Dell deletes Latitude CPU Throttling issue after link is posted here
Dell's new Latitude 5420/7420/7520 Notebooks have a ongoing issue where their CPU are severely throttled when running on any Linux distribution, even on Ubuntu OEM Certified System according to some reports at the forum.
After a link to the Dell forum post was ported to HN, the post mysteriously was deleted in the same day. Many users of this model posted to the forum thread since last year and so far no solution was provided and we only received very unenthusiastic responses from Dell support.
Shocking to see this apparent action to hide the problem.
Originally Here: https://www.dell.com/community/Latitude/Latitude-5420-7420-7...
You can still see it on Google Cache https://webcache.googleusercontent.com/search?q=cache:PFSFF4...
Screenshot here: https://imgur.com/a/Va8EMMA
300 comments
[ 2.7 ms ] story [ 269 ms ] threadBut, no. I agree, it's not good but Dell has a history of doing this and other dodgy updates/implementations for their EFI.
But I did have a similar issue with Dell XPS 13's 9350, 9360 if I used a non -oem battery, on ubuntu and windows 10, immediate throttle even if I disabled speedstep. Only going to an OEM battery did it fix the issue - though I know 100% this is not related as the users are new laptop purchases.
Still my 9350 experience was great outside that.
The US needs to adopt EU-style consumer laws.
Dell would die a quick death without corporate IT.
That, plus structural defects cause case flexing when picked up, resulting in the touchpad being inoperable and (more recently) freezing the system entirely.
This is even when the pick up is very gentle -- the case flexes under its own weight. If I type too hard or rest my hands too heavy on the system it can also occasionally cause the issue.
The issue started two days out of warranty.
Last week I was sitting in a meeting and tilted my screen up a little on the XPS 15 so I could see past some glare from lights and the hinge snapped. It has the smallest little piece of metal at the point it snapped, about 3-4mm wide and 1-2mm thick, it was no wonder it snapped, so work had to give me a new XPS 15. The hinge itself is plastic-welded to the chassis so it can't just he replaced without the whole chassis being replaced.
I was planning for my next laptop to be a Framework anyway but my (£1600) XPS 13 is only ~2 years old so I am a little way from wanting to buy a new laptop just yet.
This sort of behaviour from DELL is confirming that going Framework is probably the right choice.
EDIT: Also don't get me started on the i9 / 4K XPS 15 a friend of mine bought that seems to want to kill itself any time it's switched on, and has never worked correctly since day dot.
I've had XPS laptops all the way back in 2010, then 2013 and then 2016 - they all rocked on a table, it's like Dell's factory is crooked and they are physically incapable of making an even laptop.
Now, I agree that kind of thing shouldn't have to be thought about as a wear item -- but it seems to be one.
Neither laptop has any wear whatsoever on the rubber feet from sliding them on a desk, or on anything else for that matter. I'm a software engineer, not a savage.
From my rudimentary testing, the front right corner is a little bent upward compared to the rest of the laptop. If I touch that corner with one finger it rocks, on both models and on any level surface.
Both of my machines are older than your 7590 but they both did it since new.
I got a 2020 model this week when the hinge broke on my XPS 15 but I'm yet to test if it suffers the same issue.
I assure you all of mine did that straight out of the box. When the second one did it I was like "Huh, what are the chances" but I almost expected it with the third.
I've been through my fair share of laptop brands over the years.
Leaving Apple out of the equation, the XPS is still the best built, thinnest (though not lightest) laptop I've had that's got the grunt to get my work done.
The XPS 13 is disappointing because of the soldered WiFi meaning I can't swap out that crap Killer WiFi for Intel.
Unfortunately Dell is cheap, make (or at least used to) fairly robust computers for their price with spare parts that can easily be found. Mac is much more expensive and good luck sourcing parts from other brands. Lenovo went downhill when they were sold off.
I’d reckon equivalent M1 Macs are much cheaper.
The moment you deviate, all bets are off.
I think that was a valid argument before Apple Silicon.
I guess it's a philosophical question whether Dell should stop selling these kinds of machines, or continue to separate those who ignore basic physics from their money.
I have a semi-portable Dell Precision M6500 from 2011. It still runs and does what it should, but it weighs 4 kg and is 3.5 centimeters thick. Both those numbers are twice that of an XPS 15, and it still throttles down the quad-core i7 after about 120 seconds at full load IIRC.
This comes off as victim blaming.
As a consumer, I would expect the hardware I buy to be able to dissipate its own heat. If it can't, that's a design flaw that Dell needs to either fix or stop selling the product.
I do that with one of my GPU servers, because the alternative is switching to liquid cooling and that's more work than I want to put in right now.
Actually, I'm pretty sure that false advertising is illegal in most countries, including the relevant one here (the US).
Going off-topic, but yup. It's not victim blaming to point out when a victim didn't even take normal precautions. For example, if you're crossing the street, then even if the walk signal says WALK, you should still be looking to make sure there aren't any drivers running the red light. While yes, if you get hit by a car, it's the fault of the driver, you still failed to take the necessary actions for self-preservation.
Hell, there's a pedestrian crossing near me that isn't at an intersection, but has yellow flashing warning lights that turn on when you press a button. There's also a speaker that says "Cross street with caution, vehicles may not stop".
Back when dual GPU’s where still a thing they actually needed laptops to be thick or wide, but 100w isn’t that bad.
Also, the 2017-2019 Intel MacBook Pros are poorly suited to "professional" workloads, throttling under load or if, get this, the "wrong" (Intel) TB controller is used. There are reproducible TB bugs that are abandoned, and I still believe T2/BridgeOS was Apple's response to Intel's poor security design; they were forced to design adversarially against components they couldn't trust.
Intel website shows their most recent Vega M release is 1/27/22. If Dell isn't keeping their drivers up to date, that's their problem, but the Vega M drivers are out there, Dell probably just isn't putting out releases.
https://www.intel.com/content/www/us/en/download/19269/radeo...
anyway, more broadly, this is, unfortunately, just AMD's driver support model. For the longest time they would not even put out iGPU driver updates for their own processors, instead offloading this responsibility to the laptop vendors themselves (who, of course, did not give a shit). i.e. exactly what's happening with your dell.
https://www.notebookcheck.net/Better-late-than-never-Radeon-...
And I don't know what people really expect here - Intel is dependent on AMD for drivers, they absolutely cannot write and support a whole new driver stack for an AMD product, if AMD sandbags (like they did previously to OEMs) then Intel is pretty much stuck. AMD got shamed into supporting their own products, finally, but they're absolutely not going to give Intel an inch more than they are contractually obligated to. It's in their financial interest to make your experience as shitty as they can.
Of course, it was probably a mistake for Intel to attempt to collaborate with their largest competitor, and they definitely should have made sure the contract was airtight. But there's really no good reason for these not to be supported by mainstream Adrenalin drivers.
Is Corporate IT the real reason Dell continues to exist, or is it Intel?
[1] https://www.theatlantic.com/technology/archive/2010/07/dells...
What have you used since then that made life better?
It's soldered in so can't be replaced.
This is a laptop I bought with Linux support out of the box but I've given up expecting Dell to fix it. It's easier just to kill off and restart the wifi connection when it stalls.
Given that I'm not sure I'd buy another Linux Developer Edition laptop from Dell.
I think that XPS build is just the wrong machine for Linux and like someone else said the wrong form factor for poor power performance management.
But I also had several coworkers with the same newer 11-series XPS machines who got a series of lemons... Dell replaced the motherboards, but still.
It was the worst linux experience I've ever had, and just a bad experience in general. They gave it a broadcom wifi chip which has poor linux support, no built in drivers out of the box so whenever I reinstalled I needed to connect an ethernet wire with a usb dongle to download the driver. The wifi also just had very poor reception compared to other laptops.
Sound problems intermittently where half of the time it would have no sound on boot, and I'd have to continually reboot until it decided to work.
They had an absolutely infuriating feature which adapted brightness levels of the screen so that if you transition from dark to bright it would gradually adjust. Except this feature was broken and it does the opposite of what was intended -- it instantly cranks the brightness up to 100 and then gradually adjusts down to what the beightness should be. Even if this feature worked properly it would be undesired, but it is broken in the worst way and there is no way to disable it on my model. I have spent many hours trying.
The computer also arrived with the chasis mis-aligned. I couldnt plug anything in to any ports, we had to get a dell guy come in and reasemble the machine.
The charger also stopped working after about 1.5 years. I bought a new charger but now the charging port only works intermittently.
QA issues aside, you'd think that they would choose parts known to work well with linux but they inexplicably did not.
In S3, the system controller powers down the CPU and nonvolatile storage while keeping RAM in a low-power self-refresh mode until you wake it up. Your operating system is not running -- the microcontroller that reads the lid sensor and power button determines when the laptop will resume.
In S0ix, the CPU is still running, and the OS kernel is supposed to put itself into a low power state most of the time, while periodically doing things like checking email and performing updates. The lowest power state still uses gobs more power than S3, and Microsoft is aggressively irresponsible about when and why it wakes up. It's more like "turning off the display and spinning down disks after inactivity" and less like a real sleep state. Windows is running the show 100% of the time, even when the laptop is in your backpack.
All this because Microsoft was jealous that Apple can check email and download updates without waking up using their T2 system controller and now their own SoCs. But Apple is way less aggressive about waking up, their firmware is generally pretty okay, and T2 still uses less power than an x86 in standby.
Windows, on the other hand, has to run on every PC, so they can't specify some crazy advanced system controller, nor does Microsoft have any control over the firmware it runs... Instead they had to ask Intel to add a new sleep state that puts Windows in the driver's seat, and in order to make sure it gets used and people don't "accidentally" get stuck using the "old" S3 sleep, they made it so that you can only have either S3 or S0ix enabled, so all system builders would essentially be forced to switch. Some (Lenovo) offered a BIOS flag to surreptitiously switch between S3/S0ix before boot time, but even that seems to be going away or broken lately.
I have to believe that this will get better eventually, if only because the current state of things is so atrocious, but I wouldn't plan on buying a new laptop in the next 2-3 years.
Is AMD in a better place regarding sleep states?
As far as I'm aware, S3 and SOix both exist on AMD, but I'm not sure whether they can exist simultaneously. There seems to be just as much confused annoyance at power consumption issues on AMD laptops as with Intel, e.g. https://forums.lenovo.com/t5/Other-Linux-Discussions/P14s-Ge...
Not really sure what to make of that, but it sure is disheartening.
Separately, while AMD's better performance / watt would seem to make it a good candidate for laptops, lack of Thunderbolt across the board has also been frustrating, especially if you already have a nice TB dock setup.
Or stuff another embedded core onto those enormous CPU dies.
Unfortunately, I don't know of any good PC alternative to Dell.
I just ordered a Macbook.
[1] "...Framework Laptop is working without issue in Ubuntu 21.04.3+ (NOT 20.04.3) and Fedora 35"
https://community.frame.work/t/official-linux-and-framework-...
Because of this, my newest laptop is a Lenovo... no more Dell computers for me, sorry Dell, not sorry.
They are pretty old though and just barely fast enough. Last year I ordered a Precision 3630 to test out as an upgrade. I was so disappointed by the engineering and build quality I requested a return the day it arrived.
That said, if not anything else, Lenovo provides the best laptop keyboards in my opinion. Perfectly sized, perfectly spaced, and give the best feedback out of all else. The thermals and overall build are also easily the best on my Ideapad than any other mainstream laptop I have used, though I'd concede I haven't really used as many laptops.
I would not have bought it without the assurance that I'd be able to run Android apps. Never a word from HP about the issue (and obviously no attempt to make the situation right after they lied). I'll leave it as an exercise for the reader to decide if you should give them your money.
[1] https://sites.google.com/a/chromium.org/dev/chromium-os/chro...
The speakers are mediocre though compared to the work Mac I used to have and are mostly in line with other Windows laptops sadly. I thought they'd sound great since there are these vents behind the keyboard that look like speaker grills but they're actually for cooling.
One of the downsides is that it has coil whine when charging off the USB-C port which doesn't bother me much as I almost exclusively use it on battery or with headphones. The fan curve is also really weird where it's silent 80% of the time but it sometimes turns on and off for 15 seconds at a time which is a bit jarring. I think there's some way to tune this though but I wish the fan were a bit quieter or that there were a way to just have it cool passively. I've also not managed to get the fingerprint reader working, mostly for a lack of trying, as it seems like there's a fork of libfprint that specifically adds support for it (Elan 04f3).
I also have an HP Chromebook 14 with an Intel Celeron N4000 and 4GB of RAM which I got as a cheap laptop to bring on trips for remotely accessing my desktop and it's only real pain point is the performance - Android and Linux apps work, sound is on par with the other laptop, display is decent, no fan and no coil whine so it's very nice to use in a quiet environment and fantastic battery. I was actually surprised when I found out it can drive 2 1440p displays off a single USB-C port which is something my flatmate's M1 Macbook Air can't do.
So yeah, I can say that the consumer line HP laptops are pretty good this gen, just be careful with the displays since a lot of their laptops still use sucky 250 nit screens and also be careful with Nvidia GPUs. In my experience those never work well and you're better off getting a desktop/console/streaming service for gaming or renting a cloud GPU for compute.
My main gripes:
For first few months wifi driver was unstable and would randomly cause kernel panics, sometimes once per day, sometimes once per week. (yes I tried manually installing newer versions and using different kernel versions)... eventually it stabilised and I haven't had any issue since - Even though this was not unique to this machine, it was annoying considering it was supposed to be built for Linux.
The battery capacity diminished faster than anything I've ever owned... currently at 20% of original capacity. There was a very sharp drop last year from 80% to 60%, then another sharp drop from 60% to 40% which I'm guessing is the controller writing off cells. (Manually calibrating it by draining never gets it above 30%)... Speculating this is possibly from heat damage (sometimes I will play a game which will get it a bit hot) or just crap battery quality - expecting to find some spicy pockets when I get around to replacing it.
Other than that it's been a solid Linux machine and very comfortable lightweight form factor to use. I still don't have performance issues personally, so will replace the battery soon and see how much longer I can keep using it.
> sometimes I will play a game which will get it a bit hot
Off topic, but how did you configure i3 to work nicely with gaming? Last time I tried i3 with programs that expected to be able to arbitrarily draw windows it didn't work super great.
These days I use Pop!_shell where I might want to game, and Sway where I'll just want terminals and a browser.
However i've also tried out steam occasionally and not had any issues.
What kind of games have you had issues with?
Note i'm using X11 still and no compositor, you mention Sway, maybe these are Wayland specific issues?
[edit]
> I tried i3 with programs that expected to be able to arbitrarily draw windows it didn't work super great
Ahh, if you mean the game needs to draw multiple windows, yeah I've noticed programs that draw lots of floating windows can get into a mess, but I've not really used many of them often. I'm not sure why this is, it's not like they are using the i3wm default floating dimensions or anything.
For games windows that start floating I usually just full screen them or drop them down into tiling mode. But then i've never played a multi window game.
Mostly Windows games through Proton/Wine expecting to be able to take full control of the screen, and then having a panic attack when not allowed to.
As far as programs wanting to draw many floating windows, that's happened with Krita/Inkscape more than games, games just... didn't play nice. I'd get crashes when changing workspaces, when launching a game, so on. That was on i3 itself as well, I didn't switch to sway until after I stopped trying to enforce tiling on everything I do (just most things :P )
I suppose one good thing about Apple and MacBooks is that just because of how incredibly common they are, and the amount of mindshare they command, if there is a fatal flaw with an Apple product, there will be widespread consumer knowledge of the issues. On the other hand, models from Dell and these other manufacturers sell in low enough volumes that they can just keep pumping out broken products, and not enough people will care to hold them publicly accountable.
I'm sure this will happen around the same time a wealth tax is implemented, universal healthcare is adopted, the US stops invading countries, and we stop subsidizing suburbs. (i.e. never)
But to force the people to go fully metric would take a genereration.
Longer than that, I'd say. I remember watching videos about the metric system when I was in grade school back in the 70's. ("You can measure how far from your car to a star -- with the metric system!")
Non-metric preferred sizes are embedded everywhere.
It's already taken more.
Same thing with manufacturing. Most places switched to being able to do metric measurements by the 80s, but for older parts it's still in "thou" and other metric sounding divisions of inches.
Drug dealers on the other hand talk about eighths and teenths, though. But everyone knows that a nickel is 5 grams. Both literally and figuratively.
Milligrams? Grams are already quite a bit smaller than an ounce. A millilitre of water weighs one gram.
Occasionally microgram is used, just as i have to occasionally measure things down to ~20 microns. I'd never use inches, there, and i'd never say "a kernel of corn weighs 1/250th of an ounce". This whole thread kinda caught me off guard. Normally i am making fun of the metric system, but something rubbed me the wrong way up-thread.
If you come from a Celsius place and then move to the USA, you find it actually fairly annoying, because a bunch of useful numbers to remember is arbitrary as fuck. What is the freezing point again? 35, 37, 33? In Celsius it's 0 and very easy to remember. The freezing point is very useful, since you'll know if the rain will become snow if it's just under 0, or if ice will start forming on the road or not. Similarly what is the boiling point? 200, 205, 210, 212? vs. a nice round easy to understand 100. You can also think about how hot your water can be as a 'percentage of boiling', so 75 becomes 'hotness that is %75 the way to boiling'.
Vast majority of the world also knows that -17 (OF) is cold as fuck, and there is nothing special about that specific number too! You would also know that -16 and -18 are cold as fuck. Same with 38C (100F). 40C and 39C are also hot! And there are no unique properties about 0F and 100F, unlike 0C and 100C.
The radiator thermostats I use have a 0.5 °C gradation. A 1 °C gradation would be too coarse as far as I’m concerned. That doesn’t mean that I’d prefer Fahrenheit, but 1° C is actually a quite significant step subjectively for environmental temperature.
you were doing so well until right here.
Having said that I've never actually used cables or fathoms in practice, only metres or feet for depth and nautical miles for distance.
Given how politically polarised the US is it'd end up as political football too I think, in the UK opposition to going fully metric sometimes manifested itself as Euroscepticism and that was back in the '00s when we didn't have nearly such aggressive culture wars.
it was a political football here back in the 70s when there was talk of metric adoption.
i do hope that people like stewart brand (whole earth catalog, coevolution quarterly, how buildings learn etc) feel some level of shame for the role their proseltyzing against metric, in spite of the good things they also did.
There's some things where knowing the SI definition is irrelevant, like how many PSI a tire needs. Sure, knowing what that is in SI is useful if you're trying to figure out how much a car weighs, but if i'm at the gas station setting the pump to 35 doesn't really matter.
Industry and commerce is conducted on the global standard, though. And while the average american might not know how to convert between kilometers and miles (or hogsheads or chains or ...), but knows 500km is far, 500kg is heavy, and 500l is a lot.
Until the SI stuff is the largest part of labels, and people feel like paying a ton of extra taxes to replace every road sign in the entire country (there's millions!), it's never going to be a "metric" country. Just think of the logistics and cost of replacing every road sign, including "mile markers", exit number signs, distances on all US government roads (BLM, USDA, Forestry), overpass signs, speed limit signs and road paint - i don't see that happening any time soon. If the only benefit is "the rest of the world stops talking smack because we're backwater imperial measurement users", would you force everyone to pay for that?
This is pretty much why British roads still use miles and yards, the cost to the taxpayer of switching over would be horrendous. We do have dual measurements on heights and widths though, feet/inches and metres because many HGV drivers from the Continent aren't familiar with feet and there was concerns about them driving into bridges and so on.
America hasn't even pulled back and we still can't stop Ukraine from getting invaded. Besides, preventing Ukraine from getting invaded isn't worth a whole lot of money as an American. We can not and should not try to be the world's police.
This Utilitarian cynicism is all too common, and unworthy of a great people. Rule-lawyering our way out of helping others seems petty and mean.
This has never been true.
>Rule-lawyering our way out of helping others seems petty and mean.
Kinda like how we told Russia we wouldn't be expanding NATO eastward and after did that and pointed it out the response is akin to "neener neener did you get it in writing?"
That's not what the Budapest Memorandum said, this was the actual promise:
>Seek immediate Security Council action to provide assistance to Belarus, Kazakhstan and Ukraine if they "should become a victim of an act of aggression or an object of a threat of aggression in which nuclear weapons are used".
Such action was indeed sought at the UNSC on March 15, 2014, in spite of the fact that there was no threat of nuclear aggression. We actually exceeded our obligations under the memorandum.
>Seems a bit shitty to take their nukes
They weren't their nukes, they were Russian nukes, and IIRC Ukraine had no way to actually use them as nukes, even if they wanted to.
The tape is added by our IT department, because otherwise they'll have to replace the dock after two weeks because the plug just dismantles. Great build quality...
They’re supposed to be a premium laptop but after 2 years it’s essentially a desktop.
Windows hardware in general is junk. Saving for a MacBook at the moment.
https://en.wikipedia.org/wiki/Streisand_effect
For me it seems like everything except the "Work from Home" forum has disappeared. So its probably just some forum issues.
archive of page 1: https://archive.fo/5iQCF
There are 14 pages or so, I suggest people save every page of the discussion still in google cache and archive it, for instance page 2:
https://webcache.googleusercontent.com/search?q=cache:JO-IqO...
https://archive.fo/9BYs4
Since Google cache might be deleted at Dell's request. Don't just use screenshots, these can be easily doctored.
edit: corrected the model's name.
https://www.dell.com/support/kbdoc/en-us/000139695/balanced-...
Just move on when shopping for a laptop.
Maybe someone did a delete where post like '%Linux%'
https://i.postimg.cc/1396wMSb/Dell-HN-Issue.png
LE: I had to disable both NoScript and uBlock for me to be able to see their pop-up and after clicking on it, the image loaded properly on imgur. Still a shitty site IMO.
I'm not defending those solutions, and I don't know if Dell's builds were even more limited than some other brands or not, but I doubt those kinds of issues were unique to them.
https://html.duckduckgo.com/html?q=dell%20lattitude%20thrott...
it seems to have been a longstanding issue with not quite recent CPUs accross a number of models. the thermal throttling was not right.
there are some fixes about:
https://github.com/DivyanshuVerma/throttlestop-linux
https://www.techspot.com/community/topics/dell-fixes-latitud...
https://unix.stackexchange.com/questions/579159/who-is-throt...
https://github.com/erpalma/throttled/
Been using it on various XPS laptops for years.
Sadly it requires support for MSR writes which are being slowly deprecated by the Linux kernel. I believe the same should be achievable with thermald be I haven't been able to be bother yet.
I'm still fuming.
Writes to arbitrary MSR are still allowed by linux, but it'll squawk a warning with the PID from what poked the MSR. What has been preventing writing to MSR is microcode updates (loaded via Linux via initram/firmware, or via the vendor in a BIOS update), such as the one(s) in response to plundervolt https://plundervolt.com/ Writing to the plundervolt-related MSR was particularly helpful to undervolt a thermal-throttled laptop. So! Can either downgrade the BIOS and/or rollback the microcode that your distro is loading during boot.
I should note that the 8656u CPUs have been nothing but trouble (multiple mobo replacements and still have suspend issues).
alias cpuspeed="watch -n.1 \"cat /proc/cpuinfo | grep \\\"^[c]pu MHz\\\"\""
prints out current mhz on all cores, normally you'll see in the 2,000 mhz range (2ghz) and the 400mhz would be quite noticeable. for the record i have a dell precision 7540 with ubuntu from the oem and it has been a journey to make it workable
https://www.linux.org/docs/man8/turbostat.html
for stress testing and monitoring you could try https://github.com/amanusk/s-tui
Two reasons why Turbostat is my weapon of choice here. It's part of the kernel tools package which might be important if you don't have access to outside repos (think CoLo sites, etc).
The other reason is acces to tsc frequency which is pretty important in my line of work.
I'll definitely give s-tui a try. Looks really nice.