On Sun, Mar 15, 2020 at 8:46 AM Matt B <[email protected]> wrote: > > Hi, > > Thanks for the info on DRI_PRIME. I'm looking at doing a motherboard+cooler > swap. > > Have you run battery life tests with/without the dGPU hidden? > Isn't NVRAM the "correct" way to do it? > > -Matt > > On Wed, Mar 11, 2020 at 11:24 AM Mike Banon <[email protected]> wrote: >> >> On Sun, Mar 8, 2020 at 6:56 PM Matt B <[email protected]> wrote: >> > >> > To those who have a dual-GPU G505s and have enabled the recent support for >> > the dGPU, does DRI_PRIME GPU offloading work? >> > >> >> Yes, DRI_PRIME GPU offloading works on G505S - and actually I think >> it's the only possible way of using a discrete GPU on this coreboot'ed >> laptop at the moment. However, it's usefulness really depends on your >> software and how fast is the integrated GPU's RAM: you may have >> 1600MHz CL9 RAM modules installed, while a discrete GPU's own soldered >> chips are i.e. 1333MHz CL9 only. I recommend testing them side by side >> in all the programs you care about and remember which one is better in >> what cases. >> >> > >> > Could here be an option created to enable/disable the dGPU at boot time >> > without recompiling? >> > >> >> If there could be a real benefit from doing so (does it consume less >> power if you "hide it" from your system by disabling a PCI bus?), such >> a way could be implemented - however, I don't know how to do it >> without using NVRAM (which I would like to avoid for various reasons)
Ideally I prefer a laptop as stateless as possible at least regarding the BIOS - and also, at least some G505S laptops seem to have a broken NVRAM (strong freeze on a write attempt to NVRAM and laptop doesn't turn on until a full motherboard discharge). A10-5750M TDP is 35W, after looking at the similar CPUs from this table - https://www.eteknix.com/amd-kaveri-a10-7850k-overclocking-analysis/7/ , it seems that: Idle | GPU load | CPU+GPU load 15.6W | 23.66W | 35W 44.5% | 67.6% | 100% And "where the ACPI VFCT Table is missing, Power States may not work properly" - so a constant iGPU load, caused by the missing ACPI VFCT for iGPU in our case, adds about 8W to power consumption ~ about 12% less battery life. HD 8570M or R5 M230 might be 20W TDP at max load (hard to find exact info), but its' ACPI VFCT table is loaded by coreboot, so a power consumption difference shouldn't be as big. Currently I'm trying to add the XMP profiles support ( see https://review.coreboot.org/c/coreboot/+/40291 ) so it could take me a while to return to this research. _______________________________________________ coreboot mailing list -- [email protected] To unsubscribe send an email to [email protected]
