How many polygons?
Hi!
A Elsa Gloria III can handle 31 million polygons per second.
What does ''handle'' really mean?
If I create 100 objects with 10,000 polygons each, I will
have one million polygons. I can do 31 frames per second
in this scenario, right? And these polygons can change
their position from frame to frame, right? So, morphing
can be done with all 100 objects in realtime?
I can''t believe it. Where did I go wrong?
Thomas
1) It depends where the figure came from. The values written on the side of the box or on press releases are almost always what the raw silicon can do in a lab without a PC. The figures are also often for triangles sized between 1 and 3 pixels, without any texturing or lighting, with optimal overdraw, optimal vertex caching etc...
2) What you''re after is a real-world figure. Download BenMark from the nVidia site, this is a program which will do an almost real-world test of how many polygons you can achieve. Its still not the same as a real game because it submits long single textured strips perfectly optimal for their vertex cache scheme.
3) For true real world performance, one big limiter is how much data (vertices, textures) you can get from the PC to the graphics card - here you''re limited by the speed of the AGP bus, DMA controller, memory speeds etc.
Another big thing is how much parallelism you achieve - you don''t want the graphics chip sat idle while you calculate graphics, and at the same time you don''t want the PC''s CPU idle while the graphics card draws polys. The parallelism is related to the bus limits, and also to how you drive the card (dynamic buffers vs static buffers, locking schemes, buffer sizes, vertex formats etc).
--
Simon O''''Connor
Creative Asylum Ltd
www.creative-asylum.com
2) What you''re after is a real-world figure. Download BenMark from the nVidia site, this is a program which will do an almost real-world test of how many polygons you can achieve. Its still not the same as a real game because it submits long single textured strips perfectly optimal for their vertex cache scheme.
3) For true real world performance, one big limiter is how much data (vertices, textures) you can get from the PC to the graphics card - here you''re limited by the speed of the AGP bus, DMA controller, memory speeds etc.
Another big thing is how much parallelism you achieve - you don''t want the graphics chip sat idle while you calculate graphics, and at the same time you don''t want the PC''s CPU idle while the graphics card draws polys. The parallelism is related to the bus limits, and also to how you drive the card (dynamic buffers vs static buffers, locking schemes, buffer sizes, vertex formats etc).
--
Simon O''''Connor
Creative Asylum Ltd
www.creative-asylum.com
Simon O'Connor | Technical Director (Newcastle) Lockwood Publishing | LinkedIn | Personal site
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
Recommended Tutorials
Advertisement