PC Drivers & Software

[amdwtf]

I encountered a severe problem on my laptop HP OMEN 9 2023(AMD 7840HS + Mem 32GB + nVidia 4060).

3 non-pooled tags MiP2, Proc and EtwS just eat more and more memory, and never free any. Within 48 hours, memory usage will become 90%, and only thing I can do to alleviate this is reboot computer.

This problem disappeared after I did a clean uninstall of the driver using AMD's cleanup utility tool.

And it is returned after reinstalling the latest driver.

All these make me think the memory leak problem is caused by AMD driver.

At this point, I'm in desperate need of any help.

Thank you.

[tomoiRegr · Accepted solution]

Just updated from 23.11.1 to 24.7.1. Ran a full cmd processes test with no noticeable abnormal in memory. Might give it a try. 

[gflash · Accepted solution]

I tried the 24.7.1 driver as well. It has been running more than 24 hours and I don't see "Proc" appearing in poolmon yet.

I think this may be fixed...though I don't see anything in the 24.7.1 release notes.

[gflash]

I am seeing the same issue on my system (Beelink GTR7 PRO - AMD 7940HS + Mem 32GB). I have the latest drivers installed.

[hazart]

I have a ThinkPad P14s Gen 4 w/ 7840U and an IdeaPad Pro 5 (16″, Gen  w/ 7840HS.  With recent AMD drivers, including 24.1.1 (Jan. 2024), both systems have memory leaks through zombie processes.  The last-known good driver is 23.10.2 (Oct. 2023).

You can find more information about this type of memory leaks here: https://www.reddit.com/r/AMDHelp/comments/121wtde/amd_drivers_memory_leak_through_zombie_processes/

I'm wondering how long it will take for AMD to fix it.

[amdwtf]

Could you share some info about where can I download v23.10.2?

[hazart]

You can find it on this page: https://www.amd.com/en/support/previous-drivers/apu/amd-ryzen-processors/amd-ryzen-7-processors-rade...

[evil_doer]

I run a 7950X3D and a couple of months ago(? I forget when exactly) I started having a memory leak that wasnt traceable with task manager. And randomly figured out that disabling the internal AMD GPU halted the problem from happening. This issue will eat up all 32GB of memory within a day. 

I dont know what any of this means, but found the same program and processes and showed what mine is showing at the moment (I am at ~75% memory used)

[evil_doer]

3 hours later. ~95% memory used

And now I need to reboot my computer because the ram is about to run out.

[evil_doer]

I installed 23.10.2 (oct 19) as someone suggested and no longer have the memory leak issue.

Please fix the software.

[amdwtf]

We probably shouldn’t have high expectations, since no one from AMD has responded to us.

Zen4 is cheaper than Raptor Lake, and you get what you pay for I guess.

[Dragokar]

Contact support via the bug report tool in the driver!

[evil_doer]

24.2.1 (feb 26) still has the memory leak

I've decided to bump this topic on every driver release until its addressed.

[Dragokar]

Thats fine but please also send in reports when a new driver hits.

[amdwtf]

24.3.1 still has the memory leak.

[evil_doer]

You can say that again.

I installed the newest driver, rebooted, then about an hour later decided to record how much ram I have left over time. This graph shows from around noon, to 7:30pm where I was already running on fumes.

The spike is just me playing a video game. The flat line in the middle is it running nothing at all, I wasn't even home. Computer was just sitting idle.

Please keep sending bug reports via the app as well!

[gergalyb]

I ran into the same issue, and after many many hours of troubleshooting I was able to pinpoint the iGPU driver of my Ryzen 7900. I didn't note down the exact driver version, but I think it was v23.xx.
What I found was that if the iGPU driver is enabled, it will prevent windows from properly exiting from ANY process, and thus the page table slowly gets filled up and fragmented.
What is weird that these zombie processes don't have any handles, so it was pure trial and error finding the source. 

I have a 3060ti, so I guess I'll just leave the iGPU disabled indefinitely...

[Demiu]

Uninstall or just disable (im assuming in dev mgr)? Have you seen the improvement immediately or only after reboot?

[gergalyb]

First I disabled it in device manager, the memory leak was fixed by that, but SteamVR freaks out if you have a disabled GPU, so I had to disable it in the BIOS instead. No problems since.

[gflash]

I hope they finally fix this in 24.4.1 which should be imminent.

[evil_doer]

Not fixed. Ill provide a graph later today of the free memory dropping

[amdwtf]

you save me 10 mins of my life

[evil_doer]

Here is a graph of my memory being eaten by the AMD driver in the latest version

 

A reminder to please report this issue in the Adrenalin driver app:

 

[evil_doer]

Sigh... 24.5.1 still has the memory leak.

I wont bother with a graph, but I did install the driver last night and woke up to 0 memory free this morning.

Please report this via the app as always.

[set]

The latest good version is 23.11.1, not 23.10.2.

I've posted here how you can immediately check if the issue is present or not.

[evil_doer]

Thanks. This seems to be the case.

Unfortunately I ASSuMEd someone was correct earlier in the thread and didnt confirm myself. So Ive been reporting that 23.10.2 was the last working driver to AMD. I hope they havent been putting too much work to see what was wrong in 23.11.1 when theres nothing wrong with it.

I will do another report and point to your other thread showing how to confirm the issue.

[amdwtf]

I don't think you need to worry about that.

They probably didn't put into any work, if they even acknowledge this problem.

[Soupar]

Thanks for posting this issue. I had this problem with my new Laptop (7540U) using even the newest 24.5.1 driver and had to revert to 23.11.1 - which solves the memory leak.

I am frankly astonished something like this isn't solved over half a year and multiple driver versions, esp. because I'm running a very recent Zen4 APU. I've reported it to AMD with their bug report tool.

[felle64]

Have anyone tested the new 24.6.1 and see if it still has got a memory leak? I’m still on 23.11.1 but I wanna upgrade for anti lag 2 to CS2. I leave my pc on quite often so a memory leak is not ideal.

[war59312]

Still crap.

[blade888]

Good to know I'm not going insane over this issue,

Have the exact same issue on my partner's PC as she leaves hers on all the time.

We both run:

Ryzen 9 7950X

Asus x670e Hero MB

32GB Kingston DDR5-6000.

Samsung NVME SSDs. (500GB and 2TB)

However: Her PC has an Nvidia RTX 2070 Super, my PC has an AMD Radeon 7900XTX.

I left my PC on for a week along side hers as an experiment. I don't get the same Issue with Proc going nuts on consumption but I do see a similar RAM usage creep (i'm at 50% she's now at 75%) but not as extreme.

Will try disabling the iGPU on her PC and see if that solves it. It may be a simple conflict between the two Vidcard Vendor Drivers making her RAM issues more prominent.

[evil_doer]

Another new (broken) driver.

Don't even bother, folks. Its STILL broken.

[war59312]

To be clear, you mean 24.6.1 is still broken, yes?

[evil_doer]

Yes.

[war59312]

Please report this via AMD Bug Report Tool

Only way going to get attention.

[war59312]

Hoping Linus can bring some additional attention to this issue.

See https://linustechtips.com/topic/1575191-known-zen4-memory-leak-on-windows-11-video-to-get-amd-to-fix...

[war59312]

Please vote up:

windows - How to identify a driver causing every exited process to become a zombie, polluting page t...

[blade888]

In my Partner's PC case disabling the Radeon iGPU on her 7950X solved the issue. Her RAM now sits at 29% consumption at idle now for 2 weeks up time.

Keep in mind she was running just the iGPU driver autodetected with version 31.0.24002.92 in device manager. (not the full Adrenilane suite) Because she has an nvidia Geforece RTX as her primary.

On my PC I left my iGPU on but I run the full Adrenilane suite 24.5.1 with my 7900XTX. I get up to 50% consumption and it stops and levels out and doesn't go higher on idle after 2 weeks. 

In my case i never got the "Proc" Process Objects filling up in PoolMonX like my partner's did. I only get "EtwB" ETX Buffer being the largest item but never filling up the RAM.

My conclusion in my experience of this is:

AMD GPU driver from 24.1.x from the beginning of the year has the leak

AMD GPU drivers don't like running along side nVidia ones.

AMD GPU driver auto detect/install alone without the full Adreniline suite has the leak 

[hazart]

Recent OEM drivers for ThinkPad P14s Gen 4 w/ 7840U and ROG Ally w/ Z1 Extreme (a rebranded 7840U) have no memory leaks. I assume that the OEM's (Lenovo and Asus here) have already found a solution.

[pwalski]

Did you mean this driver https://pcsupport.lenovo.com/us/en/products/laptops-and-netbooks/thinkpad-p-series-laptops/thinkpad-... 
?
It still has driver 31.x (31.0.22048.10001)

Asus also uses 31.x https://rog.asus.com/gaming-handhelds/rog-ally/rog-ally-2023/helpdesk_download/ 

Adrenaline v24 comes with 32.x.

[hazart]

Yes!

I looked at AMD's May release (Adrenalin 24.5.1), the INF file says 31.0.24033.1003.  The INF file in AMD's June release (Adrenalin 24.6.1) says 32.0.11021.1011.

The latest drivers from Lenovo and Asus are somewhat out of date, but they no longer have memory leaks.

[pwalski]

In my case Adrenalin 24.6.1 did not help, so I reverted back to 23.10.2

My workflow includes usage of WSL2 and Docker, and I found that memory leak is caused by AMD drivers thanks to these threads
https://github.com/microsoft/WSL/issues/4166#issuecomment-2220297331
https://superuser.com/questions/1838566/how-to-identify-a-driver-causing-every-exited-process-to-bec...

(23.10.2 is not ideal either because laptop (Asus X16 with Ryzen 9 6900HS) sometimes fails to wake from hibernation, but I prefer it over memory leak)