3GSM Khronos Presentations...

Kristof: oh dear... I accidentally checked the whole presentation. ;)

anyways, interesting as a whole indeed. one thing that I spotted was amount of finnish companies listed. Nokia, FatHammer, Bitboys, Hybrid Graphics... ("mistä tämä hullu pohjoisessa elävä pieni kansa saa voimansa?" as english translation:"Where this crazy small nation living up north gets the all motivation?")

MBX looks good, but I really wasn't expecting anything less. ;) ATI didn't had a part at least on slides and nVIDIA was able to only show stuff on paper. (and at least those screenshots didn't convince me that it would really have programmable vertex or pixel processing.) Mitsubishi's Z3D-III was mentioned as low budget solution... Acceleon G30 performance/power consumption is still a mystery. (at least they are quiet. this is what ppl have been wanting right? G10 ended up on NEC's LCD pretty much without warnings forehand, so only time shows if there is interest around G30 core.)

Hybrid Graphics is a nice firm... I have few friends / I know few guys working there. (oh... if anyone reads this from Hybrid, the Gerbera mascot is excelent! :) Never destroy it. :) )
 
Wow nice catch, so they finally put in multisampling huh ...see Simon, that is what you get for taking a vacation! :)
 
MfA said:
Wow nice catch, so they finally put in multisampling huh ...see Simon, that is what you get for taking a vacation! :)

I wonder if he actually likes multisampling... :oops:
 
MfA said:
No, that is why I said it. Simon is lets say, ambivalent towards multisampling.

I hope he didn't develop the algorithm with closed eyes and wearing a gas-mask LOL :LOL:
 
Slide 46 is interesting as well - Renesas are to launch an SH-Mobile chip with MBX Lite in Q2.

So that's Intel (Carbonado), TI (OMAP 2), and Renesas all launching MBX chips this year! :D
 
presentation about MBX VGP said:
Compatible with Microsoft DX8 vertex shader 1.0 definition

:?:

Can I find somewhere the specification?

OT: Guys, don't you know that a tile based renderer can't have a geometry engine? :D
 
vb said:
OT: Guys, don't you know that a tile based renderer can't have a geometry engine? :D
Gosh, I presume that precludes programmable ones as well. Guess I'd better delete these programs then :)

The NVidia slides didn't suggest very much in the way of programmability - I'm even more curious by what they meant in their press announcement.

I was also curious as to what their "high quality texture compression" is. S3TC?

mfa said:
Wow nice catch, so they finally put in multisampling huh ...see Simon, that is what you get for taking a vacation! :)
Enough of your cheek! :)
 
vb said:
presentation about MBX VGP said:
Compatible with Microsoft DX8 vertex shader 1.0 definition

:?:

Can I find somewhere the specification?

OT: Guys, don't you know that a tile based renderer can't have a geometry engine? :D

From an older ARM .pdf:

Vertex Geometry Processor (VGP) Features

  • Hardware transformation and lighting
    engine
  • Fast floating point processor optimized only
    for 3D geometry applications
  • 4-way SIMD - 4 FLOPs per cycle
    * Calculations performed to IEEE
    single precision floating point
    * 4x4 matrix multiply in 4 cycles
  • Based on Microsoft DX8 programmable
    VertexShader spec
    * 16 vertex inputs, 192 constants,
    14 temp regs, 16 vertex outputs
    * 256 instruction memory -
    downloaded instruction store
  • Reciprocal, reciprocal square root,
    fractional, log and power functions
  • Pipelined operation allowing most
    instructions to execute in 1 clock cycle
  • 4M FLOPS / MHz - 320M FLOPS @
    80MHz peak performance

Simon,

Gosh, I presume that precludes programmable ones as well. Guess I'd better delete these programs then

Second call for another confession 8)
 
Simon F said:
I'm a bit confused (Ok, so nothing new there). I think that should have read VS1.1

I think VS1.0 was suposed to be what original Radeon(R100)/Rampage suported and was later scraped from specification.
 
Back
Top