With fan curve and 1st two positions set to 30%, GPU fans intermittently briefly drop to 0%

[davejay]

The problem I am reporting appears similar to the one reported here, which was closed due to non-responsiveness from the submitter: https://community.intel.com/t5/Intel-ARC-Graphics/0db-Fan-Mode-Sparkle-Titan-A770-and-Acer-Bifrost-A770/m-p/1542364

Using a Sparkle Arc A770 Titan with 31.0.101.5186 on Windows 11 home, first time driver installation after using DDU to clear AMD drivers from the system. Motherboard is an Asus B550i-gaming, CPU is an AMD Ryzen 5 5600.

This concern is noteworthy because a GPU idling with 30% fans (via fan curve) is much cooler than a GPU idling with 0% fans (via automatic), ~26c vs ~48c on my system, but at 30% the fan noise is loud enough that every intermittent brief drop to 0% and back to 30% calls attention to that noise.

Also, note that I say "intermittent" because the timing between each brief drop is unpredictable -- generally between 20-120 seconds -- but these drops happen consistently whenever I set the fan curve leftmost positions to 30% and 30%.

Steps to reproduce:

- Have fan speed control set to "fan curve"

- Have two leftmost fan curve percentages set to "30%" (lowest possible non-0 setting, see attached image of fan curve in Arc Control)

- Let system sit idle (~2% CPU usage in task manager)

- Observe GPU fan speed via Arc Control, Afterburner, your eyes and/or your ears

Expected behavior:

- GPU temperature holds steady (at ~26c on my machine)

- GPU fans hold steady at ~30%

Observed behavior:

- GPU temperature holds steady (at ~26c on my machine)

- GPU fans hold steady at ~30% with intermittent brief drops to 0% (see attached image of fan tachometer from Afterburner)

Impact of issue:

- GPU fan noise is quite audible at 30% and inaudible at 0%

- The transitions from 30% to 0% and back to 30% are noticeable on a system at idle when muted and constant-speed case/CPU fans are running

- The audible nature of the intermittent brief drops to 0% are therefore noticeable and annoying

Not tested:

- I did not test to see if this happens when the two leftmost positions were set to 31%/31% or any other equal value

- I did not test to see if this happens when the two leftmost positions were set to similar but non-exact values like 30%/35%

[davejay]

As I have not seen a response yet, I did some digging and found this thread, which also appears to be reporting the same thing: https://community.intel.com/t5/Intel-ARC-Graphics/Performance-alternating-coil-whine-fans-turning-on-off-all-the/td-p/1564631

As you asked for some information in that thread (that does not appear to have been provided), I've provided that info here for my installation:

• Power Supply Unit: Corsair SF750
• How did you install the driver mentioned? Intel installation utility after using DDU to clear AMD drivers from previous card
• Is it a newly built system? No
• Is it normally working before? The system, yes; the brand-new Sparkle Arc A770 card, yes *except* for the issue noted
• If it does, may we know the recent hardware or software changes made before the issue happened? See above
• How is your system utilized? Gaming, productivity, sharing this information with you  
• Have you enabled V-sync? Yes
• Please provide the SSU logs, kindly refer to this article for your complete guide, How to get the Intel® System Support Utility Logs on Windows* and attached the txt file here. -- attached you will find the SSU output and the logs from C:\Users\<your_username_here>\AppData|Local\Intel\IGN\logs as noted in the other thread

Hopefully this plus my screenshots will help you diagnose and repair.

[佑綺_Mayu]

Have you tried driver version 31.0.101.5194?  

There is a possibility that the physical fan controller of hardware itself is having an issue. This issue also appears on the Acer Bifrost Arc but that's because the driver is customized by Acer and you need to use the acer control center and the driver in their website for the fans to actually function correctly

[davejay]

Thank you for the suggestion. I had WHQL Certified Driver Updates turned on, so did not realize there was a newer driver.

I installed the driver (after a DDU uninstall) and unfortunately the behavior persists. I also tried setting the Fan Speed Control to a fixed 30%, and the behavior persists.

[davejay]

One more data point: I turned off PCIe L1 in the BIOS, in case this was something related to entering power saving mode, but the behavior persists.

If anyone here can reproduce with a Sparkle Arc A770 Titan or A750 Titan card, I would be interested to hear that, to help me determine if this is something justifying an RMA or something that many of these cards do.

[hairbag]

I have the same card and it does the same action with the fan curves. It is most likely the code in the drivers. It is very annoying and I hope that Intel will take the problem seriously. 

[davejay]

Thanks for sharing that about your card. I would appreciate one more data point, if you can provide it, to help me decide if I'll RMA the card, return it completely or live with it as-is.

Attached is a small WAV file (attached as a tar.gz due to this forum restricting WAV file uploads), recording my A770's card spooling down from 30% to 0%, and then back up to 30%. Listen for a grinding noise you hear most obviously when the fans are spooling back up from 0% to 30%. I've ruled out the fans themselves -- I can stop individual fans and it doesn't impact the noise -- so I'm trying to figure out if this an unusual sound that RMA or replacement might eliminate, or if this is a typical sound for these cards (making an RMA pointless).

Does your card make the same sound?

If you're not sure how to extract files from a tar.gz archive on Windows, you can do it from your command prompt -- try this instruction set: https://pureinfotech.com/extract-tar-gz-files-windows-10/

[hairbag]

I listened to your file, while my card sounds similar in action of the fans cycling on -off, mine is definitely just noise from the fans.

[davejay]

Very helpful, then I will likely follow up with my RMA to get one that doesn't have the grinding noise. Thank you for your help!

[davejay]

Just to follow up: original card is on its way back to the retail channel, picked up a replacement card from a different channel (to reduce the odds of getting one from the same manufacturing lot), and the grinding noise is somewhat less, although still present if I listen closely for it. I'll chalk that up to being the noise these cards make.

Also: on driver 5330 (after using DDU) the 30-0-30 behavior is still there, which is a shame. Hopefully Intel will eventually figure that out and provide an update to fix it. Thanks again for your help.

[佑綺_Mayu]

I have a A770 LE and a Bifrost A770. The LE works fine with the fan controls however the Bifrost causes it go really wild and spin extremely loud ( yes I hate the noise too ) but if I use the actual driver from Acer and their Acer Control Center, the GPU acts normally perfectly runs at the correct settings. 

My take on this is that the Bifrost originally has no 30% limit which was only imposed approximately one year after the card was launched at the tail end of 2022. Since the blower and axial fans are synchronized, Acer would have tuned the curve down to mitigate noises at idle. This is now undone by Intel Arc Control. 

Basically, there is a synchronization issue with fan configuration with the LE and OEM Arc GPU's since the OEM( Sparkle, Acer, Asrock) customizes the actual hardware so there is probably an algorithm discrepancy so yes it's a normal behavior unfortunately the fan starts to go haywire. 

[ochompskr]

having this exact problem with a SPARKLE ECO Arc A31 i just bought.  

looks like i have to return it and get an nvidia card

[Jorn-T]

Same behaviour here on a Sparcle A770, as I've described in a different thread, a few screenshots of the fan behaviour attached, now with the latest driver version. 

One problem is having the fans at 30%, with lots of drops down to 0 RPM, a different problem is having the fan curve points 0 and 1 to 0%, points 3 and upwards at 30% or more, which gives lots of fan spinning up to 1300+ RPM for a second or less, then spinning right down to 0% again. Both are extremely annoying and loud.

I found that setting point 1, 2 and 3 to around 35% makes the fan spin without those drops, but at 1500ish RPM the three fans are loud and noisy even at idle..

It should be very easy to fix, one or two lines of code, unfortunately, I don't think the Intel people read our posts or even take their time to view the attached screenshots, they just press F1, F2 or F3 giving us some pre-generated replies.

Will return my card now, and change it for a Nvidia card. 

[ochompskr]

i wonder why this isn't happening to ALL Arc cards.  do they just not notice the noise?  assuming they are constantly above 30?  or is it specific to sparkle/bifrost cards.

[Jorn-T]

I think the error might have bee. introduced with a firmware update that came along with the drivers towards the end of 2023?

[ochompskr]

if i revert to arc control center that came with driver version .4090, i can set the fan curve to below 0 and the constant on/off ramp goes away.  the software is glitchy and terrible around that time but it works to verify at least.

im still experimenting around but it seems to coincide with the 30% change they made, which doesn't appear to be firmware but rather than arc control center.  speculation on github appears to be that the sparkle/biofrost cards don't know how to handle this 30% minimum, resulting in the loops.  

[ochompskr · Accepted solution]

for what it's worth, ive been in contact with sparkle and they said they are making a FW change but it will take up to 2 months to trickle down from intel.  im returning the card as im not willing to risk the wait but a FYI for those holding on

[davejay]

Very helpful, thank you. I'm going to hang onto mine, knowing that a solution is on the way.

[yosiox]

Hi

I recently purchased a Sparkle A310 ECO and was frustrated by poor fan control.
I found a support email address on the web, so I reported the current situation, and received a reply that they were aware of the problem, planned to update the BIOS, and told me to wait for Intel's driver update.

I just wanted to use QSV for AV1 encoding, so I'll wait for an update.

[davejay]

Over four months later, I'm pleased to report that my card is no longer doing this. The fans now sit contentedly at 0 rpm until the temperature gives it a reason to spool up the fans (I noticed with driver 32.0.101.5762, but it may have been fixed in an earlier version).

[davejay]

Aaaaaaand now, another month and a half later it has returned, not sure which driver version. Ah well.