Repository navigation
Replies: 1 comment 1 reply
|
Can you translate the post in English, please. I don't think any of the maintainers speak French. |
1 reply
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
That's exactly what I observed. It should pause without overwriting the settings.
I agree with you, and from a technical standpoint, that's actually the solution I would have preferred.
The current behavior is:
The problem is the last step: writing to NVS. It turns a temporary protection mechanism into a permanent configuration change.
In my opinion, it would be better to have this logic:
Or, even better, provide an option in AxeOS:
That way, everyone could choose the strategy that works best for them.
Why would this be a better approach?
On an overclocked Bitaxe, a temperature spike can be caused by:
a particularly hot day,
a temporary reduction in airflow,
a temporary increase in ambient temperature.
Once the temperature returns to normal, there is often no reason to permanently lose 100 MHz and 100 mV. A simple pause of a few dozen seconds is enough in many cases.
I think this is worth proposing as a fix.
It's a relatively localized change:
replace the NVS write with in-memory (RAM) storage of the original parameters;
restore those parameters after the device has cooled down;
only modify NVS if the user explicitly requests it.
This would preserve thermal protection while respecting the user's settings.
I would even go a step further: prepare a small patch for AxeOS 2.14.2 that adds this behavior. Based on what you observed on your two Gammas, I think it could be useful to a significant part of the Bitaxe community.
Here's the patch I would propose
I would propose a patch that doesn't change AxeOS's overall safety philosophy, but only changes how it handles the post-overheat condition.
Option 1 (my preferred approach): Pause + resume
Principle:
Overheating detected.
Stop mining.
Force the fan to 100%.
Wait for the device to cool down.
Resume mining using the user's settings.
There is no NVS write.
Pseudo-code:
No user settings are lost.
Option 2: Temporary throttling
This is what many professional systems do.
Then, as soon as the temperature drops:
Again, without ever writing to NVS.
Option 3: User-configurable behavior
Add the following to Settings:
This is probably the best solution.
What I would change in the code
Right now, you have something along the lines of:
I would replace those lines with:
This way:
the frequency actually decreases,
the voltage actually decreases,
but nothing is saved to NVS.
Once the device has cooled down:
and it immediately goes back to 600/1150.
Even better: an "overheat counter"
I would take the logic a little further.
On the first overheat:
If it happens again immediately:
After 3 consecutive overheating events within 15 minutes:
But still without ever modifying NVS.
This means the Bitaxe recognizes that the overclock is no longer sustainable (for example, because the room temperature is 35 °C), while still preserving the original settings.
The next day, when the ambient temperature drops, it will automatically return to 600/1150 without any user intervention.
Why I prefer this approach
For someone running an overclocked miner, the settings represent a profile that has been validated after hours of testing. Permanently replacing them because of a temporary heat event means having to regularly go back into the interface and restore the values.
In my opinion, thermal protection should protect the hardware, not permanently modify the user's configuration.
I think a patch like this would have a good chance of being accepted by the ESP-Miner developers, especially if the current behavior is not intentional but simply a consequence of the way thermal protection was implemented.
All reactions