How can Nvidia be ahead of ATI but 360 GPU is on par with RSX?

[ihamoitc2005]

Memexport is not really to do with load balancing either.

Yes, but could it not be used to offload tasks from CPU, ie., physics, when graphical demands of scene is less?

 

[Alpha_Spartan]

Come now, MGS4 from what we have seen is one of the very best pieces of footage we have seen, period. The art, quality, and direction are excellent and the animation for the cut scene is a pure example of artistic genius. That capture the essense of human expression and have a very exciting trailer. The tech and the art are paired perfectly and the quality screams big budget.

You listed some things that I already acknowledged before and have nothing to do with graphics.

--Art: Subjective. Art has no bearing on the technical prowess of a film, videogame or render. It's possible to have crappy art direction and great graphics and vice versa.

--Direction: Subjective. Direction also has no bearing on the technical prowess of a film.

--Animation: Subjective. But it kinda does show the technical level of a game. However, it's moot in this case sense it wasn't a game.

I don't really care what platform it is on, what rendering tricks it uses, etc... it was great. That video is all MGS fans needs to be sold on the concept and I am sure it made a lot of PS fans really excited (and quite a few Xbox fans envious!) Really, the game looks great. The cut scene is in the Ruby class in quality and the choreography is second to none--I would put it as a tie with KZ (which, btw, if were a movie I would pay $8 to watch!)

As for whether it was exceptional or not, from a technical standpoint, is kind of pointless in that all games use shaders, particle effects, etc... Whatever rendering methods they used were obviously wisely chosen to emphasis the art. And to that degree I am not sure I can think of 5 other titles that have such a great dynamic between Art+Tech. So in that regards it is exception IMO.

Again, read my initial response to MGS4 in this thread. I didn't say I didn't like it. I didn't say it was bad. In fact, I said that I thought it was spectacular. But from a technical standpoint, there was nothing in there that impressed me.

 

[Nemo80]

Yes, but could it not be used to offload tasks from CPU, ie., physics, when graphical demands of scene is less?

No, not really. It's only used to let the CPU know the GPU has finished reading from the L2.

 

[mckmas8808]

Again, read my initial response to MGS4 in this thread. I didn't say I didn't like it. I didn't say it was bad. In fact, I said that I thought it was spectacular. But from a technical standpoint, there was nothing in there that impressed me.

So basically nothing has technically impressed you so far with the next-gen systems right?

 

[Shifty Geezer]

Again, read my initial response to MGS4 in this thread. I didn't say I didn't like it. I didn't say it was bad. In fact, I said that I thought it was spectacular. But from a technical standpoint, there was nothing in there that impressed me.

I'm really curious as to what you would consider technically spectacular.

 

[Alpha_Spartan]

Realistically we do know it, its just that many here don't want to hear/believe it so its easier not to say it.

Well, damn. You're the only one who can get away with immunity. I merely suggested that it COULD be this way and I got trashed.

It's not just that either. Any suggested weakness of the PS3 in general would attract the same negative response. Being fair demands showing both the pros and cons.

 

[EpicZero]

I'm really curious as to what you would consider technically spectacular.

What are all the announced games for the X-360 again?

(jk, jk)

 

[Dave Baumann]

No, not really. It's only used to let the CPU know the GPU has finished reading from the L2.

Errr, this is not memexport. Memexport is the capability of effectively letting the shader arrays have arbitary memory read /write capabilties.

 

[Alpha_Spartan]

I'm really curious as to what you would consider technically spectacular.

I already said that too. I said the Killzone PS3 renders. That shit got me excited! The funny thing is that certain people immediatly expected Xbox 360 to meet that level of quality graphically. However, since it was on the PS3 many neglected to realize that other PS3 games were expected to look at least that good. MGS4 isn't exempt no matter if Kojima is the Son of God or not. Since I'm neither a MGS fan or a KZ fan I am able to see things more objectively. I liked Killzone better. Alot better.

Kameo got me excited as well. The game looks sweet.

 

[Nemo80]

Errr, this is not memexport. Memexport is the capability of effectively letting the shader arrays have arbitary memory read /write capabilties.

Oh sorry, what was the name again?

 

[Acert93]

The opposite is the truth. Those ALUs are much less performant when it comes to "raw" pixel shading or raw vertex shading and it is yet to be seen if the pure number of them (48) is able to outperform a highly specialized 1337 pixel shader like in the RSX ,even if there are only 24 of them (+8 VS).

That brings new definition to "Extreme Pipelines"! ROTFLOL Move over Extreme Pipelines because 1337 Shaders are gonna kick your butt!! W00t!

Ahem. As noted in the past, the reason PS and VS can be unified is because PS already are capable of performing much of the work a VS does. The issue is not dumbing down the PS units, but making the VS units more capable.

Think of it as 48 fragment PS units that just happen to be capable of vertex shading. Obviously you have to take architecture and implimentation into consideration (a Pixel Shader units in R420 != G70 != Xenos). But ultimately these are programmable math units.

Adding an extra Mini-ALU to G70 did not make it less effecient than NV40. Adding capability to chip does not demand a performance drop. There could be one of course, but it does not stand to reason that it is a necessity, especially when Pixel Shader units already have most of what is necessary to process VS work. That is why the whole theory of USA works, because the work the shader units do is similar on the computational level.

 

[Nemo80]

:That is why the whole theory of USA works, because the work the shader units do is similar on the computational level.

Actually, we don't know if it works.

btw, well 1337 is just my way of say "normal" shader... :smile:

 

[dizzyd]

I already said that too. I said the Killzone PS3 renders. That shit got me excited! The funny thing is that certain people immediatly expected Xbox 360 to meet that level of quality graphically. However, since it was on the PS3 many neglected to realize that other PS3 games were expected to look at least that good. MGS4 isn't exempt no matter if Kojima is the Son of God or not. Since I'm neither a MGS fan or a KZ fan I am able to see things more objectively. I liked Killzone better. Alot better.

Kameo got me excited as well. The game looks sweet.

Weren't you the one that said that the MGS4 trailer was comparable to the Chrome Hounds trailer?

 

[Acert93]

I'm really curious as to what you would consider technically spectacular.

I was going to ask the same thing. I remember people saying D3 was not impressive because the technology had been used in offline rendering. The fact is, regardless of where it has been seen before, we had not seen it in games. Further, it was used in a game to make it look GOOD. Just like the first game to use HDR well will make us go "WHOA!". HDR is sooo tacked on in FarCry, SC, HL2, etc. But when we start getting games where it is part of the ART DIRECTION and make a subtle, but significant impact, THEN we can appreciate how "Advanced" it is.

Its not the lense flare technolgies that make games advanced; it is the proper use of these. In that regards MGS4 look rock solid. The effects and technology were all used in a Symphony of Balance to produce a stellar effect. I don't care if another game has all the check boxes on features. If it does not use them well then meh!

With the state of gaming/realtime engines I think the obvious answer to cutting edge, spectacular, etc... is any software that combines technologies into a package that conveys a message. On the most basic level adding a new technique, like HDR or Normal Mapping, is not "revolutionary". But we are at the stage of evolutionary steps. But the key, IMO, is when someone takes a broad array of these technologies and techniques and then puts them forward in a package that makes you go, "Whoa!"

So from my perspective I am not sure anyone could not call MGS4 spectaluar. If it is not I am not sure ANY game we have seen yet could be called such.

Alpha: Could you name 5 games that WOULD meet your criteria? That would give us a better understanding of what games in the past made you go "Whoa".

 

[ihamoitc2005]

But from a technical standpoint, there was nothing in there that impressed me.

Did you see the real-time FPS section at the beginning of the video? Much more detail than Gears of War no? Street littered, clouds of dust, crumbling buildings, light, shadows and details on soldiers, very nice soldier animation, no obvious sprite effects or other shortcuts. That section more than rest of video is probably most indicative of actual gameplay graphics. Rest of video is just for story-telling, like real-time cut-scene..

 

[Acert93]

Actually, we don't know if it works.

Of course they work. I have seen Ruby, Gears of War, PGR3, Kameo, and so forth running on the hardware. So of course they work.

If the theory of USA did not work we would not see hardware capable of using symmetric shader units to carry out both vertex and pixel shading. So I repeat:

As noted in the past, the reason PS and VS can be unified is because PS already are capable of performing much of the work a VS does. The issue is not dumbing down the PS units, but making the VS units more capable.

This has obviously progressed past theory as there are games in development using hardware that does just that.

Not to mention PVR and NV have both announced they will introduce USA in future GPUs. So of course it works.

 

[Titanio]

As noted in the past, the reason PS and VS can be unified is because PS already are capable of performing much of the work a VS does.

Some would disagree that we're at a point where it makes total sense to unify the pipe right now - that doesn't preclude it being a better choice in the future, but for now, that they still might be different enough that different, dedicated hardware is a better choice, or at least a valid alternative. Especially in a closed system with decent tools, where issues such as utilisation are not as strong a concern. Three guesses who that is. It's a different perspective, but right now we don't know who's right and who's wrong (if it's a case of being right or wrong at all! Which I doubt it probably is..)

The pixel and vertex shaders in G70, for example, are tuned and refined for their respective workloads to a very very very low level. I can only guess that accomodating each others workloads could invalidate certain assumptions and soforth that they could otherwise make. In other words, there might well be a cost in terms of the level of optimisation possible for a specific task.

VS and PS have been getting more and more similar, but are we at a point now where there is no cost in terms of optimisation and performance to unify them? That's the question.

Adding an extra Mini-ALU to G70 did not make it less effecient than NV40.

Was a mini-ALU added? I thought the difference in the ALUs was the additional functionality in one of the PS's "main" ALUs.

Regardless, another mini-alu or none, all of that functionality is purely aimed at pixel shading.

That is why the whole theory of USA works, because the work the shader units do is similar on the computational level.

Already said, but we don't know how well or otherwise USA, or Xenos's implementation of that, works. I think it's sound to expect it to work at a competent level at the very least, but expecting and knowing are two different things.

edit - scratch the last part, if you were talking about it working fullstop. I think it's a no-brainer it'll work to any degree, the question is how well.

 

[Nemo80]

Of course they work. I have seen Ruby, Gears of War, PGR3, Kameo, and so forth running on the hardware. So of course they work.

If the theory of USA did not work we would not see hardware capable of using symmetric shader units to carry out both vertex and pixel shading. So I repeat:

This has obviously progressed past theory as there are games in development using hardware that does just that.

Not to mention PVR and NV have both announced they will introduce USA in future GPUs. So of course it works.

In future yes, when there are more capable GPUs that's the reason why ATI did not bring this to the PC market yet if it would be so good they certainly would have done it. Anyways, games using the USA are cetainly not in development at all (for 360). Since the whole architecture is completly hidden by DX on the 360, it's just DX9 not more.

Also working was maybe the wrong word, better say performing. I cannot see anything spectacular on the 360 yet which does come out in front of the best looking PC games yet (FC...). Most of it has bad textures (for "next-gen") and quite low polygon models so it doesn't seem to be as good as some describe it here.

 

[Alpha_Spartan]

Weren't you the one that said that the MGS4 trailer was comparable to the Chrome Hounds trailer?

Yeah, I liked the mech models in Chrome Hounds better than the models in the MGS4 trailer. I also loved the smoke effects too.

vs.

This is an example of the "unimpressive textures" that would have gotten slammed if it weren't MGS4 or an Xbox 360 game. Things like these didn't take away from the overall impressiveness of the trailer, but it shows that this game isn't technically superior to Killzone.

 

[scooby_dooby]

Here's Richard Huddy's comments on MEMEXPORT, definately PR but also probably relevant to the discussion:

Next on my list is the hardware support for directly accessing memory from within the shader units. This makes the Xbox graphics chip work in a much more flexible way than has ever been possible before.

Now it’s relatively simple for a games developer to write code to do anything inside the graphics chip that they could do elsewhere. Accessing memory in arbitrary ways sounds like a very esoteric thing to do within a graphics chip, but actually it allows you to do some amazing things which mean that Xbox 360 games will be more like movies than you ever imagined. It’s so powerful that I’d say that this feature alone makes the Xbox 360 technically superior to any other console planned for the next five years.

Comments on the USA:

It’s not uncommon for one part of the chip to be starved of work for a large majority of the time – and that’s obviously inefficient. With a unified architecture we have hardware that automatically moulds its-self to the task required and simply does whatever needs to be done.

That all means that the Xbox 360 runs at 100% efficiency all the time, whereas previous hardware usually runs at somewhere between 50% and 70% efficiency. And that should means that clock for clock, the Xbox graphics chip is close to twice as efficient as previous graphics hardware.

And in comparison to the RSX:

What we know about the PlayStation 3 architecture is a little hazy – so some of the data is really just best guesses. We do know they are targeting a 550MHz graphics core and we know that it will be based on NVIDIA’s forthcoming G70 graphics chip.

That’s bad news for the PlayStation 3. Current informed opinion is that the G70 will have at most 24 pixel shader pipes and much of the time they will run at low efficiency. It’s a PC based design – which means that all the key customisation for a console is missing, and it’s based on some relatively old PC-based graphics architectures from NVIDIA, whereas the Xbox 360 is an all new design based on what console developers have been asking for.

Because of all this I think that in every important way the Xbox 360 graphics chip is better suited to the task, and more powerful than what we expect in PlayStation 3.

http://www.gamestar.de/magazin/specials/hardware/26071/index.html