|
Volguus posted:never. /etc is 25 years old, like god intended. I've been juggling files forward for (almost? nearly?) twenty years now. Most of that has been between different Windows installs. My plan is to make a new framework into which I'll pour almost all what I've stuffed into my current install. Volguus posted:You may want to uninstall nvidia packages though. I am selling my 3070 to just use the ryzen internal gpu (it's all I need right now, CUDA days are over) and I had to uninstall nvidia rpms before I could get a graphical interface. Check for kernel params as well, may want to remove nvidia-drm.modeset=1 though it won't do anything after you removed the packages. Given work schedule, time spent on purchase decisions, and delivery times, I might start doing this next weekend. cruft posted:Well I never!
|
# ? May 27, 2023 17:21 |
|
|
# ? May 28, 2024 13:44 |
|
You know how if you fall asleep right after waking up you might get an intense dream for a few minutes before you wake up again? Well I just had one and in it Sun released a GPL version of ZFS and someone posted about it right here in sh/sc. So when I woke up I tried finding that post, until I remembered that Sun is and remains dead. What the gently caress.
|
# ? May 27, 2023 17:37 |
|
cruft posted:Well I never! me either, that's why my systems are always a mess E: well, I missed that joke entirely
|
# ? May 27, 2023 17:55 |
|
cruft posted:Sounds like you're going down the right path there, OP. Let us know how it works out It's working I checked the app source code and fortunately it already sets transfers=1, so no need for a patch/pr. The SFTP remote config *screen* doesn't support public key authentication, but the app can just import any arbitrary rclone config file. So I created a new private key on my desktop that I use only for this, checked that it worked via desktop rclone, and then just sent the generated config file to my phone and imported it into the app. Now there is no process running *at all* on my Pi, but whenever I open the app to browse my remote files *or* the app triggers a periodic sync, openssh spawns a rclone server instead of a shell. Perfect.
|
# ? May 28, 2023 01:27 |
|
How do you handle the clock setting on a system that dual boots windows and linux?
|
# ? May 29, 2023 12:44 |
|
Hardware clock to UTC in both operating systems and everything will be fine. Windows is the one with the non-UTC default
|
# ? May 29, 2023 12:59 |
|
Will that only affect what gets written to the hardware clock or will I only be able to see the UTC time in the OS as well?
|
# ? May 29, 2023 13:19 |
|
I would blow Dane Cook posted:Will that only affect what gets written to the hardware clock or will I only be able to see the UTC time in the OS as well? From my notes since I wrestled with this exact thing a few years ago: quote:On Linux: Set the time zone properly and it will show local time on the clock on both OS's. The fast startup isn't necessary for the clock, I needed it because Windows wasn't completely powering down some device or other (maybe it was sound or ethernet or something) and leaving it in some half awake state to supposedly improve startup speeds and Linux had no idea what to do with it so the device was dead after doing a reboot from Windows to Linux without a complete power cycle.
|
# ? May 29, 2023 14:00 |
|
yeah i've been caught out by that fast startup too. It doesn't kick in if you choose restart instead of shutdown. Thanks for the help.
|
# ? May 29, 2023 14:06 |
|
Do you have to disable setting the time automatically? That’s a good way for my clock to drift enough that it breaks TOTP.
|
# ? May 29, 2023 16:40 |
You absolutely do not need to disable NTP, and doing so breaks not just TOTP but also TLS. The thing you want to disable is the setting called "Set Time Zones Automatically" in Settings > Date & Time. Fun fact: At some point after the switch to NT, the kernel itself started tracking UTC - what the registry key does is force anything that interacts with timezoneapi.h programmatically to use UTC. EDIT: Also, didn't you build a 7800X3D system? Check if your motherboard has an external clock generator. It doesn't just improve overclocking, it also improves time tracking. Also, make sure you're not disabling HPET. BlankSystemDaemon fucked around with this message at 20:29 on May 29, 2023 |
|
# ? May 29, 2023 20:22 |
|
By balancing which parts were ordered on Newegg and which from Amazon, I now have a functioning desktop again. It'll be fully up to its new ludicrous speed next weekend, when I have the GPU. A brief recap: Whenever you do stuff with microcode packages, apparently you need to poke GRUB again. If, when booting, the fsck goes fine and then nothing happens, it's that the graphical desktop environment isn't starting. Remember that Alt-F2 can bring up another login. If X can't display anything, it's nVidia's fault. It's always nVidia's fault. But, revisiting: VictualSquid posted:If you have the time and mindspace right now you should probably use the opportunity to decruft your system by reinstalling it from nothing. Just copy over your home dir and selectively copy over the etc files for the programs you want to keep. VictualSquid posted:If you are going with minimal effort you can try just installing the amd packages and run pacman's tree verification tool. It probably will work, but might not. In the meanwhile, I have a new problem: Stuff crashes. Every now and then, one process or another will crash (and generally restart). Most irritating is when Discord does this, because I lose whatever message I was typing. Have also noticed it with Firefox tabs, OpenMW, the KDE System Monitor, and the ... what do you call the 'desktop' itself? Plasma? Specifically, it's that thing which I've just noticed has frozen (clicking on icons does nothing and my clock widget is stuck at 12:39:29 and not moving). I'm not sure if this is a consequence of this weird hardware swap I've done, or of Weird Stuff happening because of how I juggled software to accommodate that hardware swap, or (most concerningly) if I have, like, very slightly bad RAM and that's triggering this. (I've never had faulty RAM before and I'm not sure what it's like.)
|
# ? May 29, 2023 20:43 |
|
Vavrek posted:By balancing which parts were ordered on Newegg and which from Amazon, I now have a functioning desktop again. It'll be fully up to its new ludicrous speed next weekend, when I have the GPU. I don't know what you can use in arch, you need to look it up. Some command to reinstall all packages or at least check that all dependencies are OK. Something like emerge -e world in Gentoo. Reinstalling all packages like that might even fix your crashes.
|
# ? May 29, 2023 20:53 |
|
Vavrek posted:In the meanwhile, I have a new problem: Stuff crashes. Every now and then, one process or another will crash (and generally restart). Most irritating is when Discord does this, because I lose whatever message I was typing. Have also noticed it with Firefox tabs, OpenMW, the KDE System Monitor, and the ... what do you call the 'desktop' itself? Plasma? Specifically, it's that thing which I've just noticed has frozen (clicking on icons does nothing and my clock widget is stuck at 12:39:29 and not moving). This might or might not be bad RAM, running memtest86+ is easy from a USB and should let you figure out if the RAM is acting up.
|
# ? May 29, 2023 21:08 |
|
VictualSquid posted:I don't know what you can use in arch, you need to look it up. Some command to reinstall all packages or at least check that all dependencies are OK. I asked because it sounded like you were familiar with Arch and were referring to some specific thing. I'll look into it later, because: drk posted:This might or might not be bad RAM, running memtest86+ is easy from a USB and should let you figure out if the RAM is acting up. 35% of the way through a first pass and Memtest86+ has found 39 errors. poo poo. The CPU cooler blocks easy access to the RAM slots, so swapping them out for individual testing will be more annoying than usual. I was hoping this was just a software glitch I could tolerate for a week before starting with a new install.
|
# ? May 30, 2023 00:37 |
|
If you really don’t want to change the memory out… http://rick.vanrein.org/linux/badram/
|
# ? May 30, 2023 00:48 |
|
Doesn't basically all RAM have lifetime warranties? I've seen absolutely ancient stuff to G.SKILL and gotten warranty replacements. I've never had DDR3 or newer fail though, only DDR1 and 2.
|
# ? May 30, 2023 01:05 |
|
I haven't had entirely bad RAM, but I've definitely had DDR4 that failed at the speed it was advertised at when I used the XMP settings. Worked fine at a still-decent speed, so I just left it there and had no further problems.
|
# ? May 30, 2023 01:37 |
|
Subjunctive posted:If you really don’t want to change the memory out… That stirs some nostalgia. Also makes me remember remapping HDDs around bad sectors to wring the last life out of them. Can't remember offhand the utility I used for that, but I think it was on the Ultimate Boot CD.
|
# ? May 30, 2023 02:16 |
|
Computer viking posted:I haven't had entirely bad RAM, but I've definitely had DDR4 that failed at the speed it was advertised at when I used the XMP settings. Worked fine at a still-decent speed, so I just left it there and had no further problems. Yep. I've never had a problem with XMP settings before, but I've only used XMP settings once before. I think cramming four sticks in, and them being kept a bit warm under the CPU heatsink (it's just a crowded area), led to the problem. Memtest86+ eventually threw loads of errors across all the memory, so if it was a hardware defect, it would've been with either all four sticks or with the CPU or motherboard ... I turned off XMP settings, left it at the default, ran another test, and it's seemed fine. My plan is to slowly tweak the RAM settings upward, repeatedly testing to see if it throws errors or not. But I have another problem! I had this problem before, late last night as I was getting this running and all earlier today. I'm not even sure if it's a Linux problem, but it seems uncommon and weird enough that no search results show up. So. This is tricky to test out, but the behavior I've observed so far: Scenario 1: Power on computer. Do not touch keyboard. MSI logo flashes for something less than a second. GRUB loads, showing the standard three options (Arch Linux, Arch Linux with Options (I forget the exact naming), and Windows 10). There is a five second countdown timer and then Arch Linux is selected and boots normally. Scenario 2: Get to GRUB OS menu as above. Hit the Down arrow once. Highlighted selection moves all the way down to the third option, Windows 10. Pressing the Up arrow does not move the selection back up. CPU fans audibly ramp to full, suggesting to me that the CPU is under some kind of mysterious load or is at least heating up. Screen appears frozen, machine non-responsive. I get a mix of worried and annoyed and power down the machine. Scenario 3: Get to GRUB OS menu as above. Hit Enter to select Arch Linux. Screen goes black. The "Booting Arch Linux" and "Loading initial ramdisk" messages do not appear. CPU fans ramp to full as in Scenario 2. I haven't tried letting this just sit there for five minutes to see what happens, in part because Scenario 1 works. (But it's weird! And annoying!) Scenario 4: I try to get into the MSI BIOS by hitting DEL during boot and somehow mistime this and I get a similar sort of behavior of non-responsive, max CPU fan. (MSI's BIOS does not appear to have any option for how long the full-screen logo should be up. I'd like the option of setting it to 3 seconds. ) I'm not sure what's going on with Scenario 4 (I tested out 2 and 3 again, before writing this, just to be sure I was remembering the details correctly). There's a blinking underscore in the upper left during this process, which sometimes stops blinking, sometimes isn't there, that I can kind of use to gauge where I am in the timing of everything. But: This means that, in addition to having to wait an extra few seconds to load into Arch every time, I'm not able to boot Windows at all to test how functional it is. It further means that something weird is going on. Both of these things are distressing. And they're unrelated to the memory issues, I believe, as I was testing them out again after setting the memory clock back down to the tragically low stock value. edit: oh, I should include Scenario 5, which is that I boot to Memtest86+ on a USB stick and the keyboard is absolutely fine for selecting a boot option there. edit 2: if it wasn't for scenario 4 and this just being weird, I'd put all this as a (frustrating) oddity to deal with between now and when I actually set up a new bootloader on the new drive, finally move into the era of GPT anything, etc. (My system drive, which has root, boot, swap, and windows, is an MBR drive because I sort of ... carried forward the partition scheme from WinXP, just cloning it onto new drives and making adjustments.) Vavrek fucked around with this message at 06:15 on May 30, 2023 |
# ? May 30, 2023 06:01 |
|
4 sticks are often iffy when applying am XMP profile, tbh. It usually isn't even the ram's fault, but the memory controller becomes unable to keep up.
|
# ? May 30, 2023 06:11 |
|
Kibner posted:4 sticks are often iffy when applying am XMP profile, tbh. It usually isn't even the ram's fault, but the memory controller becomes unable to keep up. Yeah in retrospect I'm not that surprised. I was just lured in by the idea of having all the RAM in the world.
|
# ? May 30, 2023 06:16 |
|
Vavrek posted:Yeah in retrospect I'm not that surprised. I was just lured in by the idea of having all the RAM in the world. You don't happen to be running a wireless keyboard, do you? I actually had something that sounds similarish recently when I couldn't figure out why my mouse and keyboard were seeming intermittently unresponsive, and it carried across GRUB, Linux, and Windows. Turned out it was because I had a thumb drive plugged into the USB port right next to the wireless dongle for my mouse and keyboard, and once I removed the thumb drive everything went back to normal.
|
# ? May 30, 2023 07:00 |
|
Vavrek posted:Yep. I've never had a problem with XMP settings before, but I've only used XMP settings once before. A thing to know is that XMP is officially an Intel specification, and while AMD platforms will load the XMP values it's always been a bit of a crapshoot. Works for many basic setups, very iffy with 4+ sticks or bleeding edge hardware. (Also Intel's memory controller is objectively better than AMDs.) AMD has made their own thing for DDR5 so this will be less of a problem going forward. Vavrek posted:various issues If you haven't yet, you should look for bios updates for your mobo. This being the very tail end of AM4, and the 5800x3d being the final CPU released, you might have a board with only the first release of x3d support.
|
# ? May 30, 2023 16:28 |
|
I have never heard of XMP being iffy before and on the face of it, it seems like a load of hogwash? I guess it would make sense because my first ryzen machine (3900 + aurus master) in decades was having some serious issues where it would brick the mobo (had to pop the cmos battery out to fix it) and the fix was to swap memory with my wife (the memory is fine and has been working in hers for 3~ years, always tested fine in mine). That memory was even in the QVL list specifically because i didnt want bullshit issues. Don't you literally have to use XMP these days i thought it was a legacy x86 issue or something that no memory can go above like, 2100mhz (i forget the exact value) without xmp, are you telling me the entire memory market segment is built on a lie that XMP "only kind of" works? Where can i read about this, I thought it was basically a standard edit: to be clear im not saying anyones wrong this is just blowing my mind
|
# ? May 30, 2023 19:26 |
|
Mr. Crow posted:I have never heard of XMP being iffy before and on the face of it, it seems like a load of hogwash? As mentioned in the post before yours, XMP is an Intel standard and works (near?) flawlessly with Intel CPUs. However, as AMD memory controllers are not as good as Intel ones, XMP profiles "should" work but don't always. Often, voltages have to be increased a bit and/or timings loosened a bit. EXPO is the AMD equivalent and those profiles should work (near?) flawlessly with AMD CPUs. However, there aren't very many memory modules that have EXPO profiles; they nearly all just have XMP. e: doesn't exactly answer your question, but does explain EXPO: https://www.makeuseof.com/what-is-amd-expo/
|
# ? May 30, 2023 19:41 |
|
I didn't understand XMP profiles until I realized that the profiles are stored on the stick of RAM.
|
# ? May 30, 2023 20:16 |
|
Mr. Crow posted:Don't you literally have to use XMP these days i thought it was a legacy x86 issue or something that no memory can go above like, 2100mhz (i forget the exact value) without xmp, are you telling me the entire memory market segment is built on a lie that XMP "only kind of" works? Where can i read about this, I thought it was basically a standard You're thinking of the difference between what the JEDEC standards let you encode in SPD, which are generally expected to work in any platform that advertises support for the same standard, vs. XMP. XMP is a combination of "these settings aren't guaranteed to work on all systems, pair your RAM/CPU/MB carefully" and/or "we're off the map and there's no way in normal SPD to tell the system to run this fast." I started to write more but this is probably better than any rephrasing I can come up with: https://www.crucial.com/support/articles-faq-memory/what-is-xmp Most OEM systems and virtually all servers just use SPD settings, even today. Eletriarnation fucked around with this message at 20:30 on May 30, 2023 |
# ? May 30, 2023 20:26 |
|
Mr. Crow posted:I have never heard of XMP being iffy before and on the face of it, it seems like a load of hogwash? XMP is just a little bit of ROM on the ram stick with data for the clock/MT rate and timings. That's it. When you turn on XMP, the mobo bios reads that data and sets the numbers as it says. XMP is an Intel brand, and Intel validates it on Intel platforms. So it's never iffy for Intel. They *could* go on to make sure it worked on AMD with those numbers, but for a long time they didn't because AMD sucked. That's why everyone got told to pay attention to QVL lists. Once Ryzen started getting traction you'd see memory kits like "XMP 3600, Optimized for AMD" or "Ryzen Edition" or whatever. Now AMD has made EXPO with the aim of "validated on Ryzen", but they're making an open standard so memory makers can do it themselves. Mr. Crow posted:Don't you literally have to use XMP these days i thought it was a legacy x86 issue or something that no memory can go above like, 2100mhz (i forget the exact value) without xmp, are you telling me the entire memory market segment is built on a lie that XMP "only kind of" works? Where can i read about this, I thought it was basically a standard So, no, you don't have to use XMP to go above JDEC speed. XMP ≠ memory overclocking. You can also just set the speed and timings yourself. Cheapo sticks were the most likely to be bad for AMD running XMP speed. I got a first-gen ryzen system and bought cheapo ram, because at the time memory was having a shortage and it was before everyone yelled about checking QVL constantly. I did manual OC because while the XMP profile was functional, it was slow. (First gen ryzens only did even numbers so all my timings were rounded down.)
|
# ? May 30, 2023 23:08 |
|
On a related note, why don't the JEDEC standards go any higher? They could probably foresee that DDR4 would last through some speed increases, so why not extend the high end with some more aspirational speeds than the very pedestrian 2133? It's not like this is the seventies and they had to carefully trim every possible bit of ROM used out of the spec.
|
# ? May 30, 2023 23:38 |
|
Computer viking posted:On a related note, why don't the JEDEC standards go any higher? JEDEC DDR4 goes to 3200, which is approaching the limit you can get to with reference voltage anyway. I have some absolute cheapo DDR4 3200 32GB DIMMs, they're fine. Server DDR4 is all 3200 now also. For DDR5, it goes past 7200 already: https://www.tomshardware.com/news/samsung-talks-1tb-ddr5-modules-ddr5-7200
|
# ? May 30, 2023 23:46 |
|
Twerk from Home posted:JEDEC DDR4 goes to 3200, which is approaching the limit you can get to with reference voltage anyway. I have some absolute cheapo DDR4 3200 32GB DIMMs, they're fine. Server DDR4 is all 3200 now also. Right, so I've just been unfortunate and only bought RAM with "JEDEC 2133 but XMP 3000" and the like. I guess that makes sense, I haven't exactly splurged on the nicest RAM. And yeah I haven't really got a feel for DDR5, so that was strictly a DDR4 question.
|
# ? May 31, 2023 00:00 |
|
Computer viking posted:Right, so I've just been unfortunate and only bought RAM with "JEDEC 2133 but XMP 3000" and the like. I guess that makes sense, I haven't exactly splurged on the nicest RAM. Even splurge ram made for enthusiasts will often have a very slow JEDEC ratings. 1. that reference voltage thing is a bitch, it's really hard to make it run fast with low voltage. (I think a lot of the time the late-cycle speed improvements are more about the memory die makers having tested the limits of what the platforms will tolerate than anything else.) 2. for enthusiasts & DIYers, JEDEC is your fallback when things go wrong. Setting it at 2133 instead of 2400 isn't a big difference when the kit is supposed to go 3600 or more. And it probably makes it slightly more reliable when the situation is bad.
|
# ? May 31, 2023 00:23 |
|
Klyith posted:Even splurge ram made for enthusiasts will often have a very slow JEDEC ratings. Right, and if you want a way for stick of RAM to specify a higher voltage, you probably also want a definitely-safe profile to fall back on, and from there it's not a far leap to the current state of JEDEC+XMP.
|
# ? May 31, 2023 00:34 |
|
CaptainSarcastic posted:You don't happen to be running a wireless keyboard, do you? I'll try putting the keyboard on a different port. Hadn't thought of that. Thanks for the idea. Klyith posted:A thing to know is that XMP is officially an Intel specification, and while AMD platforms will load the XMP values it's always been a bit of a crapshoot. Works for many basic setups, very iffy with 4+ sticks or bleeding edge hardware. (Also Intel's memory controller is objectively better than AMDs.) Klyith posted:If you haven't yet, you should look for bios updates for your mobo. This being the very tail end of AM4, and the 5800x3d being the final CPU released, you might have a board with only the first release of x3d support.
|
# ? May 31, 2023 01:13 |
|
I think most competent bioses do a checksum after flashing, so it would probably have warned you if it didn't work out.
|
# ? May 31, 2023 01:22 |
|
Vavrek posted:Nope! Wired keyboard, wired mouse, in what the motherboard designers clearly intended to be the KB&M USB ports. (They're the only two USB 2 ports on the back, everything else is some kind of USB 3, and they're next to the lone PS/2 port.) Are you on AMD? Some AMD boards really do not like devices using USB 2 for whatever reason. A BIOS update might fix it, but just not using the USB 2 ports is usually easier and works basically always.
|
# ? May 31, 2023 01:28 |
|
^^^ edit: that should be fixed now, USB bug took a while but is gone. (Also would be just as bad in the OS.)Vavrek posted:Nope! Wired keyboard, wired mouse, in what the motherboard designers clearly intended to be the KB&M USB ports. Ooh, is it a fancy gamer keyboard with full N-key rollover? Sometimes those are weird in bioses, though I think that was a while ago and they (the keyboards) are much better now. Vavrek posted:And the one time I used it was with my i5, yeah. I'd heard a little about EXPO, but hadn't done a ton of reading up on it, so I was mildly surprised to see XMP in a BIOS for an AM4 board. MSI and Gigabyte don't give a poo poo and call it XMP even on AMD even though that's against brand. Asus calls it D-OCP, not confusing at all. I forget what asrock does. Vavrek posted:But I imagine that if a BIOS update went wrong, the board just wouldn't work. Some nice mobos have dual bios, and if something goes wrong during flash they just flip to the other. Should be easy to check though, bios version is prominent and can be viewed in OS with inxi -M or CPUinfo.
|
# ? May 31, 2023 01:31 |
|
Klyith posted:Ooh, is it a fancy gamer keyboard with full N-key rollover? Sometimes those are weird in bioses, though I think that was a while ago and they (the keyboards) are much better now. Ha, my partner reviewed some weird splittable extremely low-keycount (65%?) keyboard, and he absolutely hated that it did that - you needed another keyboard at hand in case you ever needed to interact with your BIOS or bootloader.
|
# ? May 31, 2023 01:36 |
|
|
# ? May 28, 2024 13:44 |
|
Computer viking posted:Right, so I've just been unfortunate and only bought RAM with "JEDEC 2133 but XMP 3000" and the like. I'm pretty sure I bought this exact same RAM for my last AMD system thinking surely it must support 2933 speeds. It never did
|
# ? May 31, 2023 01:48 |