Spring 2004. Had it included two 6800 Ultra's, in internal SLI, it would still stand up today to a 7800 GTX. We'd be talking 32 pipes, much better utilized than a SLI rig.

And there we have it, in bold no less, for all to see. Why would SLI be better utilized in a console than a PC? For the same reasons that a single GPU can and will be better utilized in a console? So why the need for SLI in the console?

Take a look at PS 2 and Xbox (1) and tell me that they don't have graphics that would otherwise require a much higher specced GPU in a PC to achieve the same results.

 

[Bill]

And there we have it, in bold no less, for all to see. Why would SLI be better utilized in a console than a PC? For the same reasons that a single GPU can and will be better utilized in a console? So why the need for SLI in the console?

Take a look at PS 2 and Xbox (1) and tell me that they don't have graphics that would otherwise require a much higher specced GPU in a PC to achieve the same results.

The "need" is to essentially jump ahead a performance generation.

And yeah, imagine what Xbox could have done with two NV2A's, then. See my point? Yes they'll be used more efficiently, all the more reason it would be impressive.

Dual NV2A's would have made Xbox an 8 pipe card, doesn't sound that impressive, but an 8 pipe card can run Far Cry on PC in high res, which is a gorgeous game, at very high frame rates.

http://farcry.gamesweb.com/farcry/img/screenshots/serie07/far_cry05.jpg

Not too mention a perfect port of Half Life2. Hell 720P could have come early, if that's how Devs wanted to use the power.

 

[wireframe]

And yeah, imagine what Xbox could have done with two NV2A's, then. See my point?

Of course, you'd need to double up the CPU to take advantage of this graphical edge by making something actually interesting happen with it all, right? You'd need to double up on video memory so you can actually fuel those glorious two NV2As. right?

Yes they'll be used more efficiently, all the more reason it would be impressive.

But we know that a single GPU with twice the power of single GPUs in SLI is preferrable. At least I know this.

I'm not buying the "need" part. If nobody else has anything that is twice as graphically impressive, why should you worry? You need to sell an incremented solution next round, right? So, you'd only be setting yourself up to always doubling whatever you could have had in single configuration. This may not sound like a lot when we are talking 100M transistor GPUs today, knowing there are 300M ones, but what if the next console has a billion and you now need two billion to go SLI. This would only make sense if you have to and I don't think the first-mover advantage is there, especially with something like SLI. I could somewhat understand someone saying "to heck with it! Let's increase the transistor budget 30% and see how things turn out."

SLI more-or-less implies dual memory interfaces and a lot of redundant complexity. This can make sense in a PC when you have lots of configuration options, but not in a closed system. You don't need flexibility. You need simplicity and cost effectiveness.

Dual NV2A's would have made Xbox an 8 pipe card, doesn't sound that impressive, but an 8 pipe card can run Far Cry on PC in high res, which is a gorgeous game, at very high frame rates.

I thought Far Cry on Xbox (1) was a good example of what a lowly NV2A could do in a closed system. I haven't actually seen it first hand, but I read that it is rather impressive, especially when compared to what hardware you need to achieve similar results on a PC.

As for frame rates, again I point you to a current example: Call of Duty 2. By all accounts this game looks very similar on the PC (high settings) and Xbox 360. However, the Xbox version does not suffer from frame rate dips. Furthermore, super high FPS are squandered on a console because of the refresh rate of the output device (TV) and the total lack of benchmarking. As long as it runs smoothly it is good. There are no "oh yeah? well, the Playstation 5000 runs it 15% faster!" discussions when it comes to consoles.

EDIT (partially because the quoted post was edited):

You made a very good point by bringing up Half Life 2. Two NV2As could not do it. They'd still be lacking SM 2.0, for example. Suddenly you'd be sitting on doubled up "yesterday's news". (but I do see the performance argument you are making. I am just saying "why?" to that part.)

 

[Bill]

Of course, you'd need to double up the CPU to take advantage of this graphical edge by making something actually interesting happen with it all, right? You'd need to double up on video memory so you can actually fuel those glorious two NV2As. right?


But we know that a single GPU with twice the power of single GPUs in SLI is preferrable. At least I know this.

I'm not buying the "need" part. If nobody else has anything that is twice as graphically impressive, why should you worry? You need to sell an incremented solution next round, right? So, you'd only be setting yourself up to always doubling whatever you could have had in single configuration. This may not sound like a lot when we are talking 100M transistor GPUs today, knowing there are 300M ones, but what if the next console has a billion and you now need two billion to go SLI. This would only make sense if you have to and I don't think the first-mover advantage is there, especially with something like SLI. I could somewhat understand someone saying "to heck with it! Let's increase the transistor budget 30% and see how things turn out."

SLI more-or-less implies dual memory interfaces and a lot of redundant complexity. This can make sense in a PC when you have lots of configuration options, but not in a closed system. You don't need flexibility. You need simplicity and cost effectiveness.


I thought Far Cry on Xbox (1) was a good example of what a lowly NV2A could do in a closed system. I haven't actually seen it first hand, but I read that it is rather impressive, especially when compared to what hardware you need to achieve similar results on a PC.

As for frame rates, again I point you to a current example: Call of Duty 2. By all accounts this game looks very similar on the PC (high settings) and Xbox 360. However, the Xbox version does not suffer from frame rate dips. Furthermore, super high FPS are squandered on a console because of the refresh rate of the output device (TV) and the total lack of benchmarking. As long as it runs smoothly it is good. There are no "oh yeah? well, the Playstation 5000 runs it 15% faster!" discussions when it comes to consoles.

That's like saying Sony doesn't need to spend billions of dollars cranking out cell. Just slap a Pentium 4 in there.

The fact is more power is very useful. Both sides are pushing the technology envelope and hard.

Far Cry was on Xbox, but at 640X480 and I bet it lacked the sweeping vistas, ya know?

Oh there are no the playstation 500 runs it 15% faster discussions? That's pretty much ALL WE DO.

The problem with increasing transistor budgets is there seems to a limit based on process. For example this go round it seems to be 300m. If you made a 450m GPU, I think you'd have huge problems. That's kinda the whole point of SLI on PC as far as I can tell. Sure Nvidia could MAKE a 48 pipe GTX..actually they probably couldn't.

The memory issue is a problem..would there be a way to run two GPU's from the same pool?

 

[Guden Oden]

The way to achieve a true leap is to include not one, but TWO high end GPU's.

Imagine if Ps3 had come out in Spring 2004. Had it included two 6800 Ultra's, in internal SLI, it would still stand up today to a 7800 GTX. We'd be talking 32 pipes, much better utilized than a SLI rig.

This is a silly idea. SLI is in of itself an inefficient idea. First of all you need to duplicate all textures and frame buffers between the two chips. Second, it complicates certain rendering techniques (which are becoming very mainstream I might add). Any software that requires render-to-texture effects, or depends on past frames or such will if not fall then at least stumble on SLI, which we've already seen on the PC with various problems shown by any number of games. There's no reason it would work better on a console.

Your idea is B.A.D. from the outset; you claim greater efficiency (with no fact to back it up), and it simply isn't the case. Sorry. Terrible thread.

 

[wireframe]

That's like saying Sony doesn't need to spend billions of dollars cranking out cell. Just slap a Pentium 4 in there.

No, it's like saying Sony needs to invest billions to develop Cell, but don't need to slap two Cells in every console and increase variable costs just for the sake of having more power than can be used at this period in time.

The fact is more power is very useful. Both sides are pushing the technology envelope and hard.

They sure are. Nobody's arguing against that. You brought up the possible need to double up the GPU, thereby roughly doubling the performance of what current tehnology allows (in certain ways). I am arguing against that statement and that statement alone.

Far Cry was on Xbox, but at 640X480 and I bet it lacked the sweeping vistas, ya know?

It lacks SM 2.0 for sure.

Oh there are no the playstation 500 runs it 15% faster discussions? That's pretty much ALL WE DO.

Right, but you do that on theory. My point was that you won't be reading online reviews with pretty little graphs showing one console outpacing the other. This cannot easily be shown. Perhaps this is with good reason because no two consoles run the "same" game looking exactly the same or even with the same assets. One reason SLI/CrossFire is interesting to provide as a solution is to use it to show technological prowess and high benchmarking numbers. These are numbers that shouldn't matter to most people because the aren't running SLI, but it still influences buying behavior.

Look at the PC market. For a cosole, something like the 7800 GTX 512MB would make a lot more sense than 7800 GTX 256MB SLI. Don't you agree?

 

[Bill]

No, it's like saying Sony needs to invest billions to develop Cell, but don't need to slap two Cells in every console and increase variable costs just for the sake of having more power than can be used at this period in time.


They sure are. Nobody's arguing against that. You brought up the possible need to double up the GPU, thereby roughly doubling the performance of what current tehnology allows (in certain ways). I am arguing against that statement and that statement alone.


It lacks SM 2.0 for sure.


Right, but you do that on theory. My point was that you won't be reading online reviews with pretty little graphs showing one console outpacing the other. This cannot easily be shown. Perhaps this is with good reason because no two consoles run the "same" game looking exactly the same or even with the same assets. One reason SLI/CrossFire is interesting to provide as a solution is to use it to show technological prowess and high benchmarking numbers. These are numbers that shouldn't matter to most people because the aren't running SLI, but it still influences buying behavior.

Look at the PC market. For a cosole, something like the 7800 GTX 512MB would make a lot more sense than 7800 GTX 256MB SLI. Don't you agree?

Sorry but I have to disagree. The idea that PS3 will torch Xbox360 in power is common everywhere from Gamespot to the layman on the street.

For the two GTX argument..not neccesarily. I think the dual setup would offer greater raw power. However that's not the point. The point is TWO 512's would offer a step beyond anything else currently achievable.

I know SLI in a PC uses redundant RAM. This COULD be a big enough issue to kill the idea. However, isn't there a way to virtualize two GPU's as one?

SLI is just a term I'm using. It could be implemented differently. SLI has limitations based on the PC.

And Guden Oden, funny, I just read Nvidia CEO claiming "SLI will become more important going forward", not less.

I agree SLI has problems in a PC setup, again that's not really my point. Overall it has been a huge success for Nvidia though, such that ATI was forced to copy them.

 

[Belmontvedere]

everything gets outdated, that's a fact of life (as far as computers go atleast). the concept of something being "future proof" is a farce. by 2009 the "next-gen" consoles will be pretty mundane technologically. even with dual GPUs that would still be true. their is literally nothing that can be done about that. even if X360 had come out with 1GB of ram with 47 CPUs and 5 GPUs it would still get outdated eventually. the difference is it would cost about $5,000 (but it would still get outdated!).

 

[Bill]

This is true.

However it does nothing to negate the fundamental value of the idea.

 

[wireframe]

Sorry but I have to disagree. The idea that PS3 will torch Xbox360 in power is common everywhere from Gamespot to the layman on the street.

I'm gonna assume you mean processing power here and not electrical power. Doesn't this negate your whole argument? If PS 3 can "torch" the Xbox 360 with a single CPU and a single RSX, why would they want to put two in there?

Don't you see where I am coming from here? People compare consoles and say "well, I think this one has better graphics." Almost nobody is saying that even the top console is crap and needs twice the processing power to entertain them. That is, until you came along.

For the two GTX argument..not neccesarily. I think the dual setup would offer greater raw power. However that's not the point. The point is TWO 512's would offer a step beyond anything else currently achievable.

LOL. Stop! Nobody is saying that this isn't true. I think I, and others who have responded similarly, are thinking about the whole "box" here. So why not make the PS 3 twice as large, twice as expensive, twice as likely to fail, twice as expensive to replace, twice as costly to power....need I go on?

You are saying there is a demand for this. Perhaps you are right. I don't think you are because by looking at the market, people seem to be happy with PS 2s, Xboxes, and GameCubes just the way they are...and as comparable as they are.

I know SLI in a PC uses redundant RAM. This COULD be a big enough issue to kill the idea. However, isn't there a way to virtualize two GPU's as one?

If you think that segmented memory is the big killer you are not looking at some other basics like the costs (granted, it will be cheaper in some ways to go SLI than develop a single GPU that is twice as powerful in the same time frame). What about the fact that these GPUs needs to sit somewhere? It needs to be powered and it needs to be cooled. You also need even more logic and wiring to make this work.

SLI is just a term I'm using. It could be implemented differently. SLI has limitations based on the PC.

Sure, you used "dual GPU" in the title and that is fine. It still doesn't change the fact that having two GPUs in the device is prohibitive and maybe outright ludicrous when you consider the fact that everything is fine the way things are. (which you don't necessarily agree with, but millions of others do)

I understand wanting better graphics, but why wouldn't you argue for the consoles using GPUs that are a step ahead of what is available in the PC market instead? Why not have SM 5.0 GPUs in the consoles because they are closed systems so the features are "standard" even if they are not (so to speak).

EDIT:

SLI/CrosFire/Dual GPU works fine in a PC because it is an option. If you want to spend $1,000 on your graphics solution instead of $500, this option is there for you. A console cannot have these options. That means everyone has to be in for the $1,000 ride and this is seriously going to cut down the number of people willing to be entertained.

 

[Bill]

I'm gonna assume you mean processing power here and not electrical power. Doesn't this negate your whole argument? If PS 3 can "torch" the Xbox 360 with a single CPU and a single RSX, why would they want to put two in there?

Don't you see where I am coming from here? People compare consoles and say "well, I think this one has better graphics." Almost nobody is saying that even the top console is crap and needs twice the processing power to entertain them. That is, until you came along.



LOL. Stop! Nobody is saying that this isn't true. I think I, and others who have responded similarly, are thinking about the whole "box" here. So why not make the PS 3 twice as large, twice as expensive, twice as likely to fail, twice as expensive to replace, twice as costly to power....need I go on?

You are saying there is a demand for this. Perhaps you are right. I don't think you are because by looking at the market, people seem to be happy with PS 2s, Xboxes, and GameCubes just the way they are...and as comparable as they are.



If you think that segmented memory is the big killer you are not looking at some other basics like the costs (granted, it will be cheaper in some ways to go SLI than develop a single GPU that is twice as powerful in the same time frame). What about the fact that these GPUs needs to sit somewhere? It needs to be powered and it needs to be cooled. You also need even more logic and wiring to make this work.


Sure, you used "dual GPU" in the title and that is fine. It still doesn't change the fact that having two GPUs in the device is prohibitive and maybe outright ludicrous when you consider the fact that everything is fine the way things are. (which you don't necessarily agree with, but millions of others do)

I understand wanting better graphics, but why wouldn't you argue for the consoles using GPUs that are a step ahead of what is available in the PC market instead? Why not have SM 5.0 GPUs in the consoles because they are closed systems so the features are "standard" even if they are not (so to speak).

My whole point is about a competive edge.

The consoles are set in stone this generation, but what about next time around? Or three years ago when they were in the planning stages?

My point was people are aware of "power" differences, as portrayed in "better" graphics, because you said something to the effect of laypeople dont care, or that's how I read it. That was the only point of my comment about PS3. I'm trying to say laypeople DO care.

Sure, if PS3 is 50% more powerful, they've got it in the bag, no point in dual NOW. But three years ago they didn't have it in the bag. They didn't know what MS was doing.

Sure, everything is fine the way they are. But if MS had gone with not one but two Xenos this time around, and all the associated logic and wiring to support that, and Sony had not, guess what? The forward outlook might not favor Sony.

From a cost standpoint, MS could raise the price 100 bones, and with all the other improvements, still been far better off than with Xbox1 as far as losses.

Sony does not need Blu-Ray or Cell, I'd argue. There you go, paid for.

Competitive edge.

How about this:

Both consoles are using more or less as high end GPU's as possible. Similar transistor counts, clockspeeds, generations, everything.

What's the only way in that enviroment to jump way ahead of the other guy by more than a few percent?

 

[wireframe]

Competitive edge.

How is this competitive edge if anyone can do it by either 1) swallowing huge costs or 2) jacking up the price? If Xbox 360 was announced with dual GPUs and each of those was comparable to the RSX, don't you think Sony would have to respond in kind? Why do you think nobody wants to make this first move? I think the answer is simple: It won't be a competitive edge, it will just be a huge cost that somehow has to be managed.

Keep in mind that prices are sticky. If gaming consoles suddenly cost $1,000 people would stop and think before buying one. In the case of Xbox 360, let's pretend it had two Xenos and cost $800. Don't you think people would be more likely to wait to see about PS 3 on price alone than think "wow, this thing has two GPUs! Nothing will be able to beat it!"?

Nevermind the fact that there are cross-platform games that may completely ignore your processing advantage, opting instead to provide equal quality across all platforms. That's a problem that people are already concerned about with PS 3 (although I think this may be mostly marketing bollocks in the form of "wow, we are so powerful that maybe the World isn't ready for us!")

 

[London Geezer]

Sony does not need Blu-Ray or Cell, I'd argue. There you go, paid for.

Competitive edge.

Actually they kinda do. It's one of the very few ways to get a LOT of BDROM players out there. The more there are, the bigger their chance to "win" this bloody stupid next-gen format war with HDDVD.

So... next example? Cause this one didn't work...

 

[Bill]

I figured someone would come out with that London Boy..but what if for the extra $100 initially used on Blu Ray, again, MS doubled the PS3's rendering power roughly by including dual Xenos in Xbox360? Which Sony could then not cost effectively match due to the inclusion of Blu Ray?

You would see then that Blu Ray is like a Hard Drive. Not valuable in terms of other cost.

It's fine for them now, sure.