AMD Preferred Core Patches Updated For Linux

[phoronix]

Phoronix: AMD Preferred Core Patches Updated For Linux

On the AMD CPU side of the house, one of the patch series we are looking forward to seeing upstreamed in the Linux kernel is the AMD Preferred Core functionality that was initially sent out this summer. This AMD Preferred Core handling is built onto the AMD P-State driver and has been undergoing a few rounds of iteration with the latest "v5" patches having been posted this week...

https://www.phoronix.com/news/AMD-Preferred-Core-Linux-v5

[skeevy420]

While this won't be a problem that affects me because the 79??x3d CPUs are out of my price range, curiosity still makes me wonder if Linux scheduling and/or AMD P-State handle the x3d v-cache being on one CCD and not the other. I know that AMD has released Windows tools to assist with proper core parking & core usage so I'm just curious if Linux will get something similar. To my Layman self, staying in the correct cache group seems like it should be part of what Preferred Core handling and schedulers are trying to accomplish.

[Myownfriend]

While this won't be a problem that affects me because the 79??x3d CPUs are out of my price range, curiosity still makes me wonder if Linux scheduling and/or AMD P-State handle the x3d v-cache being on one CCD and not the other. I know that AMD has released Windows tools to assist with proper core parking & core usage so I'm just curious if Linux will get something similar. To my Layman self, staying in the correct cache group seems like it should be part of what Preferred Core handling and schedulers are trying to accomplish.

The v-cache is always added to the primary CCD which I believe also uses a higher voltage so it can clock higher. I'm not sure if the presence of v-cache is accounted for in scheduling but that CCD is still going to be preferred for the reasons I mentioned above.

[jeisom]

The v-cache is always added to the primary CCD which I believe also uses a higher voltage so it can clock higher. I'm not sure if the presence of v-cache is accounted for in scheduling but that CCD is still going to be preferred for the reasons I mentioned above.

The v-cache ccd clocks lower than the non-v-cache cdd.

[jeisom]

I tested the V4 on my zen3 system and I noticed that the system reports that the preferred cores max out/reach 6.44GHz. I assume this is not what they are actually reaching. I would think that it should report actual speed if this merged.

[Myownfriend]

The v-cache ccd clocks lower than the non-v-cache cdd.

Ah, I didn't know there was any other difference in behavior with the X3D chips.