Bad memory tends to gravitate toward me

[blueskull]

After my 7th replacement Surface Pro 7 went bad, my company offered me a new laptop, a proper one, this time an X1 Extreme Gen 2 with i7-9750H, GTX1650MQ, 16GB RAM and 1TB SSD.

I inserted a spare 2TB SSD and replaced the 16GB stick with a pair of matched serial number pair, each 16GB. So it is now 32GB/3TB.

The unit was bought on 19th, and since probably 24th it has started to show graphical defects like shown below:







Windows memory diagnosis went well, cache disabled, and it again went well (unlike the Surface), MemtestCL went well with 3100MB allocation size (any more it fails to run).

So, did I get another botched GTX16xx/RTX2000 chip or is this a software bug? I'd like to replicate it in Linux to confirm, but considering it has only showed 5 times or so in 3 weeks, I don't know when will it show up again.

I really don't want to replace it since I stripped a screw of it upgrading RAM/SSD and the seller may deny service to it.

[BravoV]

You should consider applying for a posiiton at QC dept. in memory manufacturer.

Great auditing results and findings will surely guaranteed career boost. 

[Mechatrommer]

given the repeatability of such posts, my guess it must be geographic or the workflow...

[Mr. Scram]

How do you know others don't have similar issues on that site? They may simply not notice, use a different browser or maybe have different plugins installed. It doesn't really have the hallmarks of a memory issue either.

[SiliconWizard]

That could also be just an LCD panel issue (which are more common on those laptops than memory problems IME), but yes, to make sure this is not a software/driver problem, you should definitely try booting a different OS. Just burn a bootable live Linux distro on some USB stick, it should take you a few minutes.

[rdl]

From what I've read, youtube is using a deprecated API for their user interface that's only implemented in Chrome. It tends to be slow and clunky in other browsers. I somehow got stuck with this "new" interface last week and it took me a while to get the old one back. Is it possible this could be part of your problem?

[Mechatrommer]

given the repeatability of such posts, my guess it must be geographic or the workflow...

Geographic: both US and China.

Workflow: SMF forum and Youtube.

What statistical sense can you make from the above? I don't see much.

not that. i mean something like more electrostatic damage prone area, bad production/handling factory? and as other mentioned, QC? and behavioral workflow like banging the laptop on the bed, torturing all cpu cores with some fancy programs, and then again... on the bed? drink and type etc... i dont know, you figure out that yourself, i havent heard a person with bad feng shui on 7-8 of his computing batch... even my wife with the not so good "workflow"..

[Mr. Scram]

I know MY other devices (Surface Go, phones, desktop PC) running the same Edge Insider and Win 10 don't have such issues.

I'll try Firefox for a while and see how it goes.

Those devices have different configurations and most are even based on a different platform. Why do you think the issue is RAM specifically? I'm not seeing clues that point in that direction.

[ebastler]

You are holding it wrong. 

[Mechatrommer]

as i said, it must be geographic... the last time i saw a bad pixel in a shop is like 10+ years ago. each time i want to buy a monitor/laptop, i will open up the notepad maximized so i can see bad pixels and color irregularity easier, the guy from the 10 yrs ago where i spotted the bad pixels he's selling said "this is a very special occasion!" (meaning i'm a very special/rare person) we keep mocking/repeating this word with my brother when something's up until today because he was present during that time and had a laugh together. because i cant find bad pixel nowadays (unless on some banged down tortured monitor) the test to purchase has become more stringent, brightness/contrast change at each viewing angle. so the latest monitor i bought i swing my head far left and far right, up and down before deciding to buy, the seller is like (jaw dropped) as if she's seeing a retarded cerebral palsy trying to fight with a monitor, well i guess this is the problem with a photo editor with a bit of technical knowledge... i tried to explain something to calm her down.

[rdl]

You can try adding ?disable_polymer=1 or ?disable_polymer=true to the end of the url and see if it makes a difference. Firefox has an addon to do that automatically. Don't know if it will work on non-desktop browsers, youtube may feed them a completely different version.

[magic]

The regularity of those patterns is suspicious. Perhaps something to do with GPU/VRAM as I think this is the only place where graphics are stored in a tiled format where filling a few consecutive bytes with garbage may create a rectangle rather than a line.

edit
IIRC disable_polymer stopped working long ago. I believe you have to append &f6=8 to the PREF cookie. And they show banners that the old UI may be removed soon.

[Zbig]

I feel your pain, man. You're chilling with your friends or sipping a beer watching a good movie and then suddenly, out of nowhere, a super embarrassing moment with your former girlfriend from twenty years ago flashes back like a bitch-slap to your brain. Hate those bad memories.

[rdl]

?disable_polymer still works but I have seen the banner. Seems like the banner can be stopped with uBlock Origin along with most of the other annoying crap. I have it fixed that way on Pale Moon, not sure how long it will stay fixed. Pale Moon has per site user agent override which also seems to work. I just tell youtube I'm using Firefox 42

[T3sl4co1l]

More likely disabling graphics acceleration in the browser, or Windows display adapter settings, would yield results.

I had a laptop GPU once that borked a whole row, something like 32 pixels wide, most of the way across the screen.  Fortunately that was the school laptop which I took in to IT and they swapped the card.

Also seen, I forget if it was the same one or what, but I've seen "dead pixels" even when plugged into a CRT.  Toasted frame buffer, obviously, or something with the same effect.

GPU RAM should show up on the system bus AFAIK (or be mappable as such), but I suppose it may still be inaccessible to a RAM tester.  Even then, I suppose the access pattern might still not be the same as the GPU core does, and it could be a pattern-dependent error.  (Heh, I wonder, with a carefully crafted shader, can you arrange texture buffers in such a way that row hammer is induced?)

GPU being output only, I suppose bit error rates, stuck bits, ECC, etc. aren't very important for the most part.  Perhaps that's changing with the increasing value of GPGPU applications?

Tim

[Mr. Scram]

Rest of RAM, VRAM and RAM/VRAM controllers in CPU/GPU, I can't see how pure random events can happen.

I'm not sure what the first part of your message means. They're unlikely to be pure random events although they can be. What's vastly more likely is that the whole stack of hardware and the many layers on top lead to behaviour that for many intents and purposes looks somewhat random. Relatively high level layers are usual suspects. I'm not sure why you think it's memory specifically as that tends to behave differently.

[Mr. Scram]

More likely disabling graphics acceleration in the browser, or Windows display adapter settings, would yield results.

I had a laptop GPU once that borked a whole row, something like 32 pixels wide, most of the way across the screen.  Fortunately that was the school laptop which I took in to IT and they swapped the card.

Also seen, I forget if it was the same one or what, but I've seen "dead pixels" even when plugged into a CRT.  Toasted frame buffer, obviously, or something with the same effect.

GPU RAM should show up on the system bus AFAIK (or be mappable as such), but I suppose it may still be inaccessible to a RAM tester.  Even then, I suppose the access pattern might still not be the same as the GPU core does, and it could be a pattern-dependent error.  (Heh, I wonder, with a carefully crafted shader, can you arrange texture buffers in such a way that row hammer is induced?)

GPU being output only, I suppose bit error rates, stuck bits, ECC, etc. aren't very important for the most part.  Perhaps that's changing with the increasing value of GPGPU applications?

Tim

A Rowhammer reminiscent attack using frame buffers retaining information was recently outlined. I think the page linked describes it.

https://www.vusec.net/projects/glitch

[SiliconWizard]

That could also be just an LCD panel issue

That can be screenshot?

Ah ah... given the description of it and the look of the artefacts, I just assumed that it was actually what you saw and posted some pictures - that looked close to what you can get with a bad panel.
But yeah a second look makes it obvious those are proper screenshots, so you can consider this some kind of brain fart.

Anyway, it's not quite clear if it happens in any application, or just one? In the latter case (because it looks like it's just in your favorite browser?), suspecting software rather than hardware, on first intention, seems to make more sense to me. Try running a few other graphics-intensive applications, such as some CAD software, and see what you get. If nothing like it happens, just switch to a different web browser...

That may still be a hardware issue, but until you have tested graphics intensively on your computer, I would doubt it. Interesting how hardware people would tend to suspect hardware issues first... (but yeah I remember you already got burnt lately.)

[Nominal Animal]

Do you happen to have a thermal camera you could use to check the temperature of components on the boards inside your machine?

I once had really, really odd issues with an Abit motherboard with an AM2 processor; kinda looked like rare memory corruption, but was basically impossible to duplicate.  It turned out that when I optimized the airflow, a couple of the CPU voltage regulators were stewing in their own heat, and one ended up blowing up.  I mean, literally the top of the switcher chip blew off.  These are easily corrected with small baffles, but finding out they happen is hard.

It is similar to how some people like to type on their laptops while in bed.  If you like fluffy covers, it is easy to seriously reduce the airflow, overheating the machine, because the entire cooling budget is so borderline on typical machines.

If you are anything like me, you like to have your tools work hard, so it is possible that you are just pushing these machines harder than average users, and are getting hit by marginal thermal management causing memory corruption (or like in my case, CPU voltage fluctuations that looked like occasional memory corruption or complete bus lockup).

All that said, I am not at all certain this is the case, but it might be worth investigating.

[SiliconWizard]

I just brought the computer to a nearby Lenovo service store, and they were able to issue me a diagnosis report, which allows me to replace the unit from where I bought it.

What did the report show?

[ebastler]

I was not able to replicate the problem, but I showed them some screenshots.
They said they can't issue a report on that matter without seeing that in person, but they can issue it for another reason.
They opened the back lid and found some scuffs on the edge (very hidden, but technically visible from the exterior) of it, they then issued a report saying the appearance of the computer is not up to ThinkPad standard.

Creatively working around their internal processes in order to help the customer. That qualifies as good customer service in my book! 

[Mr. Scram]

Creatively working around their internal processes in order to help the customer. That qualifies as good customer service in my book! 

It's one way of shifting particularly annoying customers to your colleagues. 

[rdl]

At least they didn't go examine it in the back room, then come out and show you it was water damage which isn't covered by warranty.

[rsjsouza]

It is similar to how some people like to type on their laptops while in bed.  If you like fluffy covers, it is easy to seriously reduce the airflow, overheating the machine, because the entire cooling budget is so borderline on typical machines.

This is a neverending battle with my wife. Sometimes the laptop is on her leg and covering the air inlet.

When possible, I even leave the laptop supported in the back to create a small angle and improve airflow without impacting the typing ao much.

If you are anything like me, you like to have your tools work hard, so it is possible that you are just pushing these machines harder than average users, and are getting hit by marginal thermal management causing memory corruption (or like in my case, CPU voltage fluctuations that looked like occasional memory corruption or complete bus lockup).

I have a similar issue but stems from a somewhat different aspect: most of our computers and laptops are either old or hand me downs, thus the natural evolution (i.e. crapification) of the OS, the browsers, the modern webpages, etc. makes them spin quite furiously for once mundane tasks...