I have a Ryzen 3 1300X at the moment and it’s always had this soft lock freezing bug on Linux. I used to dual-boot Windows on this machine and Windows never had the same problem, so I think it is an issue with the Linux kernel (I’ve also replaced nearly every bit of hardware that I originally built the PC with, except for the CPU and motherboard, so it probably is an issue the kernel has with my CPU, or possibly the motherboard firmware).
I’ve changed the kernel parameters as suggested by the Arch Wiki. The bug is pretty inconsistent about happening so only time will tell if this solves the issue. But if it doesn’t solve the issue, I’d honestly consider just getting a new CPU that doesn’t have this issue, as completely freezing up, unable to get to a tty or anything, and only being able to power off by physically holding down the power button, is a pretty major issue, even if it only happens sometimes.
So if I do get a new CPU, or maybe just for when I’m next buying a CPU for reasons unrelated to this bug (been considering an upgrade to something that’s better for compiling anyway), are there any good options out there? Intel is investing $25 billion into Israel and the BNC has called for “divestment and exclusion” from it (it’s not officially on the BDS consumer boycott list, but I’m still very much not comfortable buying from Intel). But the Arch Wiki article seems to suggest this bug is applicable to Ryzen CPUs in general, or at least it never specifies a particular model or range of models. So maybe I’m limited to non-Ryzen AMD CPUs?
I’m guessing this is one of the situations where two companies have a complete duopoly over the market and there isn’t an all-round good solution, but thought I’d ask in case anyone had some useful input.
I think if you start with political positions of bigtech companies…
Just buy used
Buy Intel used so that you’re not directly contributing?
Other than that or AMD, your only other option is ARM.
Can you go and buy an ARM CPU and build a desktop system with it?
Check out the Ampere Altra Developer Platform
The OP is concerned about stability and you’re suggesting an experimental CPU that is plagued by UEFI bugs and is overly expensive? From what I’ve been even a cheap Chinese SBC with a Rockchip CPU is more stable and reliable than that thing.
Yes.
Not many, but they exist. I think most of them come soldered to the board like laptops.
ltt made a video recently-ish showcasing a multi threaded arm cpu desktop. Not sure how availabe that is to the market though.
I have a Zen 2, Zen 3+ and a Zen 4 system and they all work well very with various Linux distros (Arch, Fedora) and recent kernels.
It’s very likely that your bug is specific to early Ryzen CPUs/chipsets. A couple of folks on those reports mentioned their issues went away after a motherboard/BIOS upgrade. So I think you’ll be fine if you went for a more recent AMD CPU+mobo.
I’ve heard that newer Ryzens play nicer with Linux.
Yeah, my 5800X3D works perfectly; absolutely zero issues. I’m guessing it’s making use of the 3DvCache too since I don’t notice any performance degredations compared to Windows.
Yeah First Gen Ryzens definitely had a Linux lock up bug. My x1700 had it all the time and could never fix it.
I experienced that issue, AMD replaced it for free for me. Still rocking that R7 1700 to this day, still going to strong!
Do you know if it’s limited to first gen Ryzens? I’m looking into getting a Ryzen 5 5600X and I want to be sure I’m not gonna have the same issue
Yes, AMD was replacing Ryzens that had that bug. I’m not sure if they are anymore though. But it’s 100% a confirmed thing. I have not heard anything Zen 2 and newer having this problem and have no experienced any Linux issues with my 3000, 5000 and 7000 series CPUs.
Ah rip. I didn’t know they were replacing Ryzens. I’ll reach out to them but the warranty on my CPU is almost definitely voided after so long.
https://www.extremetech.com/computing/254750-amd-replaces-ryzen-cpus-users-affected-rare-linux-bug
They probably won’t replace it past warranty but it’s always worth a shot
My Ryzen 9 5900X has none of these issues.
I have a Ryzen 7 5800X and I’ve had no problems
I have a 5600h system(laptop) and I have not run into the issue you mention. In fact in the past six months after the fTPM bullshit was fixed, I haven’t run into any issues.
the fTPM bullshit
Wait, what’s that?
In zen 2 and above firmware TPM was being used for a random number generator. This led to stutters during RN generation. Eventually this was fixed in 6.3.x series(or 6.4.x I can’t remember) and then the fix was backported to all lts kernels
Never experienced or even heard of Steam Deck having that bug and that’s a somewhat recent Ryzen.
I have a 3600X and haven’t had any problems under Linux.
Can confirm (but mine is the regular trim rather than the X version).
It’s better value anyway. Only got the X because it had a rebate while the standard didn’t.
This is pointless. Your tax dollars are doing MUCH more for Israel than what products you buy. Boycotts are a capitalist distraction from the real systemic issues.
His point wasn’t to find a CPU, it was to make a political post in a tech community.
I agree. We need more of these!
Buhruh! Why not just stop voting since “your vote is only a drop in the ocean” or “it only legitimises a broken system”?
Every action towards progress counts. It’s better than nothing, which is what people do if you ask them to change the world in one go. Change is gradual, change is slow, change can be achieved by the small actions of many. Not everybody has the time to “tackle the systemic issues” you perceive to be true nor does everybody agree that those are the core issues.
Belittling action, no matter how small, is discouraging and counterproductive.
Voting is good because elections can be won by a few votes and ARE won by a few votes consistently. The tiny fraction off the top of your CPU purchase that might end up possibly supporting Israel is doing nothing compared to the hundreds of millions that, may I remind you, you cannot opt out of sending to Israel.
Changing what products you consoom is literally feel-good liberal shit to make you feel like you’re doing something. Talking to representatives and protesting is way more effective, in the sense that one of them does nothing and the other actually does something.
Changing what products you consoom is literally feel-good liberal shit to make you feel like you’re doing something.
This is why big companies continue making money, influencing politics, and can have more profits than some countries have budgets. Then people like you turn around and say “ermagerd, companies are destroying the world” with an iPhone in one hand and a venti in the other while wearing fast fashion. But at least you voted, amirite?
You are severely overestimating the amount of people who, to put it the shortest way I can think of rn, are not NPCs.
Yes, what you say could be true in a perfect world where everybody or even a majority of people participate in boycotts. I still believe it’s a distraction from capitalism’s structural problems but they could still be effective that way. The difference is that the world we live in is not that world and therefore small-scale boycotts like this do basically nothing. The average person does not care what companies support Israel. Also a lot of those lists are literally wrong but that’s a different issue. Also also, the government STILL sends a SHITTON of money to Israel whether you personally support it or not.
What if you live in a country that doesn’t send any money to Israel? Also, boycotts do work.
Not when there’s an organised boycott, called for by Palestinians. You can do multiple things at once. Not buying something takes 0 hours of your time lol
Not buying something takes 0 hours of your time lol
It takes so little time you needed to make a post to ask for help lol
Making a post takes a few mins of time. Not boycotting is taking so little of your time that you needed to make 2 comments about it, wow
Omg such a waste of time that could have been spent scrolling through memes instead of trying to do the right thing.
Making a post takes a minute. Responses will take less than a minute. They were crowdsourcing specific and quite niche knowledge. In that time between making the post and reading them after a wait OP was able to do literally anything else instead of searching and trying to pull out decent looking CPUs they couldn’t guarantee. If you were buying any CPU you would hopefully look for reviews or comparisons, BDS or not, to inform you.
This smells of reductionist “no ethical consumption under capitalism” ideology.
That just means living in capitalism doesn’t exempt you from criticising the system, not that you can’t and shouldn’t use the mechanism of capitalism to help make life difficult for fascists.
It might not “fix” the problems but it sure as hell is making Israel pay while our national governments do fuck all.
I have a system with a Ryzen 1700 with the same issue and have found the only reliable way to run it is by installing and enabling the disable-c6-systemd package from the AUR. The other fixes provided in the wiki article you linked are correct but aren’t sufficient on my system, the CPU keeps reenabling the C6 state on its own and the disable-c6-systemd package works to counter that. The reason it works on Windows is they’ve disabled the C6 state by default for the CPU.
This is amazing to find out now after 7 years:) I actually adjusted voltage manually on my Ryzen R5 1600, and it became 100% reliable, apparently the fix you mention prevent voltage below 1v at idle. I wondered why my CPU wasn’t reliable unless I made manual OC with some voltage tweaks?
I never looked it up, because my OC solved the issue, but I always thought it was a bit weird.
Ah, thanks. I’m using runit not systemd (although this was happening on systemd when I was on systemd too) but I saw
amd-disable-c6
in the AUR so I’ve installed that now, fingers crossed it works (the fixes in the Arch Wiki article haven’t fixed it for me, it just happened again rip)Edit: nvm, looks like that package is a systemd service
The package is just a systemd unit to run the command
python zenstates --c6-disable
so if you install the zenstates-git package and get runit to run that command at startup it would be equivalent.Thank you!!
Edit: Tried running that, I’m getting the error that
/dev/cpu/0/msr
doesn’t exist.dev/cpu
doesn’t seem to exist at all on my machine. HmEdit 2: You need to run
sudo modprobe msr
. All good now :)
Ah, that sounds a bit unfortunate. I’ve run AMD CPUs on Linux desktops with Bulldozer / Piledriver / Ryzen 7, my current laptop is a Ryzen 7 as well, never run into that at all. Hopefully the Arch wiki will sort you out. If not that, the third option would be ‘install Linux on an M-series Mac’ - don’t know how feasible it is at the moment, and paying the ‘Mac premium for hardware and software integration and then overwriting the software’ doesn’t make a lot of sense to me.
I have has zen2 and zen3 systems and haven’t run in to that either. So Zen2+ systems should be mostly fine.
If you get an M1 or M2 mac it should mostly work. If you need thunderbolt(WIP) or vulkan(WIP) then you will have to wait. Otherwise accelerated desktops work and audio is working now. Honestly if you compare performance to competing systems, they end up pretty similar in pricing.
Print your own
…in 10 years
Or use any of these
https://www.theregister.com/2024/03/25/china_approved_tech_list/
Risc v maybe? Rock 64?
A few more years until RISC-V is at 1st-gen ryzen levels (though it looks like RISC-V is accelerating every day)
Modern Ryzen are fine on Linux. Not sure where they’re declaring it happens for “all Ryzen” from.
Not sure where they’re declaring it happens for “all Ryzen” from.
I’ve not see anyone claim it happens on “all Ryzen”, just that the Arch Wiki article doesn’t specify a particular range or model
My Ryzen 5700u work great with Debian, so as others said, consider upgrade CPU on your am4 motherboard, better buy apu since it always feels good to have backup gpu in your system in case main gpu breaks
I don’t think getting an APU ‘just in case’ is a good idea. It limits your turbo frequency and halves your L3 cache compared to the equivalent CPU variants. It also limits you to PCIe 3.0 only. Some AM4 boards have a single 4.0 x16 slot for graphics cards, so getting an APU could directly affect the graphics performance from a discrete GPU. OP should get the chip that is more suited to their typical use case.
Holding out for the Oasis, myself. Cores licensed from SiFive and compliant even with the 1.0 rev of the Vector Extension (SG2380).
Did you pre-order?
Not yet but planning to at the weekend.
I had a similar problem with a Ryzen 5 1600, at first it worked flawlessly when I did a manual OC, then when I made a reset, I didn’t care to OC again, and it began to occasionally freeze, usually when idle. When I made my manual OC again including upping voltages a bit for CPU and RAM it worked flawlessly again???
Weird since motherboard defaults aught to be stable IMO, but apparently they aren’t always.
I’ve been using this CPU for 7 years now, and it still runs like a champ, as long as I don’t use the motherboard defaults. I can’t remember last time it crashed or froze.
EDIT:
I just found out the package “disable-c6-systemd” mentioned in your link, is exactly to prevent voltage drop below 1v at idle, which sounds exactly like the bug I had. which is funny to learn 7 years later. ;)
As I understand, early ryzen processors are generally more buggy. I run 5800x on my desktop and a 5600(x?) in my server. You could try a newer ryzen and see if it works. I would recommend shopping around for a decent warranty.
That’s good to hear that you’re not having problems on newer Ryzen. Although not sure if I want to risk buying a new Ryzen CPU if there’s a chance I could have the same problem
What kernel version are you seeing that lockup bug on? I have a similar bug on Ryzen 5 2600x with kernel versions >= 6.7. 6.6 is fine.
More directly: Buy used. Lots of reputable sellers on eBay and their returns policy for defective products is unbeatable.
hardened kernel v6.7.9, but I also had the same problem on the regular Linux kernel a while back. It’s been a while since I’ve used the regular Linux kernel though so it might have gotten fixed on the regular kernel but not the hardened version?