• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Any way to share system memory with video memory?

iamtrout

Diamond Member
Just wondering. I have a 128MB 6800NU right now and just figured what the heck, if I can share 256MB of my 1GB with it, it'd be great.
 
Ok, just found out that AGP aperture size in the BIOS shares system memory with video memory. Anyone know how to make it more efficient, like making a 128MB 6800NU the equivalent of a 256MB GT through memory sharing?
 
Actually that is not quite true.

But you need the source code for the driver to do this, the video chip has to be cooperative and the memory you expand is not neccessarily usable for every memory purpose in the card (textures, framebuffer, shader commands).

For example, the DRI folks (Linux OpenSource 3D drivers) are fiddling with using additional RAM for the 4 MB Mach64 chips so that you can do more than 800x600 in 3D (which 4 MB doesn't allow).

But as I said, without source code for the driver you are screwed. And even the Linux folks don't have source for NVidia drivers and for ATI only up to Radeon 9200 (and it's a different driver than ATIs).
 
You wouldn't want to use system RAM for graphics duties, anyhow. My Radeon X800 XT PE oc'ed has nearly 40 Gigabytes per second of bandwidth as compared to 2.5 Gigabytes of bandwidth of my system RAM. It'd be a slide show using system RAM.
 
Er, yes, AGP has been created to do just that. About the only thing you can't do from "borrowed" system RAM is the final 2D surface you're going to display. (Unless it's a shared-RAM integrated VGA anyway.)
 
Permedia2 cards supposedly had some sort of pageable virtual-memory scheme that they could use, for their PCI cards, to DMA texture memory from system RAM into video RAM, something that I think they added from their higher-end workstation cards with similar VM capabilities.
That's the only way besides the current AGP aperture/GART scheme that I'm currently aware of - oh, and ATI's "HyperMemory" for their PCI-E cards, don't know anything about the actual implementation of that yet.

I think that they should add a caching IDE RAID controller to those low-end GF FX5200 and R9200/9550 cards with 256MB of memory on them, at least the additional video-memory wouldn't go to waste. 😛
 
Back
Top