Aaron44126

So, PipeWire was originally started as a project to handle video streams, but they added audio support later on. It's not just Ubuntu that is switching to use PipeWire as the default audio server, in fact Ubuntu is sort of late to the party. Ubuntu switched with version 22.10 (October 2022), but Pop!OS switched at version 22.04 (April 2022) and Fedora switched at version 34 (April 2021). It looks like they're trying to unify use cases for PulseAudio and JACK and generally fix some other architectural issues around audio with this transition. In fact, PipeWire supports to protocols/APIs used by ALSA, PulseAudio, JACK, and GStreamer, so it can be used by applications that support any of those without applications specifically needing to add support for PipeWire. Also, PipeWire supposedly addresses some issues with audio over Bluetooth that have plagued PulseAudio, and is also lighter in terms of CPU use. (When I originally tried Linux as a full-time OS back in 2008, I think that the process to move from ALSA to PulseAudio was just getting underway, so funny to see that PulseAudio is now being phased out right when I switch over again...) So, if that is clearly the way that things are going, I am going to start with trying to use PipeWire instead of PulseAudio and find a way to fix the audio drop-out issue. This guy seems to have run into the same issue and the solution appears to have been to change the value for "default.clock.min-quantum" to 1024. The default value is 16, so that's a big jump. I found other threads reporting stuttering issues and messing with the "quantum" setting seems to be a common recommendation. In fact, I found a note here on the Debian wiki specifically saying that because PipeWire has lower latency that PulseAudio by default, it can be "choppy" on systems with a high CPU load, and they also recommend increasing the quantum value. My system doesn't seem to have any sort of PipeWire config file in /etc/pipewire or ~/.config/pipewire, so I can only assume that it is using the default values for everything. I take it back, I found the config files in /usr/share/pipewire. It did have the "min-quantum" value explicitly set to 16. This command will return the current value for min-quantum: pw-metadata -n settings 0 clock.min-quantum You can change the value on the fly with: pw-metadata -n settings 0 clock.min-quantum (value) This page describes how to set up a split-file configuration so that you can adjust your own values and not worry about them conflicting with the config files that the packages wrote out, when the packages are updated later on. You can also run pw-top and it will give you a tool that shows details of connected PipeWire clients, including what the "min-quantum" value is currently. With Firefox as the only client it is reporting a quantum of 512, but if I fire up yuzu (game emulator) it appears to want to lower the quantum to 256. The config is pretty robust, you can set up rules for specific applications. ArchWiki has an example for Discord. What is the quantum, anyway? It's basically the buffer window, and thus audio latency? Best I understand it, it's the scheduling period for drivers and impacts how long applications have to fill in the audio buffer before a dropout will occur. To measure the window, you have to divide the quantum by the playback rate. On my system, PipeWire is reporting that the playback rate is 48,000 Hz, so yuzu's observed quantum of 256 means 256/48000 = 5.3ms of latency (and also a 5.3ms buffer window). The "min-quantum" of 16 means an absurd 0.3ms. ...Since I generally play at 60 FPS, frame times are 16.7 ms, so audio latency of up to around 30 ms or so is totally fine (I'm not going to be able to tell if a sound comes one frame late) and I don't mind raising the min-quantum to 1024 (21.3ms). For reference... What's latency like with PulseAudio? I'm not running it right now so I cannot easily check myself, but I found this blog post where the guy reported latency of 100ms with default values (includes commands to measure it and configuration to lower it). This bug report also mentions 100ms as the default latency. 100ms seems really high to me?? But, if that is right, it totally makes sense that if my setup used to allow for 100ms of latency and now it was pushed down to 5ms then maybe there will be issues with audio buffer underruns that were not there before. The plan for me is, I will see if just forcing PipeWire processes to run on P cores helped, and if not, I will try min-quantum = 512 and see if that helps, and if not, double it until it is good. This big post is sort of me just unpacking this as I figured it out myself, but maybe it will help someone else down the line!

Yeah, I generally agree, so I need to spend some time figuring out what the motivation and supposed benefits of the switch are here before I decide if I'm just going to just switch back to PulseAudio, which seems like it would be the easiest thing to do, or if I am going to try to make PipeWire work. I have heard about PipeWire in the context of something that exists to help with the transition to Wayland (...which I am all for, I get the benefit of Wayland, even though I am not using it yet...). I know for example that it can help screen sharing apps (i.e. virtual meeting apps, Zoom/WebEx/etc.) that are written for X11 to work under Wayland. I do not know why there is an audio component to it though. I didn't even know that there was until I ran into this issue. The Reddit post I linked above has a potential solution (adjust the "default.clock.min-quantum" value in PipeWire). Similarly, I don't want to go and make that change without first understanding what that value even does. So. Investigation to follow, when I have time. I'll post what I find out back here.

...Ran into audio stuttering for the first time when gaming after upgrading to Ubuntu 22.10. Maybe "audio dropouts" would be a better way to describe it. Come to find out, they have switched the default audio server from PulseAudio to PipeWire. I had previously added some scripting to make sure that the PulseAudio server runs on P cores only, so I guess I'll do the same for PipeWire, but if that doesn't fix it then I guess I can just switch back to PulseAudio and see if that fixes it. (Sort of a shame because the GPU / power throttling has not been an issue at all over the past two days, likely because of the BIOS settings changes I made on Friday? Fix one thing and now have something new to figure out.) [Edit] Looks like maybe it's not just me.

Confirmed, no S3 sleep in Precision 7560 or Precision 7770 here. Attempts to force it back on have not been successful. (You can edit BIOS variables and Windows will show it as an option, but it still won't work right.)

Well. Decided to go current with KDE Plasma, rather than sitting on version 5.24 which is what Ubuntu has in their current LTS repo, so I went ahead and upgraded to (K)Ubuntu 22.10 and pulled Plasma 5.27 from the Kubuntu backports PPA. (Well, it is Plasma 5.27.4, but they should have 5.27.5 shipped out any time now, it's already in the staging repo...) The upgrade was smooth and everything seems stable so far. I'll look to fully upgrade to (K)Ubuntu 23.04 in July around when 22.10 is going out of support. (I'm already using the 23.04 kernel packages...) I can already see some appreciable differences, for example, it starts "listening" for a fingerprint scan as soon as I touch anything on the lock screen. On 5.24, I had to hit the "enter" key to try to enter a blank password before it would initiate fingerprint scan.

Well, I tried really hard to reproduce this issue "on demand" in Windows on the Precision 7770 and failed. I was ready to start a new thread about this problem, but before I can do that, I need a hard reproduction case. It might be that a longer test session is needed, because I generally only run into this issue once or twice during an hour-long game session and it lasts for maybe 20-30 seconds when it happens. (But that's without pushing the CPU to full tilt, yuzu just pushes 3 or 4 CPU cores pretty hard.) Using the Windows 11 OEM image, I ran 3DMark "Time Spy" in a loop while simultaneously running Cinebench. The result over several minutes was a pretty constant 110-120W power draw on the GPU (with it complaining about still being power-limited), and Cinebench seemed to be the one suffering from low performance, which is I would argue the appropriate behavior. The CPU was stuck at 2.05 GHz (as reported by Task Manager), with all 24 threads running, until I ended the 3DMark test. I'm also taking back what I said earlier about a big dropoff between P2 and P3. I watched the power state while running this test and it was bouncing around between P0, P2, and P3, but performing fine all the while. I guess I'll revise my statement to be along the lines of "When the GPU is in the poor performance state, it can't go any higher than P3" but that doesn't necessarily mean that P3 is bad under "normal" circumstances. One thing is, I noticed that I had "Enable adaptive C-states for discrete graphics" turned on in the BIOS, which I do not want, so I turned it off before testing. That could have made a difference. I know that you (@Ionising_Radiation) also found that turned on, but still had the issue after you turned it off. I'm not sure how that got toggled on because I have tried to be careful to keep it turned off. I know that I had tested with it on before and found that it impacted graphics performance for the worse. I did upgrade the BIOS a few days ago for the first time in a while .....? For kicks, I went and tried turning on "Enable Turbo Boost Maximum Technology 3.0" in the BIOS, because the help text there explicitly states that turning this option on also turns off NVIDIA Dynamic Boost 2.0. For some reason I thought that this also negatively impacted graphics performance, but I didn't find any meaningful difference when running the same test again, so I'll try with it enabled for a while. [Edit] Wrote this cheesy script to help me check on this issue under Linux, grabbing CPU power draw, GPU temperature, GPU power draw, GPU P-state, GPU speed, and GPU utilization. (Took me longer than I thought to find a way to grab the CPU power draw with a terminal command...) Hoping that if I hit the Linux GPU performance issue again, there will be some kind of indication here (high CPU power draw + low GPU speed) confirming whether or not it is the CPU/GPU fighting each other for power....... or if it is something else entirely?

It is Linda Yaccarino, who used to be "chairman of global advertising and partnerships at NBCUniversal". https://www.theverge.com/2023/5/12/23721107/twitter-ceo-linda-yaccarino-elon-musk-nbc https://arstechnica.com/tech-policy/2023/05/twitter-users-fear-new-ceo-will-end-musks-commitment-to-free-speech-on-platform/

I was actually just looking at discussion on this. Dolphin will already squirrel away ".directory" files in a different place if you make a change on a folder that it can't write to. An option to have it do that for "all folders" would make a lot of people happy. There's a bug on the subject but it doesn't look like it is getting any real movement.

Yeah, Wayland is super interesting. I tried it briefly during my time with GNOME, and aborted mostly because it didn't work well with VMware Workstation, as under GNOME it would not allow VMware to "capture" shortcuts like Alt+Tab or the Windows key. KWin has its own solution for that, though, you can just disable global keyboard shortcuts for a given application. I'm very interested to give Wayland a go on Plasma when it is ready enough that KDE and distros are comfortable enabling it by default. Hoping it is there by the time that Ubuntu 24.04 LTS drops next year. That's kind of gross. But, macOS does that with ".DS_Store" files so nothing I haven't seen before. I wonder if it will read these ".directory" files even if that option is not checked, so maybe I can just have them sitting around where I want custom options set, rather than "everywhere" ....? I'll think about it. I actually have a small list of such things for various projects that I am considering taking a stab at. Not quite ready to get into that area while I am still sort of getting up and running on Linux. But, the Debian package system makes it easy enough to fetch the source code for a package and pull in all of the dependencies that you need to build it so it shouldn't be too hard to start messing around with stuff like this.

Replying over here so as to not throw the Precision thread that far off topic ... I have found KDE to be a bit "rough around the edges" compared to GNOME, but over all, the level of customization that it offers makes it a real "power user" DE which is exactly what I was looking for. I was only on GNOME for about a week and sort of ran into lack of choice about some things bothering me; the "who's computer is this, anyway?" issue is part of what drove me off of Windows to begin with. I'm currently still on Plasma 5.24 (which is default for Kubuntu 22.04 LTS), but I plan to upgrade to 5.27 in the next few days. Reading about the progress towards 6.0, I'm getting pretty excited about that even if it is a ways off yet. In a way, it seems like this is a good time to hop on, with 5.27 being "stable" for a while I will have time to sort of get used to things and then decide how aggressively I want to follow the latest releases once the 6.0 train gets going. I have some examples of the "rough around the edges" issues above. Konsole has a good example. I would go to settings, make a new "profile" so I could adjust the appearance to my liking, and save it. If I relaunched Konsole, my newly added profile and settings would be gone. Turns out you have to add the profile (with no changes), save it without changing anything else, immediately close and reopen Konsole, and then it will "stick"? That worked for me and I haven't had to mess with it any more since then. My opinion on Dolphin is ... meh. It's fine, better than Windows File Explorer and probably better than Nautilus or Nemo or other Linux file managers that I have tried so far. But it is no Directory Opus. I feel like a "power user file manager" on Linux is still a hole I haven't figured out how to fill yet. Some things I can do in DOpus that I haven't figured out how to do in Dolphin are: Have a subfolder of bookmarks so that I can open a certain set of folders in tabs all at once. (I can probably fudge this with a script or folder full of symlinks or something.) Using the "Details view" mode, have a custom set of columns and sorting that it will remember ... for a particular folder only. (For instance, I want to see photo metadata columns but I don't want those columns sticking around when I am in non-photo folders. Also, certain folders I would like to have automatically sorted by the file "modified date" and not the filename.) If I have a folder full of videos, I want to be able to see how long they all are, but I can't get Dolphin to show me the "duration" of video files in a column. (I can add the "duration" column from the "Audio" group, but it just shows blank for .mp4 files. It can show the duration in the F11 "Information" panel, so it does know how to fetch it...) Interested to hear if you know any tricks that I don't along these lines. But overall, I've managed to work around most of the things that really bothered me and I don't see myself switching off of KDE any time soon. I saw some discussion on how it has stuck around for so long and not fallen subject to the barrage of forks that GNOME has undergone over the years, and I think that there's something to that and they have a good trajectory going.

I did an in-place migration from Ubuntu to Kubuntu. Turns out that I love KDE...

Sigh. Case in point. (This game isn't that graphically intense, but I'm usually seeing a pull in the 50-80W range from the GPU. It's an emulator so it is pretty CPU-heavy. It had to be running at around 10 FPS for a few moments before I managed to grab the screen shot, I was full screen so the indicator wasn't shown...) Haven't taken the time to do testing on the CPU power limit yet. Maybe this weekend. @Ionising_Radiation Still good with the NVIDIA platform controller thing disabled? Have you had time to play any more?

https://www.theverge.com/2023/5/11/23517090/elon-musk-twitter-ceo-replacement Looks like Twitter has a new CEO, and "she" will start in about six weeks, but we don't know who "she" is. Musk is staying on, "overseeing product, software & sysops", so I wonder how much will really change...

Fully understandable. I've thought about this too. I'd probably keep a spare (low-end) motherboard around in case I need to swap that around. I'm not sure what else could break that would be a total emergency. And I'm not looking to switch as soon as Framework 16 comes out. I'll be watching them for a while to see how generational upgrades are going, while I ride out my time on the 7770. But yeah, being in the U.S. is probably best-case in terms of getting parts quickly from them. (Though, my experience with Dell warranty parts replacement hasn't been great lately. I waited almost a month to have my Precision 7770 display replaced, and about three weeks for the heatsink/cooler. I'm no longer considering "next business day" to be a thing that I can count on.) I really do wish someone else would put out a high-performance laptop-first power-efficient CPU. Intel's mobile CPUs are basically repackaged desktop dies with the power limits dialed down a little bit.

New Sabrent drive (warranty replacement) installed, added to RAID0 array, and it is working. ...I noticed pretty quickly that this one makes a sort of electrical buzz noise if I do several GB's of writes in a very short time. I've never heard an SSD make a noise that before, it is definitely noticeable. The other two drives of the same model that I have are silent. I don't do bulk writes all that much so I will just live with it for now. (I will ask the warranty guys about that but I doubt that they will want to spring for another replacement, if it is "working"...) I tested lots of bulk reads and it doesn't make any noise for that. [Edit] One day later, not able to reproduce the buzzing sound from the drive when doing bulk writes. Maybe it just needed some burn-in time ...?

I'm in much the same boat. (...Says the guy who has recently thrown in the towel with Windows and is currently spending way too much time "tinkering" with Linux.) Talking more about the hardware/thermal tuning side. For instance, I haven't messed with undervolting, I don't really have the appetite to figure that out given that it looks like it is on the way out (for business-focused systems). I did repaste my 7770 when it was new, but since I had the cooler replaced last month I've just been using factory paste. It is "good enough" and I'd rather just spend time using the system than fine-tuning that. I'm already annoyed that I have to spend time running tests to figure out where to set the CPU power limit to avoid this GPU throttling behavior. So... I'm also probably done with Dell. There are a lot of things I like about the Precision 7770, but it has been sort of a hassle to get consistent gaming performance out of it (less because of thermals and more because of power management getting in its own way, as we are seeing here), and also it is also definitely a step down from older Precision laptops in terms of how easy it is to swap out components, should that be needed. For example, the keyboard is no longer able to be swapped out without removing it from the bottom (requires a full disassembly), and the display panel is held in its enclosure by adhesive, not screws, making replacement of that an exercise in tedium to get it aligned just right. They also "broke" manual fan control. Well, that's for personal systems anyway. At work, we are a Dell shop, so I'll be continuing my line of Precision laptops for work probably forever. I've looked at HP ZBook and I'm not really sold there. They seem to gimp the GPU power even more than Dell does. I also really care about fan behavior. They have a "never turn off the fans" BIOS option, awesome, but it turns out that it runs the fan a a consistent 2000+ RPM which is too noisy for an idle workload. The price is consistently high, and because less people have them, there is less of an aftermarket parts market. I'm paying close attention to Framework Laptop 16 and really am interested to see the full specs reveal on that. I really like the idea of being able to have a long-term system that can be upgraded piecemeal, like you can do in a desktop. (Shouldn't all computers be that way?) Since I got the 7770 less than a year ago, it will be some years yet before I'm ready to switch again. Hopefully by then Framework has a satisfactory option. I really just want to see a high-DPI 120Hz display, a reasonably high-end NVIDIA GPU, and "not a downgrade" in terms of RAM or storage from what I have now. (Bonus if they release an 18" version.)

Any luck? I thought about this some more. If it really is a CPU/GPU power contention thing, I was wondering if it makes more sense to just lower the CPU power limit (PL1/PL2) rather than specifically downclocking each of the cores. The BIOS or whatever is setting this off probably cares more about the power use than the clock speed. I'm thinking something like... Run a combined CPU/GPU load (3DMark loop + Prime 95?) and observe that the GPU power level and performance fluctuates. ...Dial down PL1/PL2 and repeat tests until it stops happening. (Maybe just lowering PL2 would be OK. It would only be necessary to do this if you plan to put the GPU to use.) I'm going to be doing this on my system soon. I have found similar behavior with more intensive titles that I have been testing on Linux, and it definitely explains some odd behavior that I saw when trying to run games in the full 4K on Windows. GPU performance is inconsistent and it is often dropping into lower power levels. For instance, if I just run the benchmark in "Shadow of the Tomb Raider" with high resolution and graphics settings, I can see the GPU going from smooth to stuttering and back every few seconds. I can see the GPU "PowerMizer" state drop down when it goes to stutter mode. It happens more readily on the "ultra performance" thermal mode, but using the "optimized"/"balanced" mode doesn't fully stop it from happening. It doesn't happen at all if I disable turbo boost; GPU performance is consistent in that case, but I should be possible to have the CPU run at a good bit higher than the base speed and still be OK. Should be able to adjust PL1/PL2 with Throttlestop on Windows or powercap on Linux ...? I'm not thinking that taking steps to turn off "NVIDIA Dynamic Boost" will fully fix the problem. It doesn't look like that is even working on Ubuntu, the "nvidia-powerd" dynamic boost service is not engaged (even though the files to support it are present). I'm thinking the Dell EC/BIOS is at the root of this right now. ...Just too much high-wattage stuff going on in a system that is not properly set up to handle it? And really, the wattage isn't even that high. 😕

Pretty sure Pascal is the first generation where NVIDIA enforces digital signature verification on the vBIOS. So, you can't just dump the vBIOS, "adjust values", and flash it back on. (Well, you can, but the GPU will reject it and won't operate.) You can however cross-flash a different official vBIOS that comes from another P4000 card if it has more desirable behavior than yours. Maybe someone has found a way around this by now but that was the situation when I was messing with a P5000 back in around 2019.

Yeah I saw this exact page, and its "helpful" but I'm not sure if it might cause other issues. For instance, the TLP configuration seems to turn the WiFi on or off when the Ethernet cable is connected or disconnected. Great. What happens if I have Ethernet connected, power the system off, unplug the cable, and then power it back on? Or vice versa? Will I end up with both connected, or neither connected? Will have to do some testing to make sure funny state cases are covered with appropriate behavior.

I just noticed that Ethernet priority is -100 and Wi-Fi priority is 0. Maybe flipping those around would fix it? I thought maybe it was a Linux "lower priority is better" thing but it looks like that is not the case. (That's just the autoconnect priority so it might not even matter for traffic.)

It's become increasingly clear that going back to Windows would be just an exercise in never-ending frustration, so... Linux may have its own frustrations, but I'm working through things bit by bit, and the only way to go is forward. I'll ask this, maybe this is something you guys have experience with. Is there a good way to get it to disconnect from Wi-Fi if there is an Ethernet connection? (Windows does this automatically.) Network performance between my Windows VM and my network scanner (talking about like a document/photo scanner) is poop if Wi-Fi is connected. The VM only sees one network adapter, and traffic doesn't seem to always prioritize the Ethernet connection. I can script a solution to this as well, just wondering if there was an easy way to do it.

I'm not sure if that will really help you. The system might not draw more than 180W of power even if you hook up a 240W. I just thought that it might be something to try. I'm right now more interested in what happens if you try to disable Dynamic Boost... It might stop the system from trying to steal power from the GPU if there is CPU load pressure.

You can disable Dynamic Boost in the NVIDIA control panel... but not for Ampere and later GPUs? Though, maybe a procedure like this would be helpful. [Edit] Now I am also very interested in disabling or bypassing Dynamic Boost on Linux because I wonder if this might be explaining a few odd things I have been seeing with more intense GPU loads. There is documentation on it. nvidia-settings -q DynamicBoostSupport indicates that dynamic boost is supported, but I do not see a nvidia-powerd service/process running. (Using NVIDIA 525 proprietary drivers from the Ubuntu repos.)

You're on a desktop right? I think that the power management behavior is going to be different. I always see the GPU spending most time in P0. But, the laptop GPUs have a way lower max power level. I had another thought. If we're on the right track, this seems to be a case of a "not ideal" (putting it nicely) implementation of power sharing between the CPU and GPU. Dell's older systems would just power-starve the CPU if the GPU was under load, rather than the other way around, and I think that would be the preferred behavior for gaming (and most other high-GPU loads). Anyway. Do you see the "NVIDIA Platform Controllers and Frameworks" device in Device Manager? I think it is under "Software devices". That driver is at least partly responsible for handling NVIDIA Dynamic Boost (shifting power allocation between the CPU and GPU). I wonder if the system behaves any differently if you just disable it.

Yeah, on desktops or pro GPUs the scale might be wider than the ones I enumerated above. Those are what I have observed on laptops and there is a big drop-off between P2 and P3 (at least on Ampere GPUs in the Precision). You can force a certain power state with NVIDIA Inspector but I don't think that will help here. I tried messing with it before; if Dell wants you in a low power state, it's going to put you in a low power state regardless of what you try to force. And now thinking about it, I have experienced "oscillations" before where a combined CPU/GPU load would force the GPU to throttle on and off periodically. I think the best solution is along the lines of what @Etern4l suggested up above. Do what you need to do to lower the CPU power draw to the point where the system stops throttling the dGPU. You might have to be more aggressive with it. As another test, you could run a high-GPU low-CPU load for a while and confirm that this issue doesn't occur. I'd suggest 3DMark Fire Strike, just run a single GPU test in a loop, windowed. It should put the GPU up to 100% utilization and not even stop for loading between loops. (Then maybe add a separate CPU load and see if the GPU starts power throttling.) [Edit] One more thing. Do you have a 180W or 240W adapter attached to your system? Do you have access to a 240W to test with, if you are using 180W? I don't know if it will matter, but at least on Precision 7670, Dell is shipping 240W with systems with a high-power dGPU.