[AMD Public Use] Hi Alexander, The issue does not seem to be in the driver, it is in the register values populated by the firmware. We have tested this in the server platforms and it works fine. I've raised an issue internally to analyze and provide a solution for client systems. I'm still waiting for their update. Guenter, This issue is seen on client machines model ranging from 0x40 and above. Should I submit a patch to support only server specific models 0x0 to 0x3f until we triage the issue for client systems. Regards, Naveen -----Original Message----- From: Alexander Monakov <amonakov@xxxxxxxxx> Sent: Thursday, June 11, 2020 1:52 AM To: Naveen Krishna Ch <naveenkrishna.ch@xxxxxxxxx> Cc: Guenter Roeck <linux@xxxxxxxxxxxx>; Chatradhi, Naveen Krishna <NaveenKrishna.Chatradhi@xxxxxxx>; linux-hwmon@xxxxxxxxxxxxxxx Subject: Re: [PATCH 1/3 v7] hwmon: Add amd_energy driver to report energy counters [CAUTION: External Email] On Thu, 28 May 2020, Naveen Krishna Ch wrote: > > No, this isn't right. The workload is not stressing all cores, > > you're reporting 235 J for core 0 and only 35 J for core 7. > > > > If you are running the avx-turbo test, please run it like this: > > > > ./avx-turbo --iters 10000000 --spec avx256_fma_t/8 > > > > This will run AVX code on all 8 cores for about 1 second. To run it > > for longer, increase the --iters argument. > I ran with all cores loaded now and i see that sum of diff of energy > consumed by all cores is greater than the diff of socket energy > consumed in 100secs. > > Kindly, give me time. I will come back with some explanation or solution. Hello. Is there any update on this issue? Thanks. Alexander
