Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

65% - 94% mba cos value generates the same result 90% #69

Open
congxu2016 opened this issue Apr 5, 2018 · 6 comments
Open

65% - 94% mba cos value generates the same result 90% #69

congxu2016 opened this issue Apr 5, 2018 · 6 comments

Comments

@congxu2016
Copy link

congxu2016 commented Apr 5, 2018

I am using pqos library (i.e. pqos_mba_set()) to set the MBA COS dynamically. I found no matter what value I use between 65% - 94% on my machine, the MBA COS is ultimately set to 90%. My skylake platform is Xeon Gold 6150. OS is ubuntu 16.04.2 (kernel 4.13).

@mdcornu
Copy link
Contributor

mdcornu commented Apr 9, 2018

Thanks for contacting us with this issue.
We will investigate and report back when we have updates.

Regards,
Marcel

@congxu2016 congxu2016 changed the title 60% - 90% mba cos value generates the same result 90% 65% - 94% mba cos value generates the same result 90% Apr 12, 2018
@meteorfox
Copy link

I also noticed this, I tried writing the MSR directly using wrmsr and noticed it still returned "0xA" which corresponds to 100 - 0xA or 90% of available bandwidth.

MBA MSR starts at address 0xD50

To set COS1, that would be address 0xD50+1 => 0xD51.

In my case, I tried to set it to 70% available bandwidth, that's translates to 30% throttle or 0x1E.

# rdmsr 0xD51
0

# wrmsr 0xD51 0x1E

# rdmsr 0xD51
a    <== Should be 0x1e not 0xa

@gureya
Copy link

gureya commented Nov 4, 2019

Hi,

Has this issue been resolved, I'm also getting a similar behavior on Intel(R) Xeon(R) Silver 4114 i.e., the MBA levels of 70% & 80% are ultimately set to 90%

@madhura-a
Copy link

Similar behavior is observed on Intel(R) Xeon(R) Gold 6126 CPU. Not only for MBA levels, for some values of LLC cache ways values are not set properly. For example, When I try to set LLC cache ways to 0x400, it is updating to 0x600.

@limitedge
Copy link

These issues are under investigation, thanks for your patience.

@limitedge
Copy link

This is a known issue on certain generations of Xeon eg Skylake and Cascadelake which should now be in the sightings report, and this will be fixed in subsequent generations. Thanks

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

6 participants