Ryzen issues with Linux

GiGaBiTe said:
It did that automatically on the Gigabyte board when the BIOS was flashed, haven't changed the BIOS on the Asus board since it was already the latest.
Click to expand...

Do it via jumper. I am telling you this because it worked for me before.
 
thebufenator said:
That is really bad luck you are having there.

I have had a pretty decent experience with my Ryzen build, but I only run Linux in virtualbox. I am tempted to install Ubuntu alongside Windows to see if it has any issues....
Click to expand...
Would be interested if you do. :) What are your hardware specs? I also would try comparing Ubuntu 18.04 with Fedora or OpenSUSE Tumbleweed - or a comparison of the default Ubuntu with a newer kernel. That is what I will do if I decide to go with a Ryzen build - although, I might just test on Ubuntu 18.04, at first.

The reason(ing)?: some people claim that the 'no problems/issues' occur with distros using more recent kernels - 4.18.8 or later.

Edit: With up-to-date/most recent BIOS version, of course.
 
GiGaBiTe said:
Here are my settings from the "AMD CBS" page in UEFI setup:

Core performance Boost - Disable
Memory Interleaving - None
IOMMU - Auto
Global C-State Control - Enabled
Power Supply Idle Control - Typical Current Idle
Opcache Control - Disabled
Click to expand...
I'd maintain that disabling global C-state control is worth a shot, but it's moot if you can't get the memory stable. :/

Any chance one of those PSUs is bad and has been damaging other components or something along those lines? Given the RAM you're using now was in the GB board for a while, could something broken in the GB board's power delivery or config have been damaging RAM? I've had a bad mobo kill multiple CPUs before, though it took a lot longer than that.

AFAICT all AM4 boards have some kind of issue with a CMOS clear not clearing everything it should, but purports are that Gigabyte AM4 boards are the only ones that have janky enough firmware to turn it into a real problem. On ASRock, all it really means for me is that when OCing RAM it takes some time at 2133 after a fail and clear before it'll work at 2666+ again. If some RAM drive setting got corrupted on the GB board though, who knows what kind of havoc it could wreak.

The right way to CMOS clear on AM4 AFAICT is to disconnect power, move the jumper and/or disconnect the battery, push the power button a few times, wait 15 minutes, then move the jumper and/or reconnect the battery.
 
pavel said:
Would be interested if you do. :) What are your hardware specs? I also would try comparing Ubuntu 18.04 with Fedora or OpenSUSE Tumbleweed - or a comparison of the default Ubuntu with a newer kernel. That is what I will do if I decide to go with a Ryzen build - although, I might just test on Ubuntu 18.04, at first.

The reason(ing)?: some people claim that the 'no problems/issues' occur with distros using more recent kernels - 4.18.8 or later.

Edit: With up-to-date/most recent BIOS version, of course.
Click to expand...

I've been running Arch with this R7 1700 for a year or so, with no problems of that sort. Running kernels from before Zen's release seems likely to be an issue, but how common are those at this point anyway?
 
xorbe said:
g jedec spec Crucial DDR4-2133 CL15 ram? I have a set of that ram, because the first fancy ram I bought just wouldn't fscking work reliably. I also bought some DDR4-3600 CL15 that I'll be throwing in the trash since it's past return date, and doesn't seem to work no matter what settings or how slow it runs. Ev
Click to expand...
GiGaBiTe said:
So I re-installed the Patriot memory to do more testing and decided on memtesting it again just to be sure because of all the other ridiculous problems I've been having.

Sure enough, these modules went bad again. The initial 8 hour memtest when I first got the replacement modules a couple weeks ago passed with no errors.

View attachment 107150

This is the second set of memory modules I've gotten consecutively which were bad. They're not as bad as the memory in the OP video, but still bad nonetheless.

So far I've had a bad PSU, bad board, bad cpu and four bad memory modules all on a new build. maybe I should go back to buying sketchy ebay parts from Chinese import sellers, I seem to have better luck with those..
Click to expand...

I doubt you had that many out-of-the-box bad parts. The statistics on that are unreal. However you probably had one part cause others to fail and chain events after.
 
infinity9 said:
I doubt you had that many out-of-the-box bad parts. The statistics on that are unreal. However you probably had one part cause others to fail and chain events after.
Click to expand...

I've had a faulty PSU screw every single component that I was installing on a build... Capacitors were balooning like I would if I could eat all the McDonalds I crave. LOL.
 
cdr_74_premium said:
Do it via jumper. I am telling you this because it worked for me before.
Click to expand...

I don't have the Gigabyte board anymore, it went off to RMA on Monday.

ZyzzyxSilver said:
I'd maintain that disabling global C-state control is worth a shot, but it's moot if you can't get the memory stable. :/
Click to expand...

The Asus board was 100% stable with those settings and the generic Altex RAM. I know with the bad Patriot RAM it was going to fail again so I didn't continue after memtest showed it was bad. I started an RMA with Patriot and they accepted it, so I have to pack up the modules and ship them off tomorrow.

ZyzzyxSilver said:
Any chance one of those PSUs is bad and has been damaging other components or something along those lines? Given the RAM you're using now was in the GB board for a while, could something broken in the GB board's power delivery or config have been damaging RAM? I've had a bad mobo kill multiple CPUs before, though it took a lot longer than that.
Click to expand...

I tore the bad PSU apart and it failed due to shitty capacitors, can't believe reputable manufacturers are still using trash Taiwanese caps today, it's like they never learned from the capacitor plague. The Antec had OST, Teapo and UCC caps. I pulled a couple of the secondary side filter caps and they had off the chart ESR. Too lazy to recap the thing so it will sit on a shelf collecting dust until I get around to it.

I've never had a duff PSU kill any of my gear, and I've had some pretty catastrophic PSU failures. Had a couple of PSUs explode and one catch on fire, requiring a fire extinguisher. I have seen some shitty no-brand PSUs kill customer equipment though.

ZyzzyxSilver said:
]AFAICT all AM4 boards have some kind of issue with a CMOS clear not clearing everything it should, but purports are that Gigabyte AM4 boards are the only ones that have janky enough firmware to turn it into a real problem. On ASRock, all it really means for me is that when OCing RAM it takes some time at 2133 after a fail and clear before it'll work at 2666+ again. If some RAM drive setting got corrupted on the GB board though, who knows what kind of havoc it could wreak.
Click to expand...

I've only had two motherboards kill RAM twice in the past 20 years. The first was an AOpen AK77 Pro-A and the second was a P4P800-E Deluxe, in both cases the memory contacts and the RAM slot burned. I repaired the P4P800-E Deluxe by soldering in a new RAM slot, which was a royal pain and took hours.

ZyzzyxSilver said:
The right way to CMOS clear on AM4 AFAICT is to disconnect power, move the jumper and/or disconnect the battery, push the power button a few times, wait 15 minutes, then move the jumper and/or reconnect the battery.
Click to expand...

I know how to clear the CMOS memory lmao, I've been a system builder for nearly 25 years.

infinity9 said:
I doubt you had that many out-of-the-box bad parts. The statistics on that are unreal. However you probably had one part cause others to fail and chain events after.
Click to expand...

Nope, all of the parts that were bad were indeed bad. I had multiples of everything to rule the bad out.

The only other event of failure on this scale was about 10 years ago when 7 sticks of Kingston RAM from 3 different machines all went bad at the same time.
 
  • Like
Reactions: Nobu
like this
So, that sleek ultrabook I bought a few months ago is going to be stuck on windows forever? :(

I fucked up real bad not checking the Linux situation before pulled the trigger. It wasn't enough that Ubuntu no longer ships with make which absolutely drove me crazy - it had no wifi drivers but also provided no way of compiling the ones you downloaded from the internet on another computer. Thank god for Mint, yeah? But then the random freezes came and I after some reading I had no choice but to give it to the woman with windows 10 on it. This thread is really depressing
 
dgz said:
Ubuntu no longer ships with make [...] no way of compiling
Click to expand...

I'm pretty sure you can install 'make' within Ubuntu ... for instance, in openSUSE one would type 'sudo zypper install make' and it'd ask you for the root pw and install make and any dependencies automatically. For Ubuntu, looks like it might be 'sudo apt-get install make'.
 
xorbe said:
I'm pretty sure you can install 'make' within Ubuntu ... for instance, in openSUSE one would type 'sudo zypper install make' and it'd ask you for the root pw and install make and any dependencies automatically. For Ubuntu, looks like it might be 'sudo apt-get install make'.
Click to expand...
If you install the "build-essential" package (in ubuntu main repo), it pulls in make and a few other packages most compile/builds require. You can also use 'apt-get build-dep [package]' (or somesuch) if you're building something that already has a package in the repository, although that may miss some deps if your source is significantly newer or older than the package version.

Edit: of course, that doesn't really help if you have no internet connection. There is some way to download all the packages (that you want, not necessarily the whole repo) to the current directory instead of installing them, but it's been so long I can't remember anymore.
 
Oh I see. This is why I always make sure to buy laptops with an ethernet port.
 
xorbe said:
Oh I see. This is why I always make sure to buy laptops with an ethernet port.
Click to expand...
Many laptops now don't have an ethernet port....at least, the 'ultra books' or whatever they're called. Why doesn't that guy have an internet connection? Pretty difficult to use any Linux distro without one.
 
pavel said:
Many laptops now don't have an ethernet port....at least, the 'ultra books' or whatever they're called. Why doesn't that guy have an internet connection? Pretty difficult to use any Linux distro without one.
Click to expand...

The chicken-and-egg problem of wifi-only, but needing to compile that wifi driver, but pita sneaker-net required to get gcc toolchain installed apparently.
 
xorbe said:
The chicken-and-egg problem of wifi-only, but needing to compile that wifi driver, but pita sneaker-net required to get gcc toolchain installed apparently.
Click to expand...
Wait, what's the chipset of your laptop?
 
GiGaBiTe said:
Get one of those generic USB Ethernet dongles with a Realtek chipset.
Click to expand...
There's also the wifi 'devi' webpage that provides info of which hardware (chipset) works with Linux. Of course, it's best to have a very recent kernel - some of the firmware is built in the kernel - it's best to get a USB wireless adapter with one of those - I think Realtek and Atheros are two in which there is usually decent support for wifi adapters.

https://wikidevi.com/wiki/Wireless_adapters/Chipset_table
https://wireless.wiki.kernel.org/en/users/drivers
https://willhaley.com/blog/finding-linux-wifi-usb/
 
GiGaBiTe said:
Probably has a broadcom based WLAN, which never has Linux drivers.
Click to expand...
I remember reading about broadcom in Linux, previously, but people were always having trouble with them. There were articles in which Broadcom was announcing new 'Linux support' and 'open sourcing' drivers etc. etc. but even now, there is very little to read about success regarding Broadcom drivers (working) in Linux. I think Broadcom is the major hardware/drivers in recent Mac computers?
 
Broadcom is perpetually a mess, yeah. :(

GiGaBiTe said:
I've only had two motherboards kill RAM twice in the past 20 years. The first was an AOpen AK77 Pro-A and the second was a P4P800-E Deluxe, in both cases the memory contacts and the RAM slot burned. I repaired the P4P800-E Deluxe by soldering in a new RAM slot, which was a royal pain and took hours.
Click to expand...
Mechanical failures of the slot? Good thing that one isn't common.

GiGaBiTe said:
I know how to clear the CMOS memory lmao, I've been a system builder for nearly 25 years.
Click to expand...
I just mentioned it because pretty much everyone (including AMD and mobo manufacturers) seems to leave out the "push the power button a few times" bit. I haven't messed with RAM clocks and needed a clear since I heard of that fix, but the standard procedure definitely misses some stuff on AM4 boards.
 
pavel said:
I remember reading about broadcom in Linux, previously, but people were always having trouble with them. There were articles in which Broadcom was announcing new 'Linux support' and 'open sourcing' drivers etc. etc. but even now, there is very little to read about success regarding Broadcom drivers (working) in Linux. I think Broadcom is the major hardware/drivers in recent Mac computers?
Click to expand...

Broadcom may have made open source/Linux drivers, but that's only half of the necessary software to get their chips running under Linux. The other part that's needed is the firmware, which is not open source and you have to jump through hoops to get it and more hoops to get it working.

It's why NDISWrapper exists, because people found it easier to emulate Windows driver calls and use Windows drivers.

ZyzzyxSilver said:
Mechanical failures of the slot? Good thing that one isn't common.
Click to expand...

Pins burned out of the slot, it's more common than you think. If a RAM chip or capacitor fails short, it can burn a power or ground pin up in seconds.
 
Any update? I noticed a pic of a 1500X - are you using the 1st gen Ryzen?
 
pavel said:
Any update? I noticed a pic of a 1500X - are you using the 1st gen Ryzen?
Click to expand...

It's a 1500x.

The machine had been stable with the generic 4 GB DDR4 module. I got the 8 GB Patriot DDR4-2666 sticks back from RMA (they sent me a new retail kit) and it's been stable the past two days so far. I made sure to run a memtest pass to see if there was any obviously bad bits and it didn't find anything.
 
You must log in or register to reply here.
Back
Top