SGX Complies With DirectX 10

Lazy8s

Veteran
PowerVR Siggraph 2007 presentations list support for the DirectX 10.1 API in SGX.

(slide 6)
http://www.khronos.org/developers/library/2007_siggraph_bof_opengles/OpenGL ES BOF Demos Aug97.pdf

Previously, the company only went as far as describing that its functionality "exceeds... Microsoft Vertex and Pixel Shader Model 3 requirements".

The list of SGX licensees mentions the existence of multiple, unannounced customers. Prior press releases account for only one unidentified company, so SGX appears to be adding more partner wins.

Performance at 100-MHz for PowerVR's new laptop core, the SGX545, is claimed to be 31-Mtri/sec and 1-Gpix/sec. Interestingly, the SGX510's pixel fillrate is quoted as starting lower than the MBX Lite's.
 
So how likely do you think it is that we will see a PowerVR-based graphics card (or IGP) for PCs? And from who in that case? I really would like to see it, but I can't see it happen.
 
31-Mtri/sec would be a third of a 7600gt according to the benchmark prog that comes with f1 99-02

Take a current laptop IGP with the highest potential you can find and try a geometry benchmark with a shitload of lights. After you've marked the results, compare the on paper claimed triangle throughput and see the differences. Those 31M Tris/s should be perfectly achievable, since PowerVR has a long lasting tradition with claimed numbers.
 
is the point of your post to say powervr arnt lying ?
I wasnt trying to suggest they were

There's usually a large difference between claimed on paper specs and real time throughput; if they haven't changed policy they're amongst the rare exceptions yes. Considering that they've announced the first chips to be at 200MHz@90nm that 31M figure isn't even relevant. That said how many "real" triangles do you think a 7300 based IGP clocked at 200MHz would sport?
 
Performance at 100-MHz for PowerVR's new laptop core, the SGX545, is claimed to be 31-Mtri/sec and 1-Gpix/sec.
i achieved 21million tri/sec (rendered) on my old gf2mx 200, so something is definitely strange with that number
 
i achieved 21million tri/sec (rendered) on my old gf2mx 200, so something is definitely strange with that number

Ignoring that there are worlds of differences between an ancient T&L (fixed function) unit and ALUs that can handle multiple shader tasks, did it strike you that yer ole gf2mx had only one geometry unit, while that thing in question might have more than one (unified) ALUs?
 
Or it could mean mega-triangles! Let's see your GF2 do that!

Or...not. Weren't there a few other cards PVR were supposed to be releasing a while back? Or did they get canned?
 
Back
Top