X-Box 2 Speculation!

[fbg1]

It should have enough power, however I could see compatability problems rising up and such.

That's my whole point. PS2 only has backwards compatibility with PSOne b/c PS2 includes the IOP - PSOne's processor - onboard the PS2 mobo. When you stick a game into PS2, it recognizes whether it is a PS2 game or a PSOne game, and uses the appropriate hardware to run it. It's not b/c the EE is inherently backwards compatible with the IOP, it isn't at all.

Unless all Xbox games are coded completely in DX + some CPU API that can be transferred to the Xbox2's chipset, then the only way Xbox 2 can be backwards compatible with Xbox is if Xbox2 includes the Xbox's chipset onboard the Xbox2 mobo. I question whether MS will be able to put XCPU and XGPU (and MCP?) onboard the Xbox 2 mobo, simply for cost and space reasons. And I seriously doubt Nvidia and Intel will allow those chips to be combined into one die, as Sony did recently with PS2's EE and GS.

At least, that's what I think so far. If the Xbox 2 backwards compatibility situation is not as dire as I think, please enlighten me.

 

[Panajev2001a]

General purpose processors aren't designed to do everything under the sun in the quickest and most efficient way possible

No, they are designed to run the common case fast

Seriously, the more programmability you put in a chip the more you loose some of the benefits that you see with "dedicated HW"...

PlayStation 3 will have dedicated HW for texture filtering and such which are things that are extremely more efficient when hardcoded instead of using a normal CPU and doing it in software...

As far as T&L is concerned I do not see how a projected NV40 ( or the current NV30 ) is more "dedicated" than the Broadband Engine chip we saw in the patent...

VUs on PlayStation 2 show you that with a not high transistors' count you can still design two efficient Vector Units that rival ( well 102.85 MVertices/s vs 112.5 MVertices/s... I would say it is quite close, especially counting the Xbox came out 1 year after the PlayStation 2...

Vertex Shaders and Pixel Shaders are more and more programmable and they keep the speed thanks to the HUGE parallelism found in 3D graphics programming rather than simply being MORE dedicated...

The fact modern GPUs are DEDICATED chips allows them to justify such high Transistors' counts spent in laying out such a massive parallel array of execution units as we see today ( NV30 has 100+ MTransistors... ).

Vertex Shaders are startign to merge with Pixel Shaders and the difference in functions each can do are starting to disappear... you can already use textures as inputs for the Vertex Shaders, you can branch ( conditional branching for both VS and PS is coming )... it is starting not to make sense to keep them as separate units... and that is why for DX10 we are starting to talk about using the same unit for doing both... basically we want a unit flexible enough to do both...

On Cell you could see such an example.. it is the APU.

Something has to give, I know... Cell is better geared for parallel computation and Vector Processing than fast single threaded scalar processing... yet on the latter it won't be exactly a snail either...

 

[Saem]

I think when compared to a general purpose CPU I think Cell with be dramatically more powerful.

However, I believe when Intel brings out Go-Han, he'll kick Cells ass.

I'm sorry, I couldn't resist.

In anycase, yes Cell will rip most CPUs a new one. Now, the thing is that once it starts taking on graphics load, it'll have to compete against CPUs and GPU combos. THis is where Cell will start looking less spectacular. Not to say it'll suck. Just it'll have competition and won't blow everything out of the water or even close.

 

[Panajev2001a]

Saem... but Cell is only an architecture... its implementation can vary sorry could not resist ( being mr smarty pants )...

The Broadband Engine is the 4 PEs, 32 APUs ( 8 per PE ) big beast CPU...

Then you have the GPU part, the Visualizer with specific silicon for things such as texture filtering ( etc... ), 4 PEs, 16 APUs ( 4 per PE ), 4 Pixel Engines ( each comes with also some Image Cache )...

The APUs on the Visualizer should be used as Pixel Shader units, but given the nature of Cell they can be used for basically any task you want them to ( efficiency will vary as will the task )...

I do not expect PlayStation 3 to go with full software rendering...

Anyways, going back to the discussion... I do not see Playstation 2 having problems with software T&L ( well the VUs are quite flexible VLIW processors )...

 

[glappkaeft]

It should have enough power, however I could see compatability problems rising up and such.

That's my whole point. PS2 only has backwards compatibility with PSOne b/c PS2 includes the IOP - PSOne's processor - onboard the PS2 mobo. When you stick a game into PS2, it recognizes whether it is a PS2 game or a PSOne game, and uses the appropriate hardware to run it. It's not b/c the EE is inherently backwards compatible with the IOP, it isn't at all.

Unless all Xbox games are coded completely in DX + some CPU API that can be transferred to the Xbox2's chipset, then the only way Xbox 2 can be backwards compatible with Xbox is if Xbox2 includes the Xbox's chipset onboard the Xbox2 mobo. I question whether MS will be able to put XCPU and XGPU (and MCP?) onboard the Xbox 2 mobo, simply for cost and space reasons. And I seriously doubt Nvidia and Intel will allow those chips to be combined into one die, as Sony did recently with PS2's EE and GS.

At least, that's what I think so far. If the Xbox 2 backwards compatibility situation is not as dire as I think, please enlighten me.

It probably depends on how close to the metal the XBox API:s allow the programmer to go. I don't think the API:s expose to much so assuming that XBox2 is based on a X86 compatable processor and that the DirectX and Win32 API:s of XBox2 are backwards compatable there should be much like upgrading a regualar PC. The big question is probably if MS bothers to make DirectX backwards compatable since I don't see MS going custom hardware or switching to MacOS X (or something).

Patrik

 

[glappkaeft]

I have to strongly dissagre. There is basically no skill-transfer between software and CPU design.

Are you referring to the logic part, the manufacturing part, or both?

The logic, simulation, chip layour, architecture, manufacturing you name it. A compiler coder (at most one of the DEC enginners) could have some meaningsfull input in the architecture stage but I can't see OS coders, IT manegers and database engineers having much of an impact.

 

[V3]

I also don't see how it's an extension of today's trends.

Its like R300 doesn't have dedicated fixed DX7 style T&L. Its emulated through software/shader in its VS unit.

NV30 supposedly have a collection of FPUs bunched together for its VS as well.

The Broadband Engine as we know it right now, will only have 4 general purpose processors. The other 32 processors, are not something you would call general purpose processor.

 

[Panajev2001a]

thanks V3, that was nice and tight as far as explanations go

 

[Lazy8s]

Microsoft could have backwards compatibility without it having to necessarily work perfectly with every single game. Even on PS2, there's a sizeable list of PSone games that won't run properly or at all.

 

[Panajev2001a]

sizable ?

 

[Vince]

Vince, you're not understanding the use of the word dedicated. You're also taking the word dedicated to mean static hardware.

No my friend, your no seing the diffrence between an Athlon or Pentium 4 and a more scientific CPU built to specific needs.

For example, compare an Athlon against the NEC built Vector processors in the Earth Simulator - which is the fastest Super Computer presently. The Earth Simulator absolutly destroys the ASCI series (Run by the DoE and that's based on Intel processors) in not only absolute computation power, but it's being done using a much smaller number of processing elements and thanks to the more dedicated architecture (eg. akin to Cell/VU) it's not even comparable to Intel or AMD's processors.

I don't know how to make this any more clear - The Design of 'Cell' or the EE's VUs have more in common with the NV3x than they do an Athlon or Pentium4.

Your the one whose missing the big picture and instead want to evoke this "Miracle" type language in an attempt to discredit it, which is just so far off base. But, enough other people have seemed to explained it.

This is opposed to a general purpose processor which can do one thing, then it can do this thing, then it can do that thing, then it can mimmick this thing, ad nauseum.... Dedicated processors can be designed and trimmed to to do exactly what they're supposed to do in the quickest and most efficent way possible, which includes any programmability they may have.

I'm going to go out on a limb and guess this isn't a quote attributed to John von Neumann.

 

[Panajev2001a]

What still leaves me perplexed is:

Dedicated processors can be designed and trimmed to to do exactly what they're supposed to do in the quickest and most efficent way possible, which includes any programmability they may have.

:?

 

[Saem]

Saem... but Cell is only an architecture... its implementation can vary sorry could not resist ( being mr smarty pants )...

Holy crap, i can't believe I slipped like that. You're so right. *shakes his head*

 

[Panajev2001a]

Is that sarcasm that I see Saem ?



sorry I was not trying to boast anything ( it is public information after-all )...

 

[chaphack]

So PS3 might have a powerful CPU, but so did PS2. But that did not make it to be a miracle machine back then.

I think Mr Nvidia said it best, PS3 will be an extension of PS2, pure vector processing. We shall see how cool that will turn out. *looks back at PS2 hype/experienc* ..shhhhivers...

 

[Saem]

Is that sarcasm that I see Saem ?

I thought you might pick up sarcasm, but there was/is none present.

 

[fbg1]

The logic, simulation, chip layour, architecture, manufacturing you name it. A compiler coder (at most one of the DEC enginners) could have some meaningsfull input in the architecture stage but I can't see OS coders, IT manegers and database engineers having much of an impact.

Alright, point taken. Thanks for the info.

 

[Teasy]

Well teasy, if i remember correctly, PVR was one of the top seller of 3D chip when sega released the dreamcast. They were effectively producing chips and selling a lot. That's no more the case.

No they weren't. They had no top performing chip in the PC space when Sega chose them for DC. AFAIR they hadn't produced a 3d chip in the PC space for years before DC as well.

That's not true. When MS (or any other company) need something peculiar, then MS puts it in their specifications. It's up to the chip maker to fullfill the specifications. MS know what they want, that's why we don't have a NV20/25 in the X-Box but a modified chip. The extra-developpement is far less than making a chip from nothing.

You've mis-understood me. I didn't say MS have to buy only already completed PC chips from ATI or Nvidia. I said once ATI or Nvidia have produced the chip MS want MS will then have to buy those finished chips from ATI or Nvidia. As in ATI/Nvidia will produce the chips and sell the finished chips to MS. Which means that MS has no control over the chips. With PowerVR MS would have full control. They would produce the chips themselves and just pay PowerVR royalty fee's. So if MS decided they wanted the chip on a new process to save money, they wouldn't have to go to PowerVR and ask them to do it, like they would with Nvidia. They would be able to just do it themselves. See what I'm getting at?

They don't produce any more? Or more precisely, much less.

They hadn't produced a PC chip for years before they did the DC chip. Just like they haven't had a new PC chip for almost 2 years now. Nothing is different. Once again this is meaningless. All that matters is what tech they have now and wether MS likes it or not. Just like with Sega and DC.

So what? Does it means that with MS' money they can't do something?

I don't know what Bitboys can do, that's up to MS to decide. What I said is its silly to compare Bitboys to PowerVR/3DLabs. Because your comparing a company that has never sold a chip with companies that have sold millions of chips over the years.

 

[Teasy]

Finally, i'm not saying that 3DLabs or PVR can't do it, i'm saying that it will be more expensive.

Why would PVR/3DLabs be more expensive?

That's the difference between all house made and outsoursing. The contract is what is important in outsourcing, the terms and agreements

Did MS get that agreement right with Nvidia? Is that why they went to court against Nvidia? Common, just admit that MS will see being able to have full control of there chips as an attractive thing. That doesn't mean they will choose PowerVR. It just means PowerVR may be one of the companies they seriously consider for XBox2. Of course in the end its down to wether or not they can provide what MS want that will clinch any deal.

 

[Qroach]

MS and nvidia settled the dispute in court, and some people say that both sides came away as winners. either way, it's not going to have any influence on xbox 2.