I Want To Compare The Pandora Video Card With A Pc Video Card

[hede]

AFAIK the sgx is a near relative to the good old Kyro PC cards. These Cards are compared many times to its counterparts from nvidia and ati.
Maybe there are even sgx benchmarks. Anyone searched for beagleboard or Palm Pre (or maybe even iPhone 3gs)?

 

[darkblu]

The USSE shaders on SGX has three datatypes. High-precision, which are IEEE-754 compliant 32-bit floats, medium-precision, which are 16-bit floats (1 bit sign, 5 bits exponent, 10 significant bits), and low-precision, which are the 10bit fixed point format I described. In a cycle it can do one high-precision operation, two medium-precision operations, or three/four low-precision operations. What documentation/comments are available do actually say three or four, which is unclear but suggests 40bit registers.

Exophase, are you sure about the 10bit int format? the intel gma500 docs list only:

— 32-Bit IEEE Float
— 2-way, 16-bit fixed point
— 4-way, 8-bit integer
— 32-bit, bit-wise (logical only)

 

[Adventus]

Exophase, are you sure about the 10bit int format? the intel gma500 docs list only:

— 32-Bit IEEE Float
— 2-way, 16-bit fixed point
— 4-way, 8-bit integer
— 32-bit, bit-wise (logical only)

POWERVR SGX OpenGL ES 2.0 Application Development Recommendations:

To achieve best performance for a variety of tasks, the USSE supports multiple precisions. The GLSL
ES precision modifiers, lowp, mediump, and highp, map to 10-bit fixed point, 16-bit float and 32-bit
float types, respectively. While lowp computations are performed as 3 and 4 component vector
operations, mediump operations use 2 component vectors. When using highp precision, the USSE
operates on scalar values.

 

[MrPoletski]

It is practically IMPOSSIBLE to compare with PC hardware.

First of all, SGX is built for mobile devices. Meaning that it focuses on the performance per watt instead of raw performance. The microarchitecture is very much different because of the reason mentioned above. There is no Direct X support(so no Shader Models) as far as I know. Amount of RAM does not mean a thing, because the texture sizes are so much smaller because of the weaker core.

There are few things which I'd like to know about SGX and Pandora: Memory bandwidth, clocks and bus width. GPU core clock, information of the core, what kind of FPU/Vector performance can be expected and how the microarchitecture works. Also some words about the drivers.

Frankly, I haven't studied this at all myself, just popped to this thread. So yeah, going to have a great google-fest now.

SGX is designed to be scaled right from a mobile phone all the way up to high power arcade systems all on the same architecture. It very much supports directx 10+ also. In my opinion, the microarchitecture is awesome.

It has been and always will be possible - given the will to take the risk - to make a Pc graphics AIB that would wipe the floor with everything on the market for price/performance in games.

I'm amazed nobody has ever tried sledgehammering an effort in at the high end with powervr tech.

 

[Exophase]

The TI provided docs say the same thing as darkblu's snippet, so now I'm really wondering which one is right. The iPhone OpenGL ES programmer's guide says 8bits of precision but -2 to 2 range which is not a very good fit for 8bit per channel textures.

Someone ask Xmas or dmdm?