Memory bandwidth efficiency of G400 compared to Geforce

[mdhome]

I have read several times that the memory subsystems on Matrox cards are fairly efficient. If so, how does the G400 compare with the Geforce. It seems that the geforce cards have more raw power than the memory can handle and at higher resolutions this become's the limiting Factor. What is the limiting factor on the G400? Is it its memory or graphics chip (or both).


regards Michael

[Humus]

Well .. i'm programming a little 3d app in OpenGL and have found that when using large textures like 512x512x32 there's a substancial performance drop at the g400. This is probably because the texture cache is too small to hold the whole texture. Using 16 bit framebuffer, 16 bit textures, 16 bit depthbuffer etc. does not increase framerate too much, which indicates that the memory subsystem isn't the limiting factor. And overclocking the memory doesn't give too much, only when using textures like 512x512x32 i got some extra fps.

However, most nVidia cards have something like 50-80% difference between 16 bit color and 32 bit color ... so i guess they could gain a lot if they tweaked the memory subsystem a little.

[Pace]

I believe NVIDIA cards have had bandwidth limitations which actually held the card back (ie the bottleneck). Matrox cards are finely balanced cards with lovely-ness all over until an SMP NT kernel knocks then over.

Though with the current speed advantage of NVIDIA cards over Matrox then bandwidth limitations are unlikely to show until Matrox bring out a card which has similar specs but for the increased memory bandwidth.

Paul.

------------------
Pace3000 Network:
Arena | P3K | TechSupport | Portal
Join SETI@MURC!