May 31, 201610 yr . . . Edit: Might as well downgrade this one, low priority, as I appear to be the only one so far with the problem! Plus, there's an easy workaround, by adding the modprobe to the go file. unRAID OS Version: v6.1.9, and all v6 versions that I've checked, since the 6.0 betas (6.2-betas were not checked) Description: As of v5.0.6 (and many versions before), powernow_k8 was loaded for AMD CPU's, and effected constant CPU stepping, which results in power saving. For all v6 versions I've checked, it is not loaded, resulting in max CPU speed at all times. How to reproduce: Dashboard and cat /proc/cpuinfo and cpufreq-info all show CPU staying at 2411MHz. After running modprobe powernow_k8, CPU stepping is immediately in effect, going down to 1000MHz on my system. Expected results: CPU speed ranging from 1000MHZ to 2400MHz Actual results: CPU speed stays at 2411MHz Other information: I do not know if this is true for all AMD CPU's, but is true for mine (AMD Athlon 64 X2 Dual Core Processor 4600+). To summarize, powernow_k8 was loaded in v5 and the CPU stepped up and down. It's not loaded in v6, and there's no speed stepping. I assume this may have been an oversight with the development of v6. Not an oversight, acpi-cpufreq appears to work fine for all of the other AMD CPU's. Edit: fixed in 6.2
June 1, 201610 yr Author After I posted a notice about this on the 6.1.9 announcement thread, numerous other AMD users indicated they have no issues with their AMD scaling down. So clearly I'm in the minority, the only one so far! This does make sense as more would have reported it before now, if it was a wider problem. It would be interesting to know which AMD chips are working, and what drivers they are loading, with the following command - cpufreq-info -d Mine reports from the syslog "CPU0: AMD Athlon 64 X2 Dual Core Processor 4600+", with no driver loaded (unless I load powernow_k8 myself). Edit to indicate I'm the only one so far!
June 1, 201610 yr After I posted a notice about this on the 6.1.9 announcement thread, 3 others indicated they have no issues with their AMD scaling down. So clearly I'm in the minority, the only one of 4 responders. This does make sense as more would have reported it before now, if it was a wider problem. It would be interesting to know which AMD chips are working, and what drivers they are loading, with the following command - cpufreq-info -d Mine reports from the syslog "CPU0: AMD Athlon 64 X2 Dual Core Processor 4600+", with no driver loaded (unless I load powernow_k8 myself). A8-6600K root@Server_A:~# cpufreq-info -d acpi-cpufreq Sempron 3850 root@Server_B:/boot# cpufreq-info -d acpi-cpufreq And my frequencies do change based upon load
June 1, 201610 yr The powernow_k8 scaling driver is for older amd processors. powernow_k8 - "CPUFreq driver for K8/K10 Athlon 64/Opteron/Phenom processors. Since Linux 3.7 'acpi-cpufreq' will automatically be used for more modern CPUs from this family."
June 2, 201610 yr Author Interesting. On my system, lsmod shows acpi-cpufreq installed, cpufreq-info does not show it, shows no drivers installed, and it's not working. Perhaps there's a CPU profile list somewhere that needs updating? My early syslog ACPI table shows a POWERNOW entry, perhaps it could be detected some how? Perhaps this indicates it's tied to the motherboard, what technologies it supports. Jun 1 10:09:27 JacoBack kernel: ACPI: RSDP 0x00000000000F80B0 000014 (v00 Nvidia) Jun 1 10:09:27 JacoBack kernel: ACPI: RSDT 0x00000000BFFF3040 000034 (v01 Nvidia AWRDACPI 42302E31 AWRD 00000000) Jun 1 10:09:27 JacoBack kernel: ACPI: FACP 0x00000000BFFF30C0 000074 (v01 Nvidia AWRDACPI 42302E31 AWRD 00000000) Jun 1 10:09:27 JacoBack kernel: ACPI: DSDT 0x00000000BFFF3180 006732 (v01 NVIDIA AWRDACPI 00001000 MSFT 0100000E) Jun 1 10:09:27 JacoBack kernel: ACPI: FACS 0x00000000BFFF0000 000040 Jun 1 10:09:27 JacoBack kernel: ACPI: SSDT 0x00000000BFFF99C0 000248 (v01 PTLTD POWERNOW 00000001 LTP 00000001) Jun 1 10:09:27 JacoBack kernel: ACPI: MCFG 0x00000000BFFF9C80 00003C (v01 Nvidia AWRDACPI 42302E31 AWRD 00000000) Jun 1 10:09:27 JacoBack kernel: ACPI: APIC 0x00000000BFFF9900 000072 (v01 Nvidia AWRDACPI 42302E31 AWRD 00000000)
June 2, 201610 yr Author Squid, this does not appear to be a 'common problem', but would it qualify as a candidate for 'Fix Common Problems' (as a warning)? Check for CPU scaling driver and governor, cpufreq-info -d for driver and cpufreq-info -o for governor. If none set, then suggest loading powernow_k8 for AMD, unknown for Intel.
June 2, 201610 yr Squid, this does not appear to be a 'common problem', but would it qualify as a candidate for 'Fix Common Problems' (as a warning)? Check for CPU scaling driver and governor, cpufreq-info -d for driver and cpufreq-info -o for governor. If none set, then suggest loading powernow_k8 for AMD, unknown for Intel. Give me the exact output from your command as it's impossible for me to test a fail Sent from my LG-D852 using Tapatalk
June 2, 201610 yr Author Squid, this does not appear to be a 'common problem', but would it qualify as a candidate for 'Fix Common Problems' (as a warning)? Check for CPU scaling driver and governor, cpufreq-info -d for driver and cpufreq-info -o for governor. If none set, then suggest loading powernow_k8 for AMD, unknown for Intel. Give me the exact output from your command as it's impossible for me to test a fail Without powernow_k8 loaded, I get: root@JacoBack:/boot# cpufreq-info -d root@JacoBack:/boot# cpufreq-info -o minimum CPU frequency - maximum CPU frequency - governor Once powernow_k8 is loaded, I get: root@JacoBack:/boot# cpufreq-info -d powernow-k8 root@JacoBack:/boot# cpufreq-info -o minimum CPU frequency - maximum CPU frequency - governor CPU 0 1000000 kHz ( 41 %) - 2400000 kHz (100 %) - ondemand CPU 1 1000000 kHz ( 41 %) - 2400000 kHz (100 %) - ondemand
June 2, 201610 yr Squid, this does not appear to be a 'common problem', but would it qualify as a candidate for 'Fix Common Problems' (as a warning)? Check for CPU scaling driver and governor, cpufreq-info -d for driver and cpufreq-info -o for governor. If none set, then suggest loading powernow_k8 for AMD, unknown for Intel. Give me the exact output from your command as it's impossible for me to test a fail Sent from my LG-D852 using Tapatalk Just check cpufreq -d for a blank string result and that tells you there is no cpu scaling driver loaded. Nothing more than that. You don't need to get into the governor that is loaded or not loaded.
June 2, 201610 yr This issue is very easy to fix. LT just needs to load the powernow_k8 driver with 'modprobe powernow_k8' in the rc.cpufreq script.
June 3, 201610 yr Squid, this does not appear to be a 'common problem', but would it qualify as a candidate for 'Fix Common Problems' (as a warning)? Check for CPU scaling driver and governor, cpufreq-info -d for driver and cpufreq-info -o for governor. If none set, then suggest loading powernow_k8 for AMD, unknown for Intel. Give me the exact output from your command as it's impossible for me to test a fail Sent from my LG-D852 using Tapatalk Just check cpufreq -d for a blank string result and that tells you there is no cpu scaling driver loaded. Nothing more than that. You don't need to get into the governor that is loaded or not loaded. regardless of how easy it is for LT to fix, the check is in for the next release. Gotta love the one line checks
June 3, 201610 yr Author This issue is very easy to fix. LT just needs to load the powernow_k8 driver with 'modprobe powernow_k8' in the rc.cpufreq script. I suspect it's a little trickier than that, as it needs to be conditional, first on having an AMD processor, and secondly no driver installed or working. I don't think you would want to load powernow_k8 if acpi-cpufreq is working fine, and it seems to be for most AMD systems. It may be that you only want to load it when the motherboard has support for powernow (and acpi-cpufreq isn't working).
June 3, 201610 yr Author Just check cpufreq -d for a blank string result and that tells you there is no cpu scaling driver loaded. Nothing more than that. You don't need to get into the governor that is loaded or not loaded. regardless of how easy it is for LT to fix, the check is in for the next release. Gotta love the one line checks Nice! I'll test it when I can.
June 20, 201610 yr Author I decided to downgrade this one, a low priority, as I appear to be the only one so far with the problem! Plus, there's an easy workaround, by adding the modprobe to the go file.
June 21, 201610 yr I decided to downgrade this one, a low priority, as I appear to be the only one so far with the problem! Plus, there's an easy workaround Another easy workaround is to upgrade to an Intel system.
June 21, 201610 yr Author I decided to downgrade this one, a low priority, as I appear to be the only one so far with the problem! Plus, there's an easy workaround Another easy workaround is to upgrade to an Intel system. I'll be happy to provide PayPal details by PM to anyone for a FundMe - FundRobJ'sIntel System!
September 7, 20169 yr Author After upgrading my 6.1.9 system to 6.2-rc4, I discovered that powernow_k8 is now automatically loaded, so I removed the load line from my go file. (Thanks to whoever made the change!) Consider this solved and closed.
December 16, 20178 yr Sorry for digging up an old thread. I just installed recently on my old gaming rig which contains an AMD Phenom II X4 965 and powernow_k8 does not appear to be loading. Here's what I see when I try the commands on unRAID 6.3.5. root@Tower:~# modprobe powernow_k8 modprobe: ERROR: could not insert 'powernow_k8': No such device root@Tower:~# cpufreq-info -d root@Tower:~# Edited December 16, 20178 yr by _Shorty
December 16, 20178 yr edit: Aha! So the default unRAID config is working just fine with my CPU. I had forgotten that this machine was overclocked in the BIOS by manually setting the multiplier for its gaming lifetime, and now that it is an unRAID box and had this issue I remembered about this. Since I've gone back into the BIOS and set the multiplier back to auto the CPU clock speed behaviour is as expected, throttling down under low load. Doh! ---------- Ah! Underscores and dashes... root@Tower:~# ls -la /lib/modules/$(uname -r)/kernel/drivers/cpufreq/ /lib/modules/$(uname -r)/kernel/arch/*/kernel/cpu/cpufreq/ /bin/ls: cannot access '/lib/modules/4.9.30-unRAID/kernel/arch/*/kernel/cpu/cpufreq/': No such file or directory /lib/modules/4.9.30-unRAID/kernel/drivers/cpufreq/: total 24 drwxr-xr-x 2 root root 120 May 26 2017 ./ drwxr-xr-x 29 root root 580 May 26 2017 ../ -rw-r--r-- 1 root root 6483 May 26 2017 acpi-cpufreq.ko.gz -rw-r--r-- 1 root root 1880 May 26 2017 amd_freq_sensitivity.ko.gz -rw-r--r-- 1 root root 3288 May 26 2017 pcc-cpufreq.ko.gz -rw-r--r-- 1 root root 7072 May 26 2017 powernow-k8.ko.gz Edited December 18, 20178 yr by _Shorty
January 28, 20188 yr Something must be off because mine worked fine until upgrading to 6.4 any ideas on how to manually load drivers for phenom x4
January 28, 20188 yr Ok I got it to load by enabling and cool and quiet stepping in BIOS , but it used to work fine in earlier releases. Also the GUI isn’t showing the movement like it did before.
April 28, 20188 yr Just thought I would address this issue finally and I am running into some problems. tried running these commands and this is what I am getting. Do I need to manually load powernow_k8 somehow? root@Tower:~# modprobe powernow_k8 modprobe: ERROR: could not insert 'powernow_k8': No such device root@Tower:~# cpufreq-info -d root@Tower:~# cpufreq-info -o minimum CPU frequency - maximum CPU frequency - governor Any help or direction would be greatly appreciated, I cant seem to find anything else on the forums related to this EDIT - should add that I am running dual opteron 2419 processors on a Supermicro - H8DM8-2 motherboard Edited April 28, 20188 yr by benyaki
September 30, 20187 yr I'm seeing the same things after getting the warning from fix common problems. I tried the above commands and see the same outputs. I think those are the right commands. Mobo is Supermicro - H8SGL AMD Opteron™ 6212 @ 2600 tower-diagnostics-20180930-1841.zip
Archived
This topic is now archived and is closed to further replies.