Gentoo Forums
Gentoo Forums
Gentoo Forums
Quick Search: in
AM4 Mobos used for Ryzen
View unanswered posts
View posts from last 24 hours

Goto page Previous  1, 2, 3, 4, 5, 6, 7  Next  
Reply to topic    Gentoo Forums Forum Index Gentoo Chat
View previous topic :: View next topic  

What AM4 motherboard do you use. Actually use, not want.
Asus
37%
 37%  [ 29 ]
MSI
19%
 19%  [ 15 ]
Gigabyte
22%
 22%  [ 17 ]
Asrock
20%
 20%  [ 16 ]
Total Votes : 77

Author Message
Tony0945
Advocate
Advocate


Joined: 25 Jul 2006
Posts: 4297
Location: Illinois, USA

PostPosted: Fri Jul 19, 2019 1:15 pm    Post subject: Reply with quote

saellaven wrote:
I ran into the issue where the Ryzen CPUs are missing instructions that were available in the FX-8350, but even after rebuilding everything with march=k8 mtune=generic, I couldn't get anything to run once the kernel passed things off (trapping invalid opcodes). So, I did a fresh install from a stage 3.

I have B350 mobo with the AM4 bdver4 Athlon X4. I've been using these flags to make the upgrade easier. I hope to install my 2700X, that was delivered last Saturday, this weekend. I'll report back here how it worked.
Code:
CFLAGS="-march=native -mno-xop -mno-tbm -mno-fma4 -O2 -pipe"
I certainly hope to avoid a painful reinstall.
Back to top
View user's profile Send private message
Anon-E-moose
Watchman
Watchman


Joined: 23 May 2008
Posts: 5013
Location: Dallas area

PostPosted: Fri Jul 19, 2019 1:35 pm    Post subject: Reply with quote

These seem to be the ones from the fx83* series that aren't in ryzen3

fma4 lwp nodeid_msr tbm xop

Looks like when I get ready to move to ryzen3, I'll do an emerge world with these missing and after the install onto ryzen3 another emerge with proper flags.

Yeah, Tony, I hate to reinstall too.

Edit to add: out of the 5 missing from ryzen3 only 2 of them are in my cpu flags in make.conf fma4 and xop.
I'm not sure if gcc uses the other 3 but I think I'll pass the -no options with those anyway.
ETA2: seems gcc doesn't use/care about nodeid_msr and tbm, so only 3 to worry about (fma4, lwp and xop)
_________________
PRIME x570-pro, 3700x, RX 550 - 5.8 zen kernel
Acer E5-575 (laptop), i3-7100u - i965 - 5.5 zen kernel
---both---
gcc 9.3.0, profile 17.1 (no-pie & modified) amd64-no-multilib, eudev, openrc, openbox, palemoon
Back to top
View user's profile Send private message
Chiitoo
Administrator
Administrator


Joined: 28 Feb 2010
Posts: 2160
Location: Here and Away Again

PostPosted: Fri Jul 19, 2019 3:12 pm    Post subject: Reply with quote

saellaven wrote:
So, the 3700x is installed and Gentoo is running...

I'm somewhat interested in what this little test program will output for you (not my code, but something I spotted on IRC some time ago):

Code:
// https://pastebin.com/raw/CxWEU3XE

#include <stdio.h>

int main()
{
   unsigned calls = 0;
   unsigned failed = 0;
   unsigned zero = 0;
   unsigned one = 0;
   const unsigned count = 1000000;
   for(calls = 0; calls < count; calls++) {
      unsigned long long v;
      unsigned char f;
      asm volatile("rdrand %0; setc %1" : "=r"(v), "=qm"(f));
      if(f) {
         if(!v) zero++;
         if(!~v) one++;
      } else {
         failed++;
      }
   }
   if(zero>2||one>2) printf("WARNING: rdrand FAILED sanity check. CPU buggy?");
   printf("%u calls, %u failed, %u zeros, %u ones\n", calls, failed, zero, one);
   return !zero && !one;
}

When all is fine, it probably should say:

Code:
1000000 calls, 0 failed, 0 zeros, 0 ones

Quote:
Code:

 # genlop -t libreoffice webkit-gtk qtwebkit
 * app-office/libreoffice

     Fri Jul 19 04:02:45 2019 >>> app-office/libreoffice-6.2.5.2
       merge time: 33 minutes and 27 seconds.

 * net-libs/webkit-gtk

     Thu Jul 18 23:18:52 2019 >>> net-libs/webkit-gtk-2.24.2
       merge time: 49 minutes and 16 seconds.

     Fri Jul 19 07:25:21 2019 >>> net-libs/webkit-gtk-2.24.2
       merge time: 12 minutes and 57 seconds.

 * dev-qt/qtwebkit

     Thu Jul 18 21:33:41 2019 >>> dev-qt/qtwebkit-5.212.0_pre20180120
       merge time: 14 minutes and 26 seconds.

There might be a better thread for this but... USE-flags used might be interesting for comparing, too. :]

I just tested libreoffice (don't have ruby installed at all right now, so didn't try the others), and it took me 54 minutes with a Ryzen 1700 (MAKEOPTS="-j16"). Two threads may have been somewhat busy with a game during all that, but still, an interesting difference.

Code:
app-office/libreoffice-6.2.5.2::gentoo was built with the following:
USE="branding mariadb -accessibility -bluetooth -coinmp -cups -dbus -debug -eds -firebird -googledrive -gstreamer -gtk -gtk2 -java -kde -ldap -odk -pdfimport -postgres -test -vlc" ABI_X86="(64)" LIBREOFFICE_EXTENSIONS="-nlpsolver -scripting-beanshell -scripting-javascript -wiki-publisher" PYTHON_SINGLE_TARGET="python3_6 -python2_7 -python3_5 -python3_7" PYTHON_TARGETS="python3_6 -python2_7 -python3_5 -python3_7"

_________________
Kind regards,
Chiitoo.

You might remember me from Gentoo projects such as Forums, LXQt, Qt, and Wine.
Back to top
View user's profile Send private message
Tony0945
Advocate
Advocate


Joined: 25 Jul 2006
Posts: 4297
Location: Illinois, USA

PostPosted: Fri Jul 19, 2019 3:59 pm    Post subject: Reply with quote

Added "-no-wlp"
Since I have to temporarily rebuild anyway, perhaps it would be better to leave out march and set mtune=generic.
EDIT:
Wimped out and after reading this discussion I changed to
Code:
CFLAGS="-march=x86-64 -mtune=native -O2 -pipe"
and am rebuilding with
Code:
time emerge -aev @system

This should be sufficient for the ryzen to rebuild world. No sense in rebuilding Mate and thunderbird twice.
Back to top
View user's profile Send private message
saellaven
Guru
Guru


Joined: 23 Jul 2006
Posts: 577

PostPosted: Fri Jul 19, 2019 6:13 pm    Post subject: Reply with quote

Chiitoo wrote:
saellaven wrote:
So, the 3700x is installed and Gentoo is running...

I'm somewhat interested in what this little test program will output for you (not my code, but something I spotted on IRC some time ago):

Code:
// https://pastebin.com/raw/CxWEU3XE

#include <stdio.h>

int main()
{
   unsigned calls = 0;
   unsigned failed = 0;
   unsigned zero = 0;
   unsigned one = 0;
   const unsigned count = 1000000;
   for(calls = 0; calls < count; calls++) {
      unsigned long long v;
      unsigned char f;
      asm volatile("rdrand %0; setc %1" : "=r"(v), "=qm"(f));
      if(f) {
         if(!v) zero++;
         if(!~v) one++;
      } else {
         failed++;
      }
   }
   if(zero>2||one>2) printf("WARNING: rdrand FAILED sanity check. CPU buggy?");
   printf("%u calls, %u failed, %u zeros, %u ones\n", calls, failed, zero, one);
   return !zero && !one;
}

When all is fine, it probably should say:

Code:
1000000 calls, 0 failed, 0 zeros, 0 ones


Code:

$ gcc CxWEU3XE.c
$ ./a.out
WARNING: rdrand FAILED sanity check. CPU buggy?1000000 calls, 0 failed, 0 zeros, 1000000 ones


I'm assuming this is the same call that was causing systemd to hang...

Quote:

There might be a better thread for this but... USE-flags used might be interesting for comparing, too. :]

I just tested libreoffice (don't have ruby installed at all right now, so didn't try the others), and it took me 54 minutes with a Ryzen 1700 (MAKEOPTS="-j16"). Two threads may have been somewhat busy with a game during all that, but still, an interesting difference.

Code:
app-office/libreoffice-6.2.5.2::gentoo was built with the following:
USE="branding mariadb -accessibility -bluetooth -coinmp -cups -dbus -debug -eds -firebird -googledrive -gstreamer -gtk -gtk2 -java -kde -ldap -odk -pdfimport -postgres -test -vlc" ABI_X86="(64)" LIBREOFFICE_EXTENSIONS="-nlpsolver -scripting-beanshell -scripting-javascript -wiki-publisher" PYTHON_SINGLE_TARGET="python3_6 -python2_7 -python3_5 -python3_7" PYTHON_TARGETS="python3_6 -python2_7 -python3_5 -python3_7"


Code:
app-office/libreoffice-6.2.5.2::gentoo 
USE="bluetooth branding cups dbus gtk java ldap mariadb pdfimport -accessibility -coinmp -debug -eds -firebird -googledrive -gstreamer -gtk2 -kde -odk -postgres -test -vlc" LIBREOFFICE_EXTENSIONS="-nlpsolver -scripting-beanshell -scripting-javascript -wiki-publisher" PYTHON_SINGLE_TARGET="python3_6 -python2_7 -python3_5 -python3_7" PYTHON_TARGETS="python2_7 python3_6 -python3_5 -python3_7" 0 KiB


So, it looks like I'm building more into mine...
Back to top
View user's profile Send private message
Chiitoo
Administrator
Administrator


Joined: 28 Feb 2010
Posts: 2160
Location: Here and Away Again

PostPosted: Fri Jul 19, 2019 6:31 pm    Post subject: Reply with quote

saellaven wrote:
Code:

$ gcc CxWEU3XE.c
$ ./a.out
WARNING: rdrand FAILED sanity check. CPU buggy?1000000 calls, 0 failed, 0 zeros, 1000000 ones


I'm assuming this is the same call that was causing systemd to hang...

Yup.

I guess it was to be expected, really.

I like it how Phoronix says "not being able to boot newer Linux distributions". Might want to replace that 'newer' and maybe 'Linux distributions' with something, if it only affects systemd that way. (Does it only affect systemd in that way?)

Quote:
Code:
app-office/libreoffice-6.2.5.2::gentoo 
USE="bluetooth branding cups dbus gtk java ldap mariadb pdfimport -accessibility -coinmp -debug -eds -firebird -googledrive -gstreamer -gtk2 -kde -odk -postgres -test -vlc" LIBREOFFICE_EXTENSIONS="-nlpsolver -scripting-beanshell -scripting-javascript -wiki-publisher" PYTHON_SINGLE_TARGET="python3_6 -python2_7 -python3_5 -python3_7" PYTHON_TARGETS="python2_7 python3_6 -python3_5 -python3_7" 0 KiB


So, it looks like I'm building more into mine...

Thanks!

Quite a bit more, yeah. Might do some more tests here, but in any case, looks quite the nice boost indeed, even though the makeopts are the same. I did not expect that.
_________________
Kind regards,
Chiitoo.

You might remember me from Gentoo projects such as Forums, LXQt, Qt, and Wine.
Back to top
View user's profile Send private message
saellaven
Guru
Guru


Joined: 23 Jul 2006
Posts: 577

PostPosted: Fri Jul 19, 2019 6:59 pm    Post subject: Reply with quote

Chiitoo wrote:
saellaven wrote:

I'm assuming this is the same call that was causing systemd to hang...

Yup.

I guess it was to be expected, really.

I like it how Phoronix says "not being able to boot newer Linux distributions". Might want to replace that 'newer' and maybe 'Linux distributions' with something, if it only affects systemd that way. (Does it only affect systemd in that way?)


It's just systemd... so "newer distributions" is an intentional misleading to try to make it seem like anything non-systemd is outdated... the proper nomenclature, is it is causing "systemd based distributions" to fail to boot.

Given that I don't use systemd, I haven't investigated it too deeply, but my understanding is, rather than use system entropy, they ask the CPU for a random number and the random number generator in the Zen based processors is borked, so the systemd devs wrote a function that basically endlessly loops if it doesn't get enough random numbers... to which, the old programmer in me says, always validate your input. It seems like a silly beginner error and another case of NIH syndrome.

Quote:

Quote:
Code:
app-office/libreoffice-6.2.5.2::gentoo 
USE="bluetooth branding cups dbus gtk java ldap mariadb pdfimport -accessibility -coinmp -debug -eds -firebird -googledrive -gstreamer -gtk2 -kde -odk -postgres -test -vlc" LIBREOFFICE_EXTENSIONS="-nlpsolver -scripting-beanshell -scripting-javascript -wiki-publisher" PYTHON_SINGLE_TARGET="python3_6 -python2_7 -python3_5 -python3_7" PYTHON_TARGETS="python2_7 python3_6 -python3_5 -python3_7" 0 KiB


So, it looks like I'm building more into mine...

Thanks!

Quite a bit more, yeah. Might do some more tests here, but in any case, looks quite the nice boost indeed, even though the makeopts are the same. I did not expect that.


Going from the FX-8350 to the 3700x has been pretty amazing so far... other than the above instruction issue, my only complaint is the pain of having to start from a new stage 3 since "-march=k8 -mtune=generic" wouldn't guarantee working binaries.


On a side note, the 3700x has a base clock of 3.6 Mhz and the X570-Pro offered to boost me up to 3.8 Mhz just based on the cooler (stock tower cooler) I have. No stability issues so far and I'm not really one to try to max out what the processor can do in terms of overclocking.
Back to top
View user's profile Send private message
Anon-E-moose
Watchman
Watchman


Joined: 23 May 2008
Posts: 5013
Location: Dallas area

PostPosted: Fri Jul 19, 2019 9:19 pm    Post subject: Reply with quote

Chiitoo wrote:
(Does it only affect systemd in that way?)


Interesting patch /usr/portage/sys-apps/systemd/files/242-rdrand-ryzen.patch
_________________
PRIME x570-pro, 3700x, RX 550 - 5.8 zen kernel
Acer E5-575 (laptop), i3-7100u - i965 - 5.5 zen kernel
---both---
gcc 9.3.0, profile 17.1 (no-pie & modified) amd64-no-multilib, eudev, openrc, openbox, palemoon
Back to top
View user's profile Send private message
Tony0945
Advocate
Advocate


Joined: 25 Jul 2006
Posts: 4297
Location: Illinois, USA

PostPosted: Fri Jul 19, 2019 9:23 pm    Post subject: Reply with quote

time for emerge ev on the Athlon X3 -j4
Code:
real   223m6.930s
user   396m13.664s
sys   58m16.144s

I once tried j8 based on cores but the temperature got up to 100C several times resulting in a crash. I'll use -j8 on the 2700X, although the cooler is massively bigger. I hope it fiits mt mobo. It's supposed to.

EDIT: That's @system, not @world
Back to top
View user's profile Send private message
Tony0945
Advocate
Advocate


Joined: 25 Jul 2006
Posts: 4297
Location: Illinois, USA

PostPosted: Fri Jul 19, 2019 11:29 pm    Post subject: Reply with quote

Dragged up the old box. WOW! I hope that Prism heatsink is going to fit. It's HUGE!


https://imgur.com/2747Znj
Back to top
View user's profile Send private message
pjp
Administrator
Administrator


Joined: 16 Apr 2002
Posts: 18593

PostPosted: Sat Jul 20, 2019 1:03 am    Post subject: Reply with quote

One aspect that amazes me is the difference between PC / consumer cooling versus server / enterprise cooling. I've never seen anything in a server that resembled anything I've seen in PCs.
_________________
Your lips move, but I can't hear what you're saying.
Back to top
View user's profile Send private message
Tony0945
Advocate
Advocate


Joined: 25 Jul 2006
Posts: 4297
Location: Illinois, USA

PostPosted: Sat Jul 20, 2019 2:31 am    Post subject: Reply with quote

More prep, just in case:
1. rebuilt the kernel moving CPU to generic and max CPU's from 8 to 16.
2. emerge -a1v bash busybox nano openrc refind
Back to top
View user's profile Send private message
sunset_sergal
n00b
n00b


Joined: 22 May 2019
Posts: 20
Location: GA, USA

PostPosted: Sat Jul 20, 2019 2:49 pm    Post subject: Reply with quote

To everyone discussing rebuilding Gentoo packages on a machine that already has an FX-8000 series CPU so it can migrate to Ryzen 3000: the missing instructions are actually 3DNow! and related extensions. This was news way back in 2010: https://www.tomshardware.com/news/3dnow-simd-extensions-phenom-sse,11128.html

When I need a Gentoo install to run on just any amd64 system I usually use -march=core2. This STILL isn't perfect because I ran into issues with an older AMD CPU of the same age that didn't support SSSE3 (Yes, three Ss). You should probably just recompile for generic x86-64 if you're worried about unsupported instructions; if your original and target CPUs have mutually incompatible instructions, you'll have to rebuild everything twice anyway, so just pick something safe and save yourself the headache of optimizing that in-between state.
Back to top
View user's profile Send private message
Tony0945
Advocate
Advocate


Joined: 25 Jul 2006
Posts: 4297
Location: Illinois, USA

PostPosted: Sat Jul 20, 2019 7:17 pm    Post subject: Reply with quote

sunset_sergal wrote:
You should probably just recompile for generic x86-64 if you're worried about unsupported instructions.

As I said above:
Code:
Code:   CFLAGS="-march=x86-64 -mtune=native -O2 -pipe"


I was reading a Phoronix article a few days ago where he compared various benchmarks against a plain Ubuntu version and the same pacakge recompiled with -march-native. There was a ray tracing app and I think some games where it made a big difference. Gcc improved by 10%, most improvements were smaller. that got me thing about things like openrc, grub, busybox and such where any speedup is probably unnoticeable because they take no time anyway, but are vital, perhaps they should have custom CFLAGS in /etc/env.d so they don't need rebuilding.
Then there are things like T-bird, Firefox, word processors and such that spend most of their time waiting for human input or the internet. I build Palemoon, and it loads noticbly faster, but probably because the binary version has bundled old libraries and I compile with useflags to use the system libraries. Ten percent isn't much for gcc, but it can be a lot for huge packages that take an hour or two to build. In the dawn of computing, chip architectures were so different that one could get huge speed increases by custom builds, but today architecture differences are really minor.
Back to top
View user's profile Send private message
Tony0945
Advocate
Advocate


Joined: 25 Jul 2006
Posts: 4297
Location: Illinois, USA

PostPosted: Sat Jul 20, 2019 11:24 pm    Post subject: Reply with quote

Well, I did the install. One bad thing was that I didn't pull the power plug. As soon as I closed the lever on the CPU (not the heat sink, heat sink wasn't on yet) the power came on and the system fans started whirring and lights blinking while I tried to find the black power button on the black case. Also, I had to turn the heat sink 180 to get it on. Should have removed the video card instead. Hope I didn't hurt thermal paste. It ran hot, about 146 F vs around 86F at idle for the Athlon X4 with a MUCH smaller heat sink. I rebooted into the BIOS and set fans to always run at max speed. A little noisy for idle but that's OK. I rebuilt the kernel, still at -j5 and the temperature soared back to the mid 140's with very unstable frequency. Frequency hunts a lot at idle also. The kernel did build quickly.
I'm going to build -e @system and see how it compares to the Athlon X4.

The Back-UPS reports supplying 83W at idle vs 56W with the Athlon. But the Athlon idled at a much lower frequency.
Anon-E-Moose, do you have any frequency driver kernel recommendations?
It's settled at mostly around 2.0GHz and 100 F, but frequency wobbles between 1.8 and 2.2 seemingly randomly. Running on the native kernel.
Back to top
View user's profile Send private message
Anon-E-moose
Watchman
Watchman


Joined: 23 May 2008
Posts: 5013
Location: Dallas area

PostPosted: Sun Jul 21, 2019 12:08 am    Post subject: Reply with quote

I have userspace, ondemand and performance set but default is ondemand.

This is on an fx8320
Code:
cat scaling_available_frequencies
3500000 2900000 2300000 1700000 1400000


it typically bounces between 1400 and 1700 unless I'm compiling.

I'm not sure what freq's the ryzen offers but ondemand is probably going to give the best overall, while keeping temps down.

For the future,
Just remember to either unplug the power supply or flip the switch on it to off (if it has one) and give it a minute or so to discharge capacitors.
_________________
PRIME x570-pro, 3700x, RX 550 - 5.8 zen kernel
Acer E5-575 (laptop), i3-7100u - i965 - 5.5 zen kernel
---both---
gcc 9.3.0, profile 17.1 (no-pie & modified) amd64-no-multilib, eudev, openrc, openbox, palemoon
Back to top
View user's profile Send private message
Tony0945
Advocate
Advocate


Joined: 25 Jul 2006
Posts: 4297
Location: Illinois, USA

PostPosted: Sun Jul 21, 2019 1:52 am    Post subject: Reply with quote

Anon-E-moose wrote:

For the future,
Just remember to either unplug the power supply or flip the switch on it to off (if it has one) and give it a minute or so to discharge capacitors.

I'm mortified.
I usually unplug it and put the case on it's side on a work bench. It was so tempting to just drop the side. The mobo mounts on the side of the toolless case. Glad I didn't blow the chip.

with MAKEOPTS="-j8". Hardly broke into a sweat, speedwise. was mostly at 2.5 Ghz, temp at 146 F.
Code:
real   107m12.465s
user   228m41.438s
sys   29m1.438s
About twice as fast as the X4 with -j5.
I'm tempted to go for broke. But one more test with -j17...
I want to see that 3.9 Ghz.
Code:
cat scaling_available_frequencies
cat: scaling_available_frequencies: No such file or directory

I'm really surprised that the video card is so cool. That big fan must be drawing air past the video card, which has no fan of it's own.
Back to top
View user's profile Send private message
saellaven
Guru
Guru


Joined: 23 Jul 2006
Posts: 577

PostPosted: Sun Jul 21, 2019 2:14 am    Post subject: Reply with quote

3700X
Code:

/sys/devices/system/cpu/cpufreq/policy0/scaling_available_frequencies
3800000 2800000 2200000


Looking at my UPS output, I'm currently using 97.5 watts at idle, but that's also including my monitor, router, and cable modem.

I just noticed my hardware temps aren't being read... I'll need to play with modules to see if I can get it working

sensors-detect is telling me

Code:

Driver `nct6775':
  * ISA bus, address 0x290
    Chip `Nuvoton NCT6796D Super IO Sensors' (confidence: 9)


but with the module loaded, sensors says no sensors found
Back to top
View user's profile Send private message
Tony0945
Advocate
Advocate


Joined: 25 Jul 2006
Posts: 4297
Location: Illinois, USA

PostPosted: Sun Jul 21, 2019 2:24 am    Post subject: Reply with quote

saellaven wrote:

sensors-detect is telling me

Code:

Driver `nct6775':
  * ISA bus, address 0x290
    Chip `Nuvoton NCT6796D Super IO Sensors' (confidence: 9)


but with the module loaded, sensors says no sensors found

Which kernel? I have the MSI B350 Tomahawk Arctic with that chip and kernel support was not there at first. The module is table driven and at first my variant wasn't in it.
Code:
uname -a
Linux MSI 4.19.59-gentoo-ryzen #3 SMP Sat Jul 20 17:50:05 CDT 2019 x86_64 AMD Ryzen 7 2700X Eight-Core Processor AuthenticAMD GNU/Linux


Code:
cd /usr/src/linux
zcat /proc/config.gz >.config
#test for backporting nct6775.c
LONGVER=$(uname -r)
SHORTVER=$(echo -n $LONGVER | cut -f1,2 -d'.')
if [ $SHORTVER = 4.4 ] || [ $SHORTVER = 4.9 ]; then
 cp /usr/src/new_nct6775.c drivers/hwmon/nct6775.c
fi
Back to top
View user's profile Send private message
Tony0945
Advocate
Advocate


Joined: 25 Jul 2006
Posts: 4297
Location: Illinois, USA

PostPosted: Sun Jul 21, 2019 2:37 am    Post subject: Reply with quote

Code:
 cat /sys/devices/system/cpu/cpufreq/policy0/scaling_available_frequencies
3700000 3200000 2200000

But the frequency display has St. Vitus Dance. Often jiggling by 100Mhz.
Back to top
View user's profile Send private message
saellaven
Guru
Guru


Joined: 23 Jul 2006
Posts: 577

PostPosted: Sun Jul 21, 2019 5:56 am    Post subject: Reply with quote

Tony0945 wrote:
saellaven wrote:

sensors-detect is telling me

Code:

Driver `nct6775':
  * ISA bus, address 0x290
    Chip `Nuvoton NCT6796D Super IO Sensors' (confidence: 9)


but with the module loaded, sensors says no sensors found

Which kernel? I have the MSI B350 Tomahawk Arctic with that chip and kernel support was not there at first. The module is table driven and at first my variant wasn't in it.


Code:

# uname -a
Linux alpha 5.2.1 #9 SMP PREEMPT Sun Jul 21 01:14:22 EDT 2019 x86_64 AMD Ryzen 7 3700X 8-Core Processor AuthenticAMD GNU/Linux


NCT6796D support was added in 5.0 from what I can tell and a quick

Code:

# grep nct6796d /usr/src/linux/drivers/hwmon/nct6775.c
 * nct6796d    14      7       7       2+6    0xd420 0xc1    0x5ca3


shows support is there.

I do get the following message in /var/log/messages though
Code:

Jul 21 01:42:57 alpha kernel: nct6775: Found NCT6798D or compatible chip at 0x2e:0x290
Jul 21 01:42:57 alpha kernel: ACPI Warning: SystemIO range 0x0000000000000295-0x0000000000000296 conflicts with OpRegion 0x0000000000000290-0x0000000000000299 (\x5cAMW0.SHWM) (20190509/utaddress-213)
Jul 21 01:42:57 alpha kernel: ACPI: If an ACPI driver is available for this device, you should use it instead of the native driver


Other than sensors, I think I've got everything else working on this board...

going into the hardware monitor in BIOS shows me idling around 100F at boot... I'm not sure how hot I get under full load.

I was mildly amused that gkrellm ended up taking up my entire vertical screen with 16 individual processors loaded. I'm running an emerge -1 @system for you to give you a comparison there too...
Back to top
View user's profile Send private message
saellaven
Guru
Guru


Joined: 23 Jul 2006
Posts: 577

PostPosted: Sun Jul 21, 2019 6:27 am    Post subject: Reply with quote

Again, on my 3700x

Code:

# time emerge -1 @system

real   27m1.988s
user   155m32.620s
sys   16m51.608s


For most of the time, the processors were at idle during the gcc build, which took up the vast majority of the time (and, as the final emerge, was the only thing running - the first 30 packages built simultaneously):

Code:

# genlop -t gcc
 * sys-devel/gcc

     Thu Jul 18 22:29:07 2019 >>> sys-devel/gcc-8.3.0-r1
       merge time: 21 minutes and 48 seconds.

     Fri Jul 19 09:29:40 2019 >>> sys-devel/gcc-8.3.0-r1
       merge time: 20 minutes and 39 seconds.

     Sun Jul 21 02:21:20 2019 >>> sys-devel/gcc-8.3.0-r1
       merge time: 20 minutes and 57 seconds.
Back to top
View user's profile Send private message
Anon-E-moose
Watchman
Watchman


Joined: 23 May 2008
Posts: 5013
Location: Dallas area

PostPosted: Sun Jul 21, 2019 10:21 am    Post subject: Reply with quote

Tony0945 wrote:
Code:
 cat /sys/devices/system/cpu/cpufreq/policy0/scaling_available_frequencies
3700000 3200000 2200000

But the frequency display has St. Vitus Dance. Often jiggling by 100Mhz.


The scaling_cur_freq doesn't give the base freq, it gives the one it sees (in motion) so it's going to bounce quite a bit.
I noticed because I used to use cpuinfo_cur_freq which stabilizes to one of the frequencies. the scaling_cur_freq is all over the place, but typically just above one of the static frequencies. I watch it with conky and all 8 threads move every refresh (second in my case). That's just the way they changed the reporting, but it's actually a more accurate view of what your cpu is doing.
_________________
PRIME x570-pro, 3700x, RX 550 - 5.8 zen kernel
Acer E5-575 (laptop), i3-7100u - i965 - 5.5 zen kernel
---both---
gcc 9.3.0, profile 17.1 (no-pie & modified) amd64-no-multilib, eudev, openrc, openbox, palemoon
Back to top
View user's profile Send private message
Anon-E-moose
Watchman
Watchman


Joined: 23 May 2008
Posts: 5013
Location: Dallas area

PostPosted: Sun Jul 21, 2019 10:41 am    Post subject: Reply with quote

saellaven wrote:
3700X
Code:

/sys/devices/system/cpu/cpufreq/policy0/scaling_available_frequencies
3800000 2800000 2200000


I wish they would have allowed a lower freq, but that all up to the chip designers, my i7 in the laptop only has 2 freq, full out and a little less that 1/3.

Quote:
Looking at my UPS output, I'm currently using 97.5 watts at idle, but that's also including my monitor, router, and cable modem.


Not bad

Quote:
I just noticed my hardware temps aren't being read... I'll need to play with modules to see if I can get it working

sensors-detect is telling me

Code:

Driver `nct6775':
  * ISA bus, address 0x290
    Chip `Nuvoton NCT6796D Super IO Sensors' (confidence: 9)


but with the module loaded, sensors says no sensors found


The kernel (I use 5.0.13) says it understands that version (NCT6796D) of the chip
What version of lm_sensors are you using?

a little more here (from earlier this year) https://bbs.archlinux.org/viewtopic.php?id=244529
But I don't know if this will help or not.
_________________
PRIME x570-pro, 3700x, RX 550 - 5.8 zen kernel
Acer E5-575 (laptop), i3-7100u - i965 - 5.5 zen kernel
---both---
gcc 9.3.0, profile 17.1 (no-pie & modified) amd64-no-multilib, eudev, openrc, openbox, palemoon
Back to top
View user's profile Send private message
saellaven
Guru
Guru


Joined: 23 Jul 2006
Posts: 577

PostPosted: Sun Jul 21, 2019 2:13 pm    Post subject: Reply with quote

Anon-E-moose wrote:
saellaven wrote:
I just noticed my hardware temps aren't being read... I'll need to play with modules to see if I can get it working

sensors-detect is telling me

Code:

Driver `nct6775':
  * ISA bus, address 0x290
    Chip `Nuvoton NCT6796D Super IO Sensors' (confidence: 9)


but with the module loaded, sensors says no sensors found


The kernel (I use 5.0.13) says it understands that version (NCT6796D) of the chip
What version of lm_sensors are you using?

a little more here (from earlier this year) https://bbs.archlinux.org/viewtopic.php?id=244529
But I don't know if this will help or not.


I just tried lm_sensors-3.5.0_p20190505 (side note: ugh to having to keep adjusting lm_sensors in my overlay when it works perfectly fine on openrc-0.17) and still nothing... following your link and looking at the lm_sensors git repository, still no luck on any patches. According to the links, AMD and Nuvoton are both being secretive with NDAs about the sensors, so who knows if/when support will happen.
Back to top
View user's profile Send private message
Display posts from previous:   
Reply to topic    Gentoo Forums Forum Index Gentoo Chat All times are GMT
Goto page Previous  1, 2, 3, 4, 5, 6, 7  Next
Page 3 of 7

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum