← 목록으로

NVIDIA's SSD Write Bug Gets Patched in GeForce Hotfix 621.35 Driver - TechPowerUp

NVIDIA's SSD Write Bug Gets Patched in GeForce Hotfix 621.35 Driver - TechPowerUp

요약

NVIDIA today has released the GeForce Hotfix 621.35 driver to fix the bug that was hammering SSDs while idle. There is only one fix in the release: "Excessive writes to file nvlog.nvlgstg may occur in the background," and that was all that was needed on an ur…

본문

Saturday, November 21st 2026 NVIDIA's SSD Write Bug Gets Patched in GeForce Hotfix 621.35 Driver by DrUsmanSaleem Yesterday, 08:50 Discuss (30 Comments) NVIDIA today has released the GeForce Hotfix 621.35 driver to fix the bug that was hammering SSDs while idle. There is only one fix in the release: "Excessive writes to file nvlog.nvlgstg may occur in the background," and that was all that was needed on an urgent basis. We reported the issue of constant and excessive SSD writes that was spotted by several NVIDIA users earlier this week, particularly on drivers 621.14 and 621.24. The NVDisplayContainer.exe process kept rewriting even on PCs that were just sitting idle, and the nvlog file was at the center of that bug. In some extreme cases, the idle writes were up to 30 GB per hour, while the older 621.01 driver only logged around 2 MB per 5-minute idle window.

The hotfix is built on NVIDIA's latest Game Ready driver, which is currently the 621.24 release from November 12. This is an optional download, of course, as it goes through a shortened QA process and is only offered through the NVIDIA Customer Care site. Previous fixes for the SSD write bug included rolling back to 621.01 and a temporary fix that included stopping the NVIDIA Display Container LS service. Anyone who was affected by this issue is advised to grab the hotfix as soon as possible, while others can wait for the next WHQL release. Curiously, AMD also suffered from a similar SSD write bug just a few days before NVIDIA's, and it too released optional Adrenalin drivers to patch it quickly. Sources: NVIDIA, NVIDIA Customer Care

Related News

Add your own comment

30 Comments on NVIDIA's SSD Write Bug Gets Patched in GeForce Hotfix 621.35 Driver

#1 Geez, that was actually pretty quick.

Sure, it should never have been an issue in the first place, but still, good diligence from Nvidia #2 What’s with all the quick polls recently? Not saying they’re bad, just a bit curious. #3 Technology and process moves forward and things like this become a way of life, and while even coding languages had limitations and thus things like we have today couldnt really exist, you have to wonder how much of it is just new blood that doesnt think through the same lens as the older programmers.

When I started to learn you had to understand the HW side almost just as much, now its all abstracted away from your average developer.

You sure wouldnt see a delay in patching or even the bug to begin with if you were writing 30GB/hr to a 80GB WD Blue IDE drive. You would hear your mistake at around the same time you noticed your system lock up. #4 Remind me how much time it took AMD to acknowledge the issue then start fixing it ? #5
SiliconSeraphGeez, that was actually pretty quick.
You pay for it :laugh: #6 Nice to see this fixed quickly, though I'd expect this to not happen in the first place with Nvidia especially if you're paying more for it. #7 how can anyone not think this is intentional to maybe drive ssd sales? "if theyre going broke buying new ssds they cant bitch about ram or gpus" #8
InfernaRemind me how much time it took AMD to acknowledge the issue then start fixing it ?
At Nvidia's headquarters:

Engineer 1: AHAHAHAHA!!!!
Engineer 2: What?
Engineer 1: AMD's drivers are writing 500KBs/sec constantly on the drive and people are laughing at them again.
Engineer 2: Uh-oh....
Engineer 1: Isn't it funny?
Engineer 2: Oh boy.....
Engineer 1: What? ....wait..... we didn't.....Did we?
Engineer 2: .......
Engineer 1: We did. How much?
Engineer 2: Well, a couple MBs/sec.
Engineer 1: Oh boy..... Let's fix it. I read about a quick fix for the AMD drivers, maybe we can start from there.
Engineer 2: At least they haven't found yet. We might fix the drivers before they realize, or at least we will be faster than AMD.
Engineer 1: Yeah, right, yeah, that's good, it's fine, everything is fine. Well, not really with the CEO and stuff, but it's fine. Maybe no one will notice now that everyone is occupied with that thing.


That's why Nvidia fixed that faster. And while the above is an imaginary dialog, you bet someone was looking at the Nvidia drivers when the news about AMD drivers got out.
Hecate91Nice to see this fixed quickly, though I'd expect this to not happen in the first place with Nvidia especially if you're paying more for it.
It was writing more MBs/sec, I don't understand you. More than AMD's drivers. MORE. :p #9
john_It was writing more MBs/sec, I don't understand you. More than AMD's drivers. MORE. :p
Yeah that's the important thing,AMD drivers were writing anything from 500kb/s to 5MB/s at worst from reports I've seen, and apparently you were less likely to be affected updating from the Adrenalin panel.
But regardless of Nvidia drivers getting fixed faster it was writing up to 30GB per hour. It's some amazing irony everyone laughed at AMD drivers yet Nvidia's AI vibe coding one upped them on how much worse it could've been. #10
InfernaRemind me how much time it took AMD to acknowledge the issue then start fixing it ?
4 weeks to acknowledge and have a private fix done, 6 weeks to release a hotfix for RDNA 3 and RDNA 4 because of the outrage in news oulets. 7 weeks to release the fix for RDNA 1 and RDNA 2, and that was also the original planned date to publish the fix for RDNA 3 and RDNA 4 to the public before the news headline started making noise. #11 Glad I use Linux, because I would be pissed if I lost a bunch of SSD health to this nonsense at a time like this. #12 So Nvidia has the same problem as AMD with massive amounts of pointless log traffic all of a sudden?

What framework or library is to blame, or are these two unrelated events, both affecting GPU driver log-dumping craziness at coincidentally the same time? #13 So what's the punishment for this to people's ssd's? Let me guess, nothing. Do they even monitor/test this properly? Probably not. #14 I checked the total amount of writing since using the container service. It said 28 thousand MB. So, I installed this hotfix and used a workaround so that the container service won't create any more (useless) files/data. #15
FatloaderWhat’s with all the quick polls recently? Not saying they’re bad, just a bit curious.
They were apparently underused, but are now being slightly overused 🤷‍♂️ #16 www.nvidia.com/en-us/geforce/forums/game-ready-drivers/13/593169/announcing-geforce-hotfix-driver-62135-released-11/3634901/

RTX 50 series is unaffected though by the previous driver issue.. #17
Hecate91Yeah that's the important thing,AMD drivers were writing anything from 500kb/s to 5MB/s at worst from reports I've seen, and apparently you were less likely to be affected updating from the Adrenalin panel.
But regardless of Nvidia drivers getting fixed faster it was writing up to 30GB per hour. It's some amazing irony everyone laughed at AMD drivers yet Nvidia's AI vibe coding one upped them on how much worse it could've been.
Time for the AMD fanboy crowd to calm down, it is fixed, and beside not all was affected by this, including me.

Like, do you take it personally that people "laughed" at AMD? :roll: #18
Chrispy_So Nvidia has the same problem as AMD with massive amounts of pointless log traffic all of a sudden?

What framework or library is to blame, or are these two unrelated events, both affecting GPU driver log-dumping craziness at coincidentally the same time?
Likely unrelated, debug logging left on by mistake I expect. #19 And Windows (and Linux) logging: That's not an issue..!?

I have been moving this sort of incessant small writes to HDD (or DRAM/tmpfs) for years, using symlinks where other options are not available.
(Other options for Windows, but that does not include the .etl files where you have to use 'SymLink' IIRC)

Doing this not only adds years of lifespan to your SSD; it also makes your system noticeably more snappy.


www.tweaktown.com/reviews/6860/samsung-850-evo-1tb-3d-nand-ssd-review/index.html

I always thought of this as 'Obvious' to anyone clever/informed enough to give a damn.
I never realized it was a... 'New Issue'!? :) #20 I havent seen this on 612.24 on the 4070ti super. #21
SiliconSeraphSure, it should never have been an issue in the first place, but still, good diligence from Nvidia
When it's on AMD side it's a total disaster somehow for nvidia it's a normal thing. #22
FatloaderWhat’s with all the quick polls recently? Not saying they’re bad, just a bit curious.
I see the use case to find interests of the audience when the data is evaluated every 3 hours and shared. Dropping less interesting topics for those users who visit the webpage and choose an option on the quick vote. The quick votes should include more possible options, or should be adapted by more staff when people write in teh comments some options are missing.
Event HorizonGlad I use Linux, because I would be pissed if I lost a bunch of SSD health to this nonsense at a time like this.
We should stick to windows, the topic was microslop windows.

Anyway. Does windows even has a proper system logger like

syslog-ng
dmesg
sensors
package manager log
all services writing to /var/log

?

nope. and microslop hardly reacts to bugs

--

i read out smartctl etc myself quite regularly. windows is far too complicated need fancy overblown gui with 100mb for the same task
smartctl 7.5 2025-04-30 r5714 [x86_64-linux-7.3.5-gentoo-r1_11_21_2026] (local build)

I do configure my box to write less. I did use the faulty amd driver(s) for more than 8 months in windows 11 pro for gaming with the epic game store. Windows does not work with those 3 swap files or was it up to 5 swap files disabled. I tested this in year 2026 and shared it here in the windows section. 8gib dram is the same as what i use on my other box, but windows fails.

windows has also too many writes to the drive, it is not only the amdgpu driver or the nvidia gpu driver which ruins the storage = C: in microslop windows.
Model Number: KINGSTON SKC3000D2048G

Available Spare Threshold: 10%
Percentage Used: 1%
Data Units Read: 109,466,707 [56.0 TB]
Data Units Written: 51,898,253 [26.5 TB]
Most likely cheap SAMSUNG NAND, QLC, whatever else what people are using.

--

I did download several times those fancy windows 10 / 11 pro add on package from computerbase.de/downloads where the not mentioned process explorer is included. i'm not fond of using winzip gui for every basic task
LogicI have been moving this sort of incessant small writes to HDD (or DRAM/tmpfs) for years
We should stick to windows, else a certain moderator will delete those posts again. like my previous ones

anyway. windows does not support this. It is very old knowledge, dating pack to linux kernel 2.0.0. TMPFS was a lot discussed in year 2008 in certain forum posts elsewhere to speed up computation.

--

NVIDIA fixed the issue much faster. That is also a given fact. That bug was known on the amd side for the windows amd gpu driver for microslop windows over 9 months what i saw, even on this forum.
note: i mostly bought amd related notebooks, ~ 8 different amd desktop cpus, ~6 different amd motherboards, ~5 different amd gpus. #23
RootinTootinPootinwww.nvidia.com/en-us/geforce/forums/game-ready-drivers/13/593169/announcing-geforce-hotfix-driver-62135-released-11/3634901/

RTX 50 series is unaffected though by the previous driver issue..
Enough reports that are suggesting otherwise, including mine with the 5080. #24
_roman_That bug was known on the amd side for the windows amd gpu driver for microslop windows over 9 months what i saw, even on this forum.
This bug first manifested in 26.5.1 and if my math is correct that's 6 months, not over 9 months.

Also based on worst case write and the unlikely fact that someone had not turned off their computer in the last 6 months this bug could have caused at worst 17TB of total writes. Who knows how many of them even landed on the NAND itself instead of being absorbed by some sort of cache like DRAM on SSD or some host side cache. #25
TomorrowThis bug first manifested in 26.5.1 and if my math is correct that's 6 months, not over 9 months.

Also based on worst case write and the unlikely fact that someone had not turned off their computer in the last 6 months this bug could have caused at worst 17TB of total writes. Who knows how many of them even landed on the NAND itself instead of being absorbed by some sort of cache like DRAM on SSD or some host side cache.
I often talk to someone who uses a 7700 XT, apparently they say it's been a thing since RDNA3 release, but was an issue kept silent for whatever reason. The root issue probably has to do with a specific configuration (in software and hardware) since it would have been subject to a fix way earlier if it was widespread, though it could have also been this supposed logging issue.
The timing of reports about this issue in particular just matches with the downgrades from 32GB to 16GB RAM systems (especially on the laptop side) due to current DRAM prices, but it may just be a coincidence as well. Add your own comment
← 목록으로