Developers? maybe you can help fixing this GPU bug on AMD Advantage Edition Laptops (G513QY)

[ryzenmaster]

To Developers!

As a developer.... reaching out to see if you could help us,

I am one of the many owners of the ASUS AMD Advantage Edition laptop G513QY, there is a whole discord server community of this laptop owners that suffer from GPU related reboots issue.

Basically, there is a bug that so far neither ASUS or AMD fixed. Instead people getting ASUS RMAs and new units getting same reboots (at least few people reported to be the case).

Now at the Discord community we found workaround, and we thought maybe you could help fixing it at the AMD control panel level.

How replicate the issue on this particular laptop model:

1. Run Blender GPU Render
or
2. Run a heavy game

Within 1-30 minutes, the dGPU (RX6800M) would go sleep and laptop crash and reboots.
It --seems-- as the system not properly transitioning the GPU tasks from dGPU to iGPU which causes this crash and reboot.

Workaround we found:

1. Run GPU-Z in background and keep it running while and after gaming/render, will keep RX6800M awake and avoid reboots.
2. Installing ASUS GPU Tweaker 3, will keep RX6800M Awake and wont reboot.

Problems:

1. GPU-Z as long as running, will keep rx6800m awake 24x7 which reduces battery life as iGPU wont be utilized.
2. ASUS GPU Tweaker 3 will keep rx6800m awake 24x7 which reduces battery life as iGPU wont be utilized.

Solutions:

If we dont use ASUS GPU Tweaker 3 and instead sticking to GPU-Z we close GPU-Z whenever we need long battery life (lets say working outside) and keep it ON when we plugged to AC. Alternatively, if we want game while on battery and don't want a reboot to occur, we can keep GPU-Z open.

Now that you understand the problem, if you could work-out a solution within the driver/control panel/maybe some little tool in task-bar with that we could press on/off, to make a solution more reliable and easy for everyone, would be interesting.

We got a great community that would beta test this for you/with you, discord: https://discord.gg/NwWySb43
with many owners of this laptop from worldwide, you can chat to everyone, ask beta testing on their units and work with community on a solution

I will truly appreciate your feedback, we hope for the best from you soon!

Thank you,

 

[Naki]

What a curiously horrible idea!
Anyway, good luck with that.

 

[ryzenmaster]

As per my little chat with @W1zzard, basically, the community guys asking if we could get something bit more friendly, let's say something that runs I'm background, maybe with a little menu at (clock area in windows) and shortcut keyword combination, for example:

Ctrl+x+1 = Gaming Mode (Reboots Protection ON)
Ctrl+x+2 = Power Saving Mode (dGPU can sleep)

Maybe one could even see the deeper root cause to the reboots (can also join our discord with many users with this laptop there), could remotely debug few units, maybe could come up with even more suffice fix.

At this moment guys already almost fed up with hope for Asus or AMD, although we all still waiting. Maybe the rescue will come from TPU community

Thank you so much !

 

[W1zzard]

Maybe you could even see the deeper root cause to the reboots

This seems like just another kind of AMD driver issue that will probably never get fixed

let's say something that runs I'm background, maybe with a little menu at (clock area in windows) and shortcut keyword combination, for example:

Good idea

So without bother you further as I imagine you are busy, if you can give any idea for the community on what to expect, would be awesome, many people waiting for good news and be sure, Community will help you to test it before getting to more public.

No idea on these details yet, I like the concept, but finding time isn't easy. Will try to get something working for you this year

 

[ryzenmaster]

This seems like just another kind of AMD driver issue that will probably never get fixed


Good idea


No idea on these details yet, I like the concept, but finding time isn't easy. Will try to get something working for you this year

Thank you so much!!!!

This seems like just another kind of AMD driver issue that will probably never get fixed


Good idea


No idea on these details yet, I like the concept, but finding time isn't easy. Will try to get something working for you this year

@W1zzard community guys shared one more hypothesis, something for you to consider while developing the solution.

We suspect that originally, once game or render done and dGPu prepares for the sleep state, when dGPU goes power saving mode, something fails in the process, maybe a command to go sleep causes the gpu to hang and this causes a hang and a reboot.


How user with a unit like this that has reboots can replicate this theory:

1. Once after you run a known program (game/render) that causes the reboot, you can get it to hang/force reboot by disabling and reenabling this PCI device:
   

   http://imgur.com/dZTs5W5

2. As long as the pcie port does not hang (no reset command issued), the gpu will continue working.
3. But, the moment you try to issue a pci device reset
4. it reboots/hangs

As we see it "low power state" is broken and causes a reboot... and keep dGPU idle make it consume 6W which cuts battery life by 50%... so we somehow need dGPU's "Low Power State" to work without cause crash/reboot.

To clarify, dGPU has few modes, as we know it:

1. Idle (no use but alive)
2. Low Power State

So keep it idle/alive would still consume approx. 6W which causes 40-50% on-battery life cut, but if we can fix so Low Power State works and doesnt causes the reboot issue, we will gain back the battary life and avoid reboots issue.

GPU-Z and ASUS GPU Tweaker 3 keeping dGPU alive, avoids from it going "low power state" (LPS = as we suspect has a bug which causes it to reboot) which means, even if nothing uses the dGPU (Idle) it still eats approx. 6W and hence cuts battery life by 40-50%.

Which leads Josh from our Discord community to following theory, The GPU should never go into power saving mode whenever a app is using it. But in some apps, that is possible. Hence, the bug is this: Whenever an multithreaded app that is using the 6800M is suspended (debugger attached or for whatever reason), and the GPU goes into power saving mode, then resumes after. It will cause the failure to reset.

In other words, whenever an app is using the GPU and does not pause using the GPU (or the GPU does not go into power saving mode) it will not cause this bug.

So hopefully, we can fix it, so dGPU could enter low power state after gaming/render sessions, and doing so without crashing.... if this not possible to fix, then we need the solution that would basically put it to sleep/idle, less effective as this one would keep taking 6W ... and shorten battery life.

NEW-WORKAROUND by Josh from Discord Community:
" My final version of this hack fix, so that you dont need Disable ULPS or GPU Tweak or GPUZ. This app will keep the gpu alive when AC Power is plugged in. When AC Power is not plugged in, it will stop keeping gpu alive to allow gpu to go to power saving mode. The opposite happens with ac power plugged in. Its resource usage is very low, less than 5 MB RAM, and barely any cpu/gpu usage. You can add this to task scheduler to run on startup. WARNING: Side effect is the laptop will never successfully sleep on AC Power NO SUPPORT PROVIDED. Use at your own risk -- Download this tool: https://t.me/AMDadvantage/61 mirror: https://mega.nz/file/PtFADDiQ#Bb-nrC3h55u5X-aj8XZ6KNoP6Q1qrC8XDz38yVl7jKI "

@W1zzard sharing this with you, as it might be additional support to your development

Thank you!

 

[ryzenmaster]

** Josh released updated version that fixes potential memory leak:

File on MEGA

mega.nz

For further updates please join the Discord where Josh sharing it https://discord.gg/NwWySb43 as I can't keep and edit here for each update he shares**

Now all we can do is wait and hopefully before end of this year @W1zzard will come up with some magic for us