How is "Intel Graphics" split between microprocessor and chipset?

Question

I'm running a Xeon E3-1275 v6 (Kaby Lake) processor - that has Intel HD P360 for a GPU - on a motherboard with C236 chipset - which doesn't support graphics. Naturally I knew that there was no monitor connector on this motherboard (HDMI, DVI, etc.) but I did expect to see an Intel HD P360 GPU in the Windows (Server 2016) device manager and be able to use the built-in GPU for GPGPU type applications.

But no, it's missing from the device manager.

So my main question is: how is the implementation of Intel graphics/gpus split between the microprocessor and the chipset such that you need both in order to get any graphics?

Alternatively, maybe I did something wrong or forgot to configure something I needed to and I should see the Intel GPU available - if that's the case, what do I need to do to fix it?

Edit: This is definitely related to Can a processor gpu still be used for video transcoding if the motherboard is disabled from using processor gpu? but that doesn't have an explanation (or answer!).

Answer 1

According to a block diagram from Intel featured in this article: https://hothardware.com/news/intel-skylake-gen9-series-graphics-architecture-unveiled:

This diagram is included in this technical paper that details the graphics-processor in the same processor, though it does not go into detail about the Display Controller component.

• As you may already know, a "GPU" is only part of a graphics system: you also need graphics memory (to serve as a framebuffer and as texture-memory or RAM for GP-GPU computation), you also need a bus to get to/from system main memory (if it isn't sharing system RAM), and finally, a display-controller which is where the electronics are that read the framebuffer out to the various ports (DVI, HDMI, and RAMDACs for the analog VGA, S-Video, Component and Composite ports, if applicable).

• In the diagram from Intel, we see that the graphics-processor part of the GPU system resides inside the processor-package. This is illustrated by the light blue block on the far-left being contained within the large grey rectangle representing the processor-package. The graphics-processor connects to the Display Controller through the Ring Interconnect built-in to the package. (The memory bus to the EDRAM (dedicated GPU memory) or shared system memory is not in these diagrams).

• The question then, is how do display signals from the Display Controller get out? According to this semi-technical paper from Intel ( https://www.intel.com/content/dam/www/public/us/en/documents/white-papers/ia-introduction-basics-paper.pdf ) it describes how the processor itself generates its own display connections for eDP, DisplayPort 1.2, DVI and HDMI (page 3). Note these are all digital signal formats.

• That same document describes how the chipset (a H81, Q87 or C226 in this case) provides a "Platform Controller Hub" (PCH) that provides additional display output options, such as VGA, but also that it uses Intel Direct Media Interface (DMI) to access the Display Controller directly - so the VGA signal is generated from the framebuffer's internal representation in the Display Controller instead of being transcoded via DisplayPort or HDMI first.

• And then the PCH will have the necessary conductors routed directly to the VGA port on your motherboard. The other ports (HDMI, DisplayPort and DVI) will apparently be wired-up directly to pins on the processor package itself (perhaps via a signal-amplifier or relay circuit?).

As for your specific question: it's likely you simply have the processor built-in GPU disabled in your EFI/BIOS setup area. You said that your motherboard already has an ASPEED GPU built-in, so I'll bet your motherboard is configured to disable the CPU-integrated graphics and to substitute ASPEED instead (this may be preferable for many reasons). If you disable the ASPEED graphics then the system should default back to the Intel CPU-integrated graphics.

For further reading, search for "Intel Graphics Technology Interface".