I have 64GB of ram and a GTX 1060 6gb, and this has been a thing i’ve been wondering and actively searching for but never find a clear answer on. Is there a way to dedicate say for example 4gb of it to make it essentially act like extra VRAM?
Not with a discrete GPU like that, no. An integrated GPU uses system RAM as VRAM, but even there I’m not sure you could directly control how much is used.
Integrated generally has the option to adjust the frame buffer size, usually up to around either 64MB or 256MB.
Dedicated graphics cards, however, have no such option. Even if they did it wouldn’t help - the bandwidth of system memory is significantly lower than that of video memory, and that’s before taking into account that it would have to share time with the CPU itself, and still cross the PCIe interface.
To put some numbers on it, PCIe 4.0 16x is capable of just shy of 16GBps in both directions simultaneously, dual-channel DDR4-3200 is around 50GBps (I’m unsure if this is full duplex), and the memory bandwidth of a GTX 1060 is between 192GBps and 216GBps (again, duplex unknown), depending upon variant.
However, since AGP was a thing, GPUs have been able to store textures in system memory. I don’t think it gets a lot of use any more outside of on-die/-package GPUs (due again to the large disparity between available bandwidths) but it’s technically still possible. It doesn’t perform relevantly better than just caching in RAM (even via an OS-level disc cache) but it would have a slightly lower overhead - not that multi-core systems really even notice.
Ultimately, though, there would be no point. Dedicating system RAM to graphics only would offer zero performance increase while restricting the amount of RAM available for general use. The only possible outcome would be a reduction in performance, assuming any change at all.
Darn it. 