Fan speed fluctuation - Intel A750

[LadyMikea]

Application [Required]

Intel Arc Control

Processor / Processor Number [Required]

Amd 7800x3d

Graphic Card [Required]

Intel Arc 750

Rendering API [Required]

• Vulkan
• OpenGL
• DirectX12
• DirectX11
• DirectX10
• DirectX9
• Not applicable

Windows Build Number [Required]

• Windows 11 23H2
• Windows 11 22H2
• Windows 11 21H1
• Windows 10 22H2
• Windows 10 22H1
• Other (Please specify)

Other Windows build number

No response

Describe the feature [Required]

Hello,

Bellow is a screenshot from Intel Arc Control (.5382), as you can see fan speed varies from 0 to 600 rpm in relatively short amount of time. Temperatures are fine - 45-48 C in idle. My question is, how this look from fan longevity perspective? Maybe low (or medium) but constant speed is better for card fans, because start-stop cycles can fast wear moving parts(?).

Additional notes

No response

[chkhld]

I am experiencing the same issue.

Had this happen on a Sparkle A750 Orc and Sparkle A310 Elf, although I do recall the A750 being more annoying, maybe because of 2 fans vs. the 1 on the A310.

I followed the ASPM power-saving instructions provided by Intel (L1, Windows power plan, etc.) but the problem existed before I even did that, so I don't see any relation.

It doesn't matter if the GPU is idling at 10-15W, effectively off (due to ASPM and no connected screen) at 0W, or blasting through a video encode at full speed taking up ~29W at full media encoder load. The card is basically quiet, but the fans ramp up and down and up and down all the time.

With the A750 I thought at first it might be an issue with my aging PSU (bequiet! Dark Power Pro 550W) or its cable, but on a brand new NZXT C1200 with new cables it was no different. And since the A310 also exhibits this behaviour, which doesn't even have PCIe power connectors, I guess it's safe to rule out the PSU.

Then I thought it might be my mainboard, under-delivering to the PCIe 4.0 x16 port or something, despite (back then) no other GPUs or power-eating components were installed, except for an i5-14500 (now i9-14900; both non-K). So I swapped a perfectly good Asus Prime B760 board out for an NZXT N7 Z790 (I think it's ASRock based) which DEFINITELY doesn't have any issues providing stable power to its PCIe slots ... and it's the same. Fans spinny uppy, spinny downy, spinny uppy, spinny downy, all the time.

My PC case (NZXT H5 Flow) has an angled fan in the front of its base, it's dedicated to shoveling cool air at GPUs. So I set up this angled fan to blast air at the A750/A310 once the GPU temperature hits 30°C, to hopefully make it run cooler and quieter, lowering the need for the ARC fans to ramp up all the time, but to no avail. The ARC GPU temperature stays solidly around 45-50°C at idle times, and highest at 50-55°C under continuous load, even with the angled case fan putting through at full blast. (Yes, it's spinning in the right direction. 😄 ) The ARC's fans don't care, they just go on ramping up and down in the same intervals as before.

[freak2fast4u]

The GPU fan behavior is out of whack and somewhat always has been, even on Intel LE cards.
Here's on A770 16GB LE :

Here's my fan curve :

The hard part of this type of algorithm is, at the very least, answering these key questions :

• does the fan have to stop immediately as soon as target temperature is reached ?
• if not, how long does it have to stay on ? 1s ? 2s ? 10s ?
• do we even track the time ? or do we simply track the chip temperature ?
• what tolerance to the target temperature is acceptable ? None ? +/- 0.1C ? +/- 0.5C ? +/- 1.0C ?
• once the fan has stopped how far does the temperature have to deviate from the target before it comes back on ?
• when it comes back on, should the fan run slowly until target temperature is reached (if ever) ? or should it run fast so the target temperature is maintained and tracked as close as possible ? somewhere in the middle maybe ?

You can look into Hysteresis and Fuzzy control system if you're curious, though at the end of the day that's Intel's job :p If other manufacturers can get this right, surely Intel and their partners (sparkle, acer, asrock, etc.) will get it right at some point, but Intel ... plz fix.

There's a pretty simple problem to fix first, which is to have a properly graduated graph in Arc Control, and allow settings below 30% other than 0%. Right now, it's either 0% or 30%+, nothing in between.
30% is roughly 1300rpm, which is wierd because ... 100% is 2600rpm, so 30% should really be 780rpm. Unless somehow 100% should be 4400rpm ? I don't remember buying a rocket ;)
1300rpm starts to get noisy already, so having that limitation is problematic in itself.

I personally don't care because the fans are quiet below 1000rpm, so having them yo-yo between 0 and 700rpm is fine for me, but as usual, YMMV.

[rcliftonharvey]

I personally don't care because the fans are quiet below 1000rpm, so having them yo-yo between 0 and 700rpm is fine for me, but as usual, YMMV.

It is quite bothersome to me. I can not concentrate if rhythms or repetitive sequences distract my brain from working.

My entire machine is absolutely quiet. I can hear birds chirping in the trees, I can hear mice roughing each other up in my garden, I can almost hear the worms boxing their way through the soil. But there is zero noise coming from any AIO, fan or other device in my box, everything is cool and quiet since I'm not doing much.

Except for the A310 fans that always spin up and spin down again, "vrooooooaaaaaaaammmmmmffffff", every 30-or-so seconds. Even while I'm typing these few lines, I hear nothing but the clacking of my keyboard, and periodically "vrooooooaaaaaaaammmmmmffffff" from under my desk.

It's distracting and annoying and should get whoever is responsible for this nuisance relieved of their job/s, because this is no state to release a product in.

[i5heu]

On windows one can use custom fan profiles, but for my Linux work station this is not a option.
I have FW version DG02_2.2353.
I think i might wait another day maybe this post here will bring some relief to this mental hazard, but if not i will need to return this Intel Arc GPU because of these Driver issues regarding the fan.

I am a bit disappointed how something like this, that renders this thing useless, gets so little attention over months and months... soon a full year.

#430

[Toetje585]

Same issue on my Sparkle A750, reading all the comments and previous replies from Intel I'm disappointed this issue has not been resolved. Either the people that are affected need a custom vbios with fan stop disabled or they lower the fan stop range 10c down. I bought the card for Linux so I have no fan controll options. Might just return it and never look at Intel again.