Power VR 5

Something is happening at PowerVR. I'm not sure what it is but something's going on. I know a few of the investors have suddenly taken a bit more of an interest this last week.

Perhaps they're gearing up to some sort of announcement. Details on the chip for Sega perhaps?
 
hey guys cool down, I know the summer is warm, but that's no excuse :p
Let's wait and see, as always.
 
*pours bucket of cold water on head*

bah, I'm not exactly happier now. oh well, i guess its better than getting all worked up just to be let down bigtime.
 
What sort of a beast (assuming high end pc/arcade) is assumed here? Can they take the crown soon or at all?

An 8x/12x/16x pipe setup (2/3/4 quad) with possible dual chip usage in Sega-Sammy's low-high end arcades? Maybe they could run in pairs as apposed to quads and keep efficiency due to the deferred nature???

What bandwidth would be required? Would a 256-bit memory interface be necessary? Radeon 9700/9800 level perhaps??
2 pipes = 128bit sdram (kyro 1/2)
4 pipes = 128bit ddr (kyro 3?)
8/12/16 pipe = 256bit ddr1/2/3 (pvr5?)

And what will happen with the pixel & vertex shading (shaders v3+). Will this be an area which has more advancements then other architectures?

Could we finally see a PPP or elements included here... could this be the issue with dx9c delays or is this more likely in dxNext??
 
The most important info about PowerVR Serie 5 is: Could it render Mercedes Benz Truck Racing benchmark?
And if we still could see the "infamous" tire track "that should clearly be behind the truck"-march 2001.
And will PVR5 will support hardware transform and lighting, because it's important in games such as Diablo 2.

:LOL:

BTW it could be funny if IMGtech decide to answer to Nvidia's slides with a joke à la AMD (with the "Prescott survival kit").
 
I dont think T&L will be an issue this time, arcade has had it for eons (from series 2 - ELAN unit) and even the tiny MBX Pro has a (dx8 level- v1.1) vertex shader unit.

Remember this is an architecture with a potential for Arcade(SEGA)/PC/Console so its obviously has to be significantly more powerful and advanced than at least the current Arcades and Consoles on the market else noone would use it.

On the PC Front, the shading power will be more of an issue this generation with Half Life 2, and other games based on it and other dx9 engines. The stencilling speed of the previous generations from PowerVR-namely Kyro, if continued through accordingly, should provide some impressive performance in the Doom3 engine by possibly offering increased Z-performance of recent architectures.

These games may be what is holding back the release, as well as SEGAs own arcade launch first, or is a PC Partner still needed, and this going to be Arcade-only?

I think many of the compatibility issues arise from developers not strictly sticking to the Open Standard APIs and carrying out techniques which are not recommended, but may work well on one particular common testing architecture/extension (shadermark tests anybody??).

Early drivers, I believe were responsible with Mercedes Benz, but Im pretty sure that a little testing would resolve these as with any new architecture. Ive read reports of the new NV40 failing to run/crash running some programs... its always going to be an issue early on in product's life cycle, and Im sure all the companies out there have buggy drivers, which crop up in one app or another.
 
aZZa said:
Early drivers, I believe were responsible with Mercedes Benz,
IIRC, (AND I COULD BE WRONG), the application 'forgot' to tell D3D that it wanted to read the Z buffer. I think the driver update simply added an app hint in to correct that oversight.

Amusingly, I seem to recall another vendors system not working with MBTR...
 
Simon, you should keep quiet, else the 3D-Police from the Village :oops: will be after u guys for cheating Kyro drivers through app-detection (not compatibility profiles) :D
 
Simon F said:
Amusingly, I seem to recall another vendors system not working with MBTR...

Too add to that it was also ironic how when that happened the web site that had been used so heavily in a certian PDF pulled the game from its benchmark suite due to rendering issues with hardware :) Sorry for the OT post.
 
aZZa said:
An 8x/12x/16x pipe setup (2/3/4 quad) with possible dual chip usage in Sega-Sammy's low-high end arcades? Maybe they could run in pairs as apposed to quads and keep efficiency due to the deferred nature???

I still don't see any need YET for multichip sollutions for Arcade machines, but that's just probably me.

What bandwidth would be required? Would a 256-bit memory interface be necessary? Radeon 9700/9800 level perhaps??
2 pipes = 128bit sdram (kyro 1/2)
4 pipes = 128bit ddr (kyro 3?)
8/12/16 pipe = 256bit ddr1/2/3 (pvr5?)

I guess you were able to still find pipelines in the usual sense up to Series4.

And what will happen with the pixel & vertex shading (shaders v3+). Will this be an area which has more advancements then other architectures?

Could we finally see a PPP or elements included here... could this be the issue with dx9c delays or is this more likely in dxNext??

*lifts shoulders*

As for future APIs there's a Geometry Shader and an (optional) Tesselation unit mentioned in the WGF specs; I doubt anything regarding topology or tesselation in that sense has all that much to do with former APIs (at least not in a similar sense).
 
Back
Top