Lack of a third fan reported by drivers is really starting to cause issues

Graphics / Linux Linux
,
22

I thought this may have been posted here as well, but I can go ahead and tell everyone that this problem is actually 100% unfixable and will never be solved - though it shouldn’t NEED to be solved, because using anything to control your fan speed like GreenWithEnvy DOES change the fan speed for all 3 fans now.

Anyway, it’s impossible for Nvidia to implement fan monitoring/control for 3 individual fans, because the Nvidia drivers DO NOT control the third fan. Nvidia’s drivers only support two fans (think about it, none of their own GPUs have 3 fans, only 2). I received an email in regards to this bug report (since I’m the one that filed it) and it turns out that the third fan’s control is implemented through the AIB partner.

This is EXACTLY why MSI Afterburner on Windows ONLY let’s you monitor/control 2 fans for, for example, an EVGA RTX 3090. Or a Gigabyte Gaming OC 4080. It ONLY shows 2 fans.

But If I launch Precision X1 from EVGA, my XC3 Ultra 3090 WILL show all 3 fans. It’s a firmware/software thing.

So unless these GPU AIBs release Linux software, it’s impossible.

@BlueGoliath NVML doesn’t even have anything to do with it, either.

23

but I can go ahead and tell everyone that this problem is actually 100% unfixable and will never be solved

If it’s possible to interface with custom RGB controllers on Linux then I doubt this is true.

though it shouldn’t NEED to be solved, because using anything to control your fan speed like GreenWithEnvy DOES change the fan speed for all 3 fans now.

This isn’t the case for the second fan on EVGA’s ICX cooler.

Nvidia’s drivers only support two fans

Doubt that. NVML has no such hard limits on the number of fans that are exposed. Other Nvidia APIs do have an internal max number(3, ironically) of “coolers” a GPU can have but since it’s an internal thing that number could probably be changed as needed.

. I received an email in regards to this bug report

They can be bothered to answer you but couldn’t be bothered to answer the various threads / replies about rendering / crashing issues in 525 / 530 drivers, including me. Really @aplattner? People might be throwing away perfectly working hardware thinking that it’s defective meanwhile Nvidia can’t do the absolute minimum by just saying “This is likely a driver issue and we are looking into it” or something.

@VulkanGuy is right. The communication sucks.

it turns out that the third fan’s control is implemented through the AIB partner.

Knew that already.

NVML doesn’t even have anything to do with it, either.

Nvidia probably could work with board partners to expose AIB-controlled fans. Would they be willing to? Probably not.

24

I am pretty sure most AIB RGB controllers are communicated with via i2c, I know that’s how mine is done, which is why it was able to be made to work with OpenRGB.

Doubt that. NVML has no such hard limits on the number of fans that are exposed. Other Nvidia APIs do have an internal max number(3, ironically) of “coolers” a GPU can have but since it’s an internal thing that number could probably be changed as needed.

This has nothing to do with NVML. It’s not NVML that controls the 3rd fan on Windows. It is 100% outside of Nvidia’s drivers/APIs.

Like I said, everything I posted was directly from Nvidia. I still have the email. I’ll consider posting relevant ■■■■ here.

They can be bothered to answer you but couldn’t be bothered to answer the various threads / replies about rendering / crashing issues in 525 / 530 drivers, including me.

I mean, the 530 drivers haven’t really been out that long, and rendering/crashing issues are COMPLETELY different to diagnose than an issue you can immediately point to, such as “yeah the 3rd fan is not controlled by us, the AIBs handle all that, and it’s controlled through their firmware.” They’d have to actually diagnose the crashing/rendering issues first (actually, first they’d have to even reproduce the issue).

Also, I had opened an email bug report at the same time as I posted here, and I kept emailing them until they replied, and even then it took a couple months. 530 has only now been out like a month and a half or something.

I’m not defending their notorious seemingly random decision-making when it comes to whether they reply or not, I have had the same frustrations with other issues too. Honestly I think they should figure out a way to have more consistent bug reporting. The open kernel modules GitHub is a good start, but they close bugs that aren’t limited to the open gpu modules. I think that’s bullshit, if the bug exists in the open drivers people should be able to post bug reports there.

So yeah, I totally agree with you that their slowness to respond to some people on bugs is unacceptable, but it honestly does seem like you reacted without actually thinking about how they’re two completely different issues, and mine took an unacceptable amount of time to be resolved even when they immediately knew what the issue was.