• 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.

Ue video memory as main ram

HFS+

Senior member
is it possible to use video memory as ram? I got 6gb of ram but my gpu has 1gb ram how can i use that 1gb of video ram as system memory so that i have 7gb ram?
 
Certainly not by default. As far as I know it isn't yet possible but AMD are talking about it for their new cards, but not quite for the purpose you are looking at.

That card is using some of that RAM to achieve the nice pretty screen you look at currently, its not all going to waste.
 
Currently not that I'm aware of, however with memory prices why would it matter? And you wouldn't be able to use all of the memory from your primary card anyway since that's being used for your display.
 
A very special driver I would have thought 🙂 A graphics driver would naturally assume that it should address all available video RAM, so the 'special driver' would have to cut in before the graphics driver and somehow convince the graphics driver that there's less video RAM to address.

When playing around with RAM disks, I used to use boot.ini to set the maximum amount of physical system memory that the system could see/use, then a RAM disk driver (I used RAMDiskNT) would grab the remaining RAM and set that up with its own drive letter.

I wonder what sort of latency the 'video RAM masquerading as system RAM' would have.
 
Last edited:
it is possible to use it with elaborate opencl/cuda programing, but don't expect anything like this to be implemented into your OS kernel. The main problem would be the PCIE bus latencies/bandwidth.
The memory is better thought of as a large cache for the GPU. IE: transfer a large amount of data there from the host system once then do millions/billions of operations on/with it before replacing it, ideally.
Maybe eventually with APUs, SOCs, and the whole fusion thinking things will be more shared I dunno.
 
Last edited:
A very special driver I would have thought 🙂 A graphics driver would naturally assume that it should address all available video RAM, so the 'special driver' would have to cut in before the graphics driver and somehow convince the graphics driver that there's less video RAM to address.

When playing around with RAM disks, I used to use boot.ini to set the maximum amount of physical system memory that the system could see/use, then a RAM disk driver (I used RAMDiskNT) would grab the remaining RAM and set that up with its own drive letter.

I wonder what sort of latency the 'video RAM masquerading as system RAM' would have.

With Linux there's 2 ram drive drivers to choose from in the kernel, ramfs is a pure ramdisk and tmpfs is a swap backed ramdisk. The latter is preferable since if you run low on memory it can be paged out to the swap partition/file. The main thing ramfs is used for is the early boot process so the initramfs archive can be extracted, udev started and modules loaded so that the root filesystem can be found and mounted.

Video ram would have many orders of magnitude higher latency than main memory, but would still be a lot faster than an SSD.The problem would be that you would end up gimping your display unless the video drivers were written to coordinate with the memory stealing driver so that higher resolutions, games, etc could be run.
 
Back
Top