Is the display card ram or the main memory used when a display is connected to the motherboard display connector but rendered by the display card GPU?
I’m concerned mainly with memory bandwidth issues (not space), but generally I’d like to know what is actually happening and what performance consequences might be.
I connected one of two monitors through the display card and one through the motherboard. Unlike the BIOSs in most of the computers I use, this one has a choice to avoid disabling on board graphics when a display card is used.
I’m pretty sure I have verified that the display card GPU is used (at least by default) to render images on both monitors.
But for the second monitor, does it render into system ram and the on board graphics then refreshes from system ram through the motherboard display connector? Or does it somehow refresh from graphics card memory to the monitor?
That may be the same question as which graphics system is actually doing the refresh job.
BTW, on systems where the BIOS never enables the on-board graphics (because a display card is used), is there a way to override that to use those extra connectors for more monitors?
I would believe each uses its respective vram. The GPU card has its own ram and the iGPU uses system ram as normal.
The act of adding an additional PCIe GPU would not change the way the iGPU in the processor functions. Likewise, using only the GPU on the card would not change the way the vram is allocated for each. They seem to be discrete vram allocations.
The question about the bios automatically disabling the intergrated graphics when a GPU is installed in the PCIe slot would depend upon the BIOS. If the on-board GPU is disabled it would imply that the connectors are also disabled and cannot be used. They certainly could not be used as extra connections for the PCIe GPU since that is a hardware design. The on-board GPU supports only the on-board connectors and the PCIe GPU supports only the connectors on that card. The vram is similarly restricted by hardware design.
I doubt that is true unless both monitors are attached directly to the card.
In the past some GPU designs allowed a direct connection between the PCIe GPU cards so one could control the rendering on both (each card could only drive one monitor). That is a thing of the past with newer GPUs having multiple display connections (HDMI & DP) so one card can drive multiple monitors. (My RTX 3050 has HDMI + 3 DP connections and will drive up to 4 monitors directly connected)
I’m still pretty sure that the PCIe GPU was doing the rendering for the performance benchmark rotating image I tested on the monitor connected to the motherboard connector.
Can you think of any convincing test?
When you say “supports” I would hope to understand that separately for rendering vs. refreshing. PCIe allows the card to read/write system memory. Rendering for the onboard connector is just a matter of writing to whatever ram that image is kept in. If that is system ram, there is no hardware reason it can’t work that way.
I don’t know/understand what is involved in actually sending a signal through the connector. I would guess only the onboard graphics has a path to send through that connector. But I don’t really know. Refreshing means reading from the ram where the image is stored and sending that to the connector. Regardless of which ram the image is in, the hardware clearly allows either graphics system to read that ram for refresh. If the PCIe can’t directly transmit to that connector, that would imply the onboard graphics must be doing the refresh. But that still would be possible with either type of ram and with either graphics system doing the rendering into the ram.
I hope someone who understands this better than either of us will add some information.