AM4 Mobos used for Ryzen

[shimbob]

3900X on an Asrock B450M Steel Legend, love it!

Only issue is the latest firmware (v2.90) is glitchy and keeps clearing BIOS at boot. Went back to v2.70 which is rock solid stable.

[Ant P.]

Haven't voted in this, but I'm planning on getting a Gigabyte one in the future for several reasons:

1) this model I've found has no stupid LEDs (a rare occurrence)
2) I already have an AM2 one from them and it's lasted a decade with very few issues; everything on it works and the BIOS update routine doesn't have any obnoxious win32 dependencies.

I'll be sure to document every tiny thing wrong with with the firmware when I get it.

And I'm back.

Gigabyte X570-UD
Ryzen 9 3900X
64GB (4 sticks) of Corsair Vengeance LPX 16GB DDR4-3600
…and my old RX550

The BIOS was a minor pain in the ass, but overall feels just like an Unnecessarily Edgy Gamer version of the UIs they had on old Compaq machines. The first two(!) USB keyboards I used played havoc with it so I had to dig out an old nasty PS/2 one. I went and updated it from version F1 to F11, which was drama-free. My usual daily keyboard works fine but I don't know if that was the update's doing or not.

RAM was unstable at 3600. Given that that's an overclock rating and the default otherwise is 2733 or something I shouldn't expect miracles. It managed to boot an Ubuntu live USB but things went out to lunch pretty fast once I tried to untar a stage3 so I lowered it to 3200. atop tells me I'm still getting far above 1.0 IPC under load, so the RAM isn't a significant bottleneck.

I can't get any temperature readings at all besides a single k10temp value, and an ACPI thermal zone that gives invalid readings (dmesg complains too). I'm pretty sure it has a bog standard nct* chip on the motherboard because the BIOS offers to let me tweak 5 temperature trip points per fan, but sensors-detect finds nothing. There's no loss of functionality but being unable to tweak values under a real workload is super annoying.

CPU idles at 34°C as I'm typing this and maxes out just below 75°C. With single-threaded CPU-bound stuff (waiting for the emerge spinner) it hovers around 50°C and atop shows individual cores getting up to 4.4-4.5GHz.

It makes extremely short work of nightmare packages, as I'd hoped:

Code: Select all

dev-lang/rust: 13′46″ average for 1 merge
net-libs/webkit-gtk: 22′17″ average for 1 merge
sys-devel/clang: 11′08″ average for 2 merges
sys-devel/gcc: 25′05″ average for 6 merges
sys-devel/llvm: 10′20″ average for 2 merges
www-client/firefox: 14′47″ average for 1 merge

Overall, very boring after the initial setup. Which was exactly what I wanted.

And to hell with the UFO lights on that CPU fan! eix -R app-misc/cm-rgb if you just want it to go away.

[Anon-E-moose]

And to hell with the UFO lights on that CPU fan! eix -R app-misc/cm-rgb if you just want it to go away.

Tried emerging but it didn't like something.



I bit the bullet the few days.
Said farewell to the M5a99fx/fx8320/32g mem
and replaced it with PRIME x570-pro/3700x/64g mem.

I've set -j14 and it still compiles pretty fast leaving me with 2 other cores to play with.

Just minor probs, same as saellaven earlier, no cpu temp reporting.

I have been using the MuQSS scheduler for a while, but when I try to boot, it stalls at that point, haven't looked at it in detail yet.

[bammbamm808]

And to hell with the UFO lights on that CPU fan! eix -R app-misc/cm-rgb if you just want it to go away.

Tried emerging but it didn't like something.



I bit the bullet the few days.
Said farewell to the M5a99fx/fx8320/32g mem
and replaced it with PRIME x570-pro/3700x/64g mem.

I've set -j14 and it still compiles pretty fast leaving me with 2 other cores to play with.

Just minor probs, same as saellaven earlier, no cpu temp reporting.

I have been using the MuQSS scheduler for a while, but when I try to boot, it stalls at that point, haven't looked at it in detail yet.

As far as CPU temperature reporting, the out-of-tree zenpower module works for my 2700x/x470 Taichi. It doesn't do fans, but I've set curves in the UEFI. I think 5.6's k10temp does everything that zenpower does for me, but I'm fine with 5.4 for now.

Code: Select all

sensors
zenpower-pci-00c3
Adapter: PCI adapter
SVI2_Core:    +0.76 V  
SVI2_SoC:     +1.12 V  
Tdie:         +43.5°C  (high = +95.0°C)
Tctl:         +53.5°C  
SVI2_P_Core:   5.51 W  
SVI2_P_SoC:   12.99 W  
SVI2_C_Core:  +7.27 A  
SVI2_C_SoC:  +11.54 A  

[Ant P.]

And to hell with the UFO lights on that CPU fan! eix -R app-misc/cm-rgb if you just want it to go away.

Tried emerging but it didn't like something.

Sorry to hear that. emerge problem or does it do weird things at runtime? I've seen some of the latter myself, the ebuild could do with some improvement.

Just minor probs, same as saellaven earlier, no cpu temp reporting.

Kernel 5.6 fixes that. I haven't been able to get 5.6 to boot on mine yet, it hangs after refind. I applied the 5.6 hwmon patches to my 5.5 tree and it works fine there though. Weirdly there's no "maximum" temperature in the sensor output any more, there was with the old one.

Turns out my mainboard sensors are an IT8688, which is never going to be supported because Gigabyte fired all their good engineers or something. (That'd explain all the other BIOS headaches I've been having with the board. Guess I won't be buying hardware from these idiots again.)

I have been using the MuQSS scheduler for a while, but when I try to boot, it stalls at that point, haven't looked at it in detail yet.

Had a similar problem here. It works when there's sufficient load, but the system goes comatose when cores are idle and there's no source of external interrupts. This isn't the only CPU I've had those symptoms on, but it's far more pronounced when there's so many cores to go idle...

[Chiitoo]

Indeed, there's a bit more now with gentoo-sources 5.6.2 and ASUS PRIME X370-A (used to have just Tdie and Tctl I believe):

Code: Select all

k10temp-pci-00c3
Adapter: PCI adapter
Vcore:        +1.04 V
Vsoc:         +0.92 V
Tdie:         +40.6°C
Tctl:         +40.6°C
Icore:       +20.00 A
Isoc:         +5.50 A

I've unfortunately not been able to use ck-sources since 2018 or so, looking at the old list of configs I have laying around. Last I tried 5.4.11, and still got issues (boots okay, but hangs random'ishly).

Not sure if it all first started with the Ryzen or not. Been meaning to bring that up in the ck-sources topic, but eh...

[Anon-E-moose]

And to hell with the UFO lights on that CPU fan! eix -R app-misc/cm-rgb if you just want it to go away.

Tried emerging but it didn't like something.

Sorry to hear that. emerge problem or does it do weird things at runtime? I've seen some of the latter myself, the ebuild could do with some improvement.

I'm sure it's a minor thing, I had just downloaded the stuff and tried from scratch and it complained about corruption in ebuild, I'll try again later today.

Just minor probs, same as saellaven earlier, no cpu temp reporting.

Kernel 5.6 fixes that. I haven't been able to get 5.6 to boot on mine yet, it hangs after refind. I applied the 5.6 hwmon patches to my 5.5 tree and it works fine there though. Weirdly there's no "maximum" temperature in the sensor output any more, there was with the old one.

Interesting, I'll probably look at doing that.

I'm also looking at zenpower/zenmonitor, though I just downloaded it last night.

Overall I love the 3700x and the increase in power and throughput, though I only push a -j14 (I like to leave some cores for other stuff)

A pleasant surprise (April 4 dates are ryzen 3700x, March 11 was fs8320)

Code: Select all

     Mon Mar 11 06:53:06 2019 >>> sys-devel/gcc-8.2.0-r6
       merge time: 41 minutes and 23 seconds.

     Sat Apr  4 12:01:34 2020 >>> sys-devel/gcc-8.2.0-r6
       merge time: 15 minutes and 7 seconds.

     Sat Apr  4 12:23:06 2020 >>> sys-devel/gcc-9.2.0-r2
       merge time: 18 minutes and 22 seconds.

Had a similar problem here. It works when there's sufficient load, but the system goes comatose when cores are idle and there's no source of external interrupts. This isn't the only CPU I've had those symptoms on, but it's far more pronounced when there's so many cores to go idle...

Ah, something for me to start looking at, might be just a minor thing.

Thanks Ant P.

[Anon-E-moose]

About the MuQSS hang, I just googled a little, I'll play with it later to see if it works.

http://ck-hack.blogspot.com/2017/11/run ... muqss.html

Gets stuck booting the kernel when running with RQSHARE_MC on my AMD Phenom X6.

Setting it to RQSHARE_NONE boots just fine.

Seems to typically stop right around PCI initialization or after the following line:
NOHZ: local_softirq_pending 02

...

similar issue encountred: MuQSS 0.170, core i7 2760qm

setting notrheadirqs boots just fine with whatever rqshare=smp,smt,none setting, if either SMT/HT is enabled or not in BIOS.

threadirqs locks up boot if HT is disabled in BIOS and rqshare=smp, with rqshare=none threadirqs does not lock up boot. conclusion is that threadirqs conflicts with any type of rqsuare=smp|smt|mc

i've encountered a similar issue with a NUMA machine (dual-socket xeon x5670), but didn't track down wether it is related to threadirqs for sure, but i suspect it is because boot lockups were observed too. otherwise the following setup boots just fine with NUMA awareness: rqshare=smp and properly shared runqueue per socket, HT disabled in BIOS, threadirqs enabled but boot lockup was observed with HT=enabled in BIOS, rqshare=smp and threadirqs

and from the admin-guide/kernel-parameters.txt

Code: Select all

    rqshare=    [X86] Select the MuQSS scheduler runqueue sharing type.
            Format: <string>
            smt -- Share SMT (hyperthread) sibling runqueues
            mc -- Share MC (multicore) sibling runqueues
            smp -- Share SMP runqueues
            none -- So not share any runqueues
            Default value is mc

[Anon-E-moose]

And to hell with the UFO lights on that CPU fan! eix -R app-misc/cm-rgb if you just want it to go away.

Tried emerging but it didn't like something.

Sorry to hear that. emerge problem or does it do weird things at runtime? I've seen some of the latter myself, the ebuild could do with some improvement.

A bit more than some, IMO.

Once I figured out what it was trying to do and what other packages needed to be installed to make it work.

Bottom line, I've turned off the annoying wraith rgb.

Thank you, thank you, thank you for the original comment

[Makersmarx]

Ive been running GA-AB350-Gaming (rev. 1.x) with the Ryzen 1600 since it came out without any issues (even overclocked a bit to 3.8GHZ with the stock cooler this entire time). I just upgraded the CPU yesterday with a 3600 without issues (after upgrading BIOS). Only real difference I notice is my temps jump UP DOWN UP DOWN very often. Could be the thermal paste or as I see all over the place its what they do, but my 1600 seemed much more stable temp wise (They cooler seemed to be built better as well with the 1600). Not that this cpu is throwing anything extreme temp wise, just jumping up and down a lot unless idle. I'm running 5.6.2 Kernel

[fcl]

Ive been running GA-AB350-Gaming (rev. 1.x) with the Ryzen 1600 since it came out without any issues (even overclocked a bit to 3.8GHZ with the stock cooler this entire time). I just upgraded the CPU yesterday with a 3600 without issues (after upgrading BIOS). Only real difference I notice is my temps jump UP DOWN UP DOWN very often. Could be the thermal paste or as I see all over the place its what they do, but my 1600 seemed much more stable temp wise (They cooler seemed to be built better as well with the 1600). Not that this cpu is throwing anything extreme temp wise, just jumping up and down a lot unless idle. I'm running 5.6.2 Kernel

I doubt changing the thermal paste will help much. Ryzen CPUs, especially the Zen 2 ones* have those temperature spikes**, that often cause annoying fan sound. I'd recommend you to change you fan settings in the BIOS. As an example I've set my CPU fan speeds as follows: 35% until 55C, 55% 72C, 80% 82C. Use those values only as a general guideline but adjust as needed for your cooler etc. Do the same for the case fans. Many motherboards also have a "fan smoothing time" (may be named differently). Increasing it can help. Of course keep an eye on the CPU temperature after changing these settings, things can get toasty especially when compiling heavy stuff with all 12 threads.

* because of small 7nm manifacturing
** "ryzen sawtooth" search shows interesting results

[Anon-E-moose]

New 3700x and haven't noticed any temp spikes, of course I went through the bios to set things my way.

Temps stay moderate, upper 30's at idle and it ran to 61 degrees while compiling gcc (9.3) (14/16 cores used)

I always turn off what used to be called turbo core/cool'n'quiet as it will boost your core speed, by pushing more voltage which translates to more heat.

As far as the fan, I don't notice it (wraith prism), but I've set 25C - 20%, 45C - 50%, 65 - 100%

[Ant P.]

That more or less matches the temps I've had. I've been running folding@home 24/7 on my 3900X (with boost off, it doesn't help at all), so it stays more or less at 75°C… idles at ~34 otherwise.

[Makersmarx]

Ive been running GA-AB350-Gaming (rev. 1.x) with the Ryzen 1600 since it came out without any issues (even overclocked a bit to 3.8GHZ with the stock cooler this entire time). I just upgraded the CPU yesterday with a 3600 without issues (after upgrading BIOS). Only real difference I notice is my temps jump UP DOWN UP DOWN very often. Could be the thermal paste or as I see all over the place its what they do, but my 1600 seemed much more stable temp wise (They cooler seemed to be built better as well with the 1600). Not that this cpu is throwing anything extreme temp wise, just jumping up and down a lot unless idle. I'm running 5.6.2 Kernel

I doubt changing the thermal paste will help much. Ryzen CPUs, especially the Zen 2 ones* have those temperature spikes**, that often cause annoying fan sound. I'd recommend you to change you fan settings in the BIOS. As an example I've set my CPU fan speeds as follows: 35% until 55C, 55% 72C, 80% 82C. Use those values only as a general guideline but adjust as needed for your cooler etc. Do the same for the case fans. Many motherboards also have a "fan smoothing time" (may be named differently). Increasing it can help. Of course keep an eye on the CPU temperature after changing these settings, things can get toasty especially when compiling heavy stuff with all 12 threads.

* because of small 7nm manifacturing
** "ryzen sawtooth" search shows interesting results

Agree, I have a fan curve I have set in bios and haven't had any annoying fan sounds as of yet. Was just noting the differences in "spikes" vs my 1600. Ex...At idle I am 34-37 open a browser it will spike to 47 or so then drop back down. I didn't notice these changes with my 1600, but as you mentiond it appears to be the Zen 2 changes.

I just installed this 3600, has anyone messed with minor OC'ing with the stock cooler? I was successful with my 1600 and figured I would mess with this 3600 soon.

[Anon-E-moose]

If you're running with Performance Core Boost (bios setting), you might disable it and see if your temp surges don't stabilize.

As far as the difference between Ryzen and previous gen chips, it might be more the lack of reporting in previous gens vs a real difference.

OC'ing should be fine with the stock cooler, with a little moderation in mind. And be aware that with OC comes increased overall heat, not just the cpu itself.

[Makersmarx]

If you're running with Performance Core Boost (bios setting), you might disable it and see if your temp surges don't stabilize.

As far as the difference between Ryzen and previous gen chips, it might be more the lack of reporting in previous gens vs a real difference.

OC'ing should be fine with the stock cooler, with a little moderation in mind. And be aware that with OC comes increased overall heat, not just the cpu itself.

Thanks for the idea, I will try disabling Performance Core Boost this afternoon to see if it stabilizes. I had great success in OC'ing the 1600 so will report back on this 3600 once I am able to tinker some more with it.


Edit: Can confirm PCB being off does stabilize temp spikes currently running at a constant 36.

[sdauth]

Don't want to create a thread for that so, can anyone using a Ryzen 3950X or similar tell me how long it takes to compile ungoogled-chromium on that beast ? (With thinlto, optimize-lto, optmize-webui USE flags on. )

[Ant P.]

Here's a neat trick:

Code: Select all

make -C /usr/src/linux/tools/power/x86/turbostat/ BUILD_OUTPUT=$PWD turbostat

That tool will give you per-core power measurements in a format like iostat. Much more detailed than sys-power/cpupower.

[saellaven]

Don't want to create a thread for that so, can anyone using a Ryzen 3950X or similar tell me how long it takes to compile ungoogled-chromium on that beast ? (With thinlto, optimize-lto, optmize-webui USE flags on. )

3700x with 32GB

Code: Select all

[ebuild  N     ] www-client/ungoogled-chromium-81.0.4044.138_p1::pf4public  USE="clang cups kerberos proprietary-codecs system-ffmpeg system-harfbuzz system-icu system-jsoncpp system-libevent system-libvpx system-openh264 tcmalloc vdpau -cfi -closure-compile -convert-dict -custom-cflags -enable-driver -gnome -hangouts -optimize-thinlto -optimize-webui -ozone -pulseaudio (-selinux) -suid -system-openjpeg -thinlto -vaapi -wayland -widevine" L10N="-am -ar -bg -bn -ca -cs -da -de -el -en-GB -es -es-419 -et -fa -fi -fil -fr -gu -he -hi -hr -hu -id -it -ja -kn -ko -lt -lv -ml -mr -ms -nb -nl -pl -pt-BR -pt-PT -ro -ru -sk -sl -sr -sv -sw -ta -te -th -tr -uk -vi -zh-CN -zh-TW" PYTHON_TARGETS="python2_7 python3_7 -python3_6 (-python3_8)" 768,203 KiB


 Sat May 23 04:36:17 2020 >>> www-client/ungoogled-chromium-81.0.4044.138_p1
       merge time: 2 hours, 30 minutes and 24 seconds.

So, I would expect a 3950X to be roughly twice as fast given it has double the cores/threads.

[sdauth]

Thanks saellaven.
So still more than 2 hours on a recent cpu without any lto (-optimize-thinlto -optimize-webui -thinlto).. wow.. that chromium is a huge bloat. I was curious so I tried on my old AMD (A10 6700 / 32GB, tmp/portage in ram) It took 14 hours with all lto enabled.

[Ant P.]

I used to run Chromium on my Atom netbook. Back then it was reasonable enough I could build it on the machine in only 1GB of RAM, and the compile time was similar to the above (with a bit of distcc).

But last time I checked, 3 or 4 years ago, it took 78 hours. I haven't gone back to it since. At the rate things are progressing it'll be just as bad on a heavy desktop soon.

[poe_1957]

My complete specs are AMD Radeon 5700XT Challenger and Ryzen 9 3900X WRAITH (Stargate?).
In the names i see a lot Mantisse and Stargate )
I had in the past very good experiences with Asrock and AMD processors. Unlike MSI, and Gigabyte (none anymore in my possesion due to failures in MB. ) So now I have ONLY ASROCK.
The new board is the second one, due the first was DOA. But MSI had even worse and they do not live as long as the ASROCK board.
I have still to configure the Radeon GPU together with the LG 4K 27" screen. I have found the drivers but not yet a good X configuration. (If anyone has this you may post it I tried xrandr but it fails with no display.
The system is VERY fast : compiling libreoffice in 10min. Total price was =< 2000€
I heard that the Asrock is the AMD reference MB. For Linux i recommend it.

[poe_1957]

@ sdauth
I cannot speak about Chrome but firefox and Libreoffice (10min) are VERY fast on Ryzen 9 12 cores (1TB M2 and 32GB RAM) so it will even faster on 16 real cores. It takes more time to configure it for compiling than the compiling time itself. I used to compile libreoffice 2 days on my old intel core i5 Lenovo laptop with ssd.
So if you can afford it go for it ! it is faster than any intel processor I know. And no below zero temp cooling as with Intel. It runs at 60°C But do not forget to update your BIOS. The build bios is faulty on ALL motherboards: so UPDATE!

[Tony0945]

Torn between the ASUS Prime X570-Pro and the ASUS AM4 TUF GAMING X570-Plus
The gaming board has more features but is $100 cheaper, but I look at that all black board and can't see much. Bad cataracts. It would be a bear finding were to plug things in. I have an older all black board but it has red connectors. Easy to find.
Anyone owning either of these care to comment?

[saellaven]

Torn between the ASUS Prime X570-Pro and the ASUS AM4 TUF GAMING X570-Plus
The gaming board has more features but is $100 cheaper, but I look at that all black board and can't see much. Bad cataracts. It would be a bear finding were to plug things in. I have an older all black board but it has red connectors. Easy to find.
Anyone owning either of these care to comment?

My Asus Prime X570-Pro has been rock solid since I got it at launch... and despite the cheaper price, it has more room for expansion than the TUF board, though the TUF board has some more built-ins (wifi, bluetooth, etc).