Xenos Questions

Perhaps you can translate these into sensible form, Dave. If not, then forgive my ignorance.

As I understand it, the ROPs are integrated with the EDRAM module. Does this in any way impede their access to main memory?

I'm also curious to know exactly how much tiling can occur (720p w/4xAA and FX8 buffers will tile less often than the same with FP16 buffers) before it becomes less efficient than just having kept the buffer in main RAM?

Finally, if tiling does impact speed noticably, will games come with an option for either 480p w/4xAA OR 720p w/o AA (in the sense that both will fit into EDRAM w/o tiling)?

 

For those with regular TV's or EDTV's (480p max), will games be rendered internally at 720p or 480p or is 720p really downscaled?

(I haven't seen this answered, but please PM me instead of posting if it has been answered in another thread. This may be an idea something others will want to do to cut down on the size of this thread).


Don't think I'll get an answer for the following...

Backwards compatibility: everything and anything they are willing to say or comment about it.


PS/VS auto-balancing...stuff: The programmer does not have to consider it because it's automatic, but is it possible to make it not do that automatically, say in order to possibly to be a bit more efficient if the designer knows exactly what he wants by some chance? Can I look at it as being X amount of "pipes" assigned to do VS and 48-X doing PS?


Anything to help a particular shadowing algorithm? I read there was something that could assist in stencil shadows (I could be wrong), but what about shadow buffers?


Edit:

Programmable AA ala R3xx+?

 

Find out how many quads the ALU pipes work-on simultaneously.

 

I would like to know what IP from Microsoft’s portfolio of IP ended up being in the R500. Also is the Arbiter programable?

 

does the xbox output to dvi or hdmi so i can use a pc monitor?

 

No.

 

Dave -

Ask them if this is true

"Given the Xbox 360 GPU's multithreading and balanced design, you really can't compare the two systems in terms of shading operations per clock. However, the Xbox 360's GPU can do 48 ALU operations (each can do a vector4 and scalar op per clock), 16 texture fetches, 32 control flow operations, and 16 programmable vertex fetch operations with tessellation per clock for a total of 48*2 + 16 + 32 + 16 = 160 operations per cycle or 160 * 500 = 80 GOps per second."

80 Gops?

 

so other then AA, Do you think they will be using any AF?

ATI has always had great AF, with very little performance hit (unless you crank it all the way up)

 

1. Is the shader language for the Xenos unified?

2. If the shader language has the same syntax and featureset for pixel & vertex shading, can Xenos do texturing inside the vertex shader (e.g. for general-purpose displacement mapping)?

3. Does the Xenos have a programmable topology processor?

4. Since the Xbox 360 has a HDD, are there any "HyperMemory" features designed into the Xenos?

5. Does the Xenos shader model include a complete integer instruction set?

6. Like everyone else, I want to know more about the higher order surfaces Also, have there been any tessellator improvements that will help displacement mapping?

7. Does Xenos support direct access to the frame buffer inside the pixel shader? (Since Xenos is a "tile based render" of sorts I am curious about this as TBDR do not have this issue... could this be the answer to some of our questions relating to fixed function blenders?)

8. On those, what FP blending does Xenos support?

My last question may be outside the realm of Xenos, but is related to the unified shaders:

In the B3D "Direct Next Preview" it was said: "Unifying the shader models is great – there are several instances where you want to do things at the vertex level that’s mostly a pixel level operation, or something at the pixel level that’s usually done at the vertex level". What are some examples of this that game developers may use?

I am also curious about the virtual memory/General I/O Model features MS have planned for WGF 2.0, what (if any) of these features have been incorporated into Xenos?


Thanks Dave for letting us forward some questions, even if they are irrelevant and do not get asked, it is the thought that counts!

 

First there were stories that the Xbox2 isn't backwards compatiable. Now there's a story that it is backwards compatiable. How backwards compatiable is the Xbox2 against the original Xbox? Will people who bought any Xbox game be able to play it on the Xbox2 or will only some games be supported?

 

I think it's small damage control.Theu don't know really how to do that.

 

Well if you don't know . .then I think it's a fair question for Dave to ask.

US

 

Can you run your own physics algorithms on the GPU?

 

about R500:

1 does it still works on quads or 8s, 16s, or what?
2 how many texture units can one batch of pixels (see above) use?
2b is it able to use more texture units to do faster aniso?

 

What's the cost of a failed branch prediction, in cycles?

I'm curious to know if Xenos really is SIMD across the entire shader engine. If so branching is going to be pretty costly isn't it?

Jawed

 

...as an addendum to this, was it actual hardware/software differences and complexities between the platforms that makes it necessary to emulate on a one-by-one basis, or might it stem more from working around proprietary nVidia procedures that MS would otherwise be infringing upon?

 

- Are there any new texture compression formats?
- Does the chip support multiple rendertargets (MRT)?
- Do the 16 "memory access units" that go along with the TMUs have access to all the memory (except eDRAM), or are they limited to accessing a small portion? Do they support memory writes?

 

Can its units work on both vertex and pixel streams, simultaneously, within the same clock cycle?

 

Is it pretty agreed upon that FP10 is used as a default for HDR?

What HDR formats are supported and what's the performance hit for using HDR FP16 if it is supported?

 

Since ATI took over a sizable portion of the former Real3D division of Lockheed Martin (the ATI Orlando design center as one example) what did the former Lockheed Martin Real3D engineers contribute to Xenos, how much former Real3D IP is in there, what are the specific former Real3D technologies, etc ??