Xenos Questions

[jvd]

Shit. I'm really beginning to understand why there are lots of confusion about how this actually operates. It really doesn't operate in a manner in which fits in with the conepts we've understood before.

Dave I'm really hoping you do an article about this . I'm really interested in learning more

 

[DemoCoder]

The end of "pipelines" as we know them. No wonder the R400 got delayed, this thing is a really ambitious design. The MS IP aspect concerns me tho, clearly ATI should retain a cross-license to use this on future desktop parts.

Of course the patent issue concerns me tho. If MS essentially makes WGF2.0 = R500-like architecture (e.g. most efficiently implemented this way), I have to be concerned about all other IHVs being precluded by restrictive patents from trying similar techniques. They'd essentially be bared by a monopoly adopting an API and controlling the very IP that is must be used to efficiently implement it.

 

[ralexand]

The end of "pipelines" as we know them. No wonder the R400 got delayed, this thing is a really ambitious design. The MS IP aspect concerns me tho, clearly ATI should retain a cross-license to use this on future desktop parts.

Of course the patent issue concerns me tho. If MS essentially makes WGF2.0 = R500-like architecture (e.g. most efficiently implemented this way), I have to be concerned about all other IHVs being precluded by restrictive patents from trying similar techniques. They'd essentially be bared by a monopoly adopting an API and controlling the very IP that is must be used to efficiently implement it.

Excellent points, but I would think they will try to avoid any look of impropiety from here on out. The government is watching.

 

[Jawed]

These would be ATI's patents, not M$'s.

Jawed

 

[nAo]

Yes, it's an ambitious design!
Let's hope it succeedes too, so other IHVs would be forced to accelerate on the innovation too

 

[Danack]

These would be ATI's patents, not M$'s.

I believe there is an 'interesting' licensing agreement between ATI and MS.

Alledgedly in early 2002 MS said to both Nvidia and ATI "We need your patents to beat Sony, sign them over or never work with us again". Nvidia said "No - you'd pwn us if we did that". ATI said "sure thing" to get the XBox2 contract.

All, highly alledgedly.

 

[Dave Baumann]

No, thats was nothing to do with Sony, or consoles. It was more a case that MS didn't want the IHV's forcing a license for everything that went into DX, so for DX9 and beyond they basically said that they'll only give a hardware license if the IP issues are swept aside.

 

[Danack]

so for DX9 and beyond they basically said that they'll only give a hardware license if the IP issues are swept aside.

But doesn't the Xbox360 have DirectX on it? Presumably Microsoft wouldn't have limited the IP license to just a Windows box....though obviously I'm only guessing here.

 

[nelg]

Also is the Arbiter programable?

From anohter thread..

The engineers nearly groaned when I ask for an explaination of the load balancer - in conceptual terms its simple, in real terms it seems fairly complex logic. At the basic level it analyses the sizes of the vertex and pixel buffers and tries to apportion the load so that they are equalised as much as possible (dependant on the program load). It can also recieve hints from the OS and application so the programmer can give a bias to the load balancer, increasing the priority of load types.

Cool.

 

[DemoCoder]

Dave,
Does that mean if NVidia implements unified shading, Microsoft can grant a license to NVidia even if ATI doesn't? In traditional standards working groups these days, people are forced to declare IP up front, and in many cases, given up said IP for uniform licensing.

I'd also be curious as to the separation of patents. Obviously, there can be patents related to the WGF2.0 "concepts" much like NVidia held pixel shading patents, but there can also be patents related to "implementation" of those concepts.

So while MS may own the IP for say, unifed shading, maybe they don't own stuff like "The Arbiter"

 

[blakjedi]

so for DX9 and beyond they basically said that they'll only give a hardware license if the IP issues are swept aside.

But doesn't the Xbox360 have DirectX on it? Presumably Microsoft wouldn't have limited the IP license to just a Windows box....though obviously I'm only guessing here.

No the X360 API and os is supposedly unique... it may be derivavtive of Directx or NT but they arent the same.

 

[Robbitop]

Shit. I'm really beginning to understand why there are lots of confusion about how this actually operates. It really doesn't operate in a manner in which fits in with the conepts we've understood before.

Then it would be nice if you could explain the whole things to us. Maybe in an seperate article with correct blockdiagramms?
I think there will be many people who are interested in reading it.

by the way: too many questions and too less answers in this thread IMO

 

[Shifty Geezer]

so for DX9 and beyond they basically said that they'll only give a hardware license if the IP issues are swept aside.

But doesn't the Xbox360 have DirectX on it? Presumably Microsoft wouldn't have limited the IP license to just a Windows box....though obviously I'm only guessing here.

No the X360 API and os is supposedly unique... it may be derivavtive of Directx or NT but they arent the same.

This can't be right as XNA is intended for cross-platform development. MS even suggested Sony use it in PS3. Whatever API is used for XB360, it'll find it's way to PC if it isn't there already, save a few custom OS pieces I imagine.

Regards unified shaders, this is hardware covered by patents, surely? The DirectX/XNA API calls for vertex and pixel shaders, and the hardware either has these as seperate units or unified, but their implementation shouldn't matter to the API.

 

[London Geezer]

This can't be right as XNA is intended for cross-platform development. MS even suggested Sony use it in PS3. Whatever API is used for XB360, it'll find it's way to PC if it isn't there already, save a few custom OS pieces I imagine.

That makes me think... What are the chances of seeing opengl engines on X360?

 

[version]

 

[version]

http://pc.watch.impress.co.jp/docs/2005/0526/kaigai183.htm

 

[MechanizedDeath]

Sure wish that new block diagram had the bandwidths posted to different parts of the chip. :? Two steps forward...two steps back. PEACE.

 

[nelg]

How would the R500 perform without the eDram and associated logic? Assuming the missing logic would be transferred back the GPU. Is there anything inherent in the design (unified shaders) that would make it perform better or worse than a traditional design, WRT the functions the eDram logic was doing?

 

[Pete]

Re: NP, Lumi.

I'm pretty sure the internal organization is different and that the ALUs are not "outputting 2 samples per clock each". It's quad-based.

Xmas, just so I'm clear, does he jump from shader ALU to ROP functionality and back in this full quote? If so, is that incorrectly mixing terms, or is it part or R500's new architecture?

The 48 ALUs are divided into three SIMD groups of 16. When it reaches the final shader pipe, each of the 16 ALUs has the ability to write out two samples to the 10MB of EDRAM. Thus, the chip is capable of writing out a maximum of 32 samples per clock. At 500MHz, that means a peak fill rate of 16 gigasamples. Each of the ALUs can perform 5 floating-point shader operations. Thus, the peak computational power of the shader units is 240 floating-point shader ops per cycle, or 120 billion shader ops per second at 500MHz.

Or is he just saying there are 16 ROPs, despite the (8 ROPs) rumors to the contrary?

 

[Xmas]

Seems like incorrectly mixing terms to me. It's simply not the task of an ALU to write out samples to memory.