Hello, That's an interesting question: Even if the GPU is physically-located inside the die, it is exposed as a "virtual" PCI device (vendor number 1002 and model number 130f), and that's how we detect it, and that's how the driver configures it. Many components of the CPU die are configured through PCI just because it's an easy way.
Ideally, the AMD driver would put some flag somewhere in the sysfs attributes to tell us that the GPU is actually inside the package but I am not aware of any such way to get that information. So that's something we may want to improve. That said, it's not such a big problem since the locality is still correct: the PCI device appears connected to the main machine memory. What could wrong is the L3 cache, but I don't know if that one is shared between the CPU and GPU in AMD packages. Brice Le 27/03/2015 10:13, Samy CHBINOU a écrit : > Hello, > I run lstopo on my APU A10-7850K (4CPUs + 8 GPUs), they are detected > (see included picture) but the 8 GPUs are detected on the PCI bus, > while they are on the same die as the CPUs and directly share parts of > the RAM. > ... I don't understand the signification of the numbers PCI 1002:130f > ... to what does it correspond? > a dmesg gives me : pci 0000:00:01.0: [1002:130f] type 00 class 0x030000
