Alternatives to IPMI Fan Controller (Need Slower RPM): Low Noise Adapter vs. Additional 3rd Party Controller?

The Problem: I’ve posted elsewhere re: my inability to lower my fan speeds below 20 percent of full power due to the IMPI’s hard floor: Asrock X570d4u, x570d4u-2L2T discussion thread - #671 by johntdavis . I’m trying to use consumer-grade actively-cooled CPU components with “let’s take off for the sky” server fans meant to provide airflow for passive cooling, and at the IMPI fan controller’s lowest RPM (20 percent duty), it’s just way more airflow than I need. I need to slow down the fans some other way.

Question: Does anyone have any opinions on or experience with doing this using any of the three options below:

  1. IPMItool: I can’t figure out how to control individual fans with this yet. It’s very complicated and I haven’t found a good idiot’s guide yet. :stuck_out_tongue:
  2. Noctua low-noise adapter: Simplest hardware solution; should drop speed by 20-30 percent? I’m looking at these: NA-SRC7
  3. Add-on fan controller: Something like this: NA-FC1

Noctua really doesn’t want you using the low-noise adapter on non-Noctua fans. I’m wondering if it’s possible to damage those fans if you do.

The add-on fan controller does exactly what I want, but costs $25, and I need three of them since I’m wary of running three high-powered Supermicro fans off a single motherboard header.

I’d like to try IPMItool, as it’s free, but I can’t find an example that doesn’t involve controlling all the fans as a single entity (e.g., set all case fans to x percent). I need to be able to control fans individually, as I’m running rear exhaust and front intake at different speeds because they’re different models of fan.

I also need to get this done so I can get the server off my table and into the rack, so I’m very tempted just to throw hardware at it if that woud be effective.

Thanks!

I have used the low noise adapters on non noctua consumer fans but never server fans… even 2U server fans can pull 20+ watts each at full chooch and that would be a lot to ask of that little adapter to handle. I believe you’d end up toasting the adapter, not the fan.

I was a bit worried about that, too.

I think IPMItool is the safest way to go. I just can’t figure out how to use it. :stuck_out_tongue:

Noctua LNAs are just cables with a resistor soldered in the middle. Looking online the LNA is 50 ohms. 20W at 12V means 1.66A. Plugging that into Ohm’s law and electric power calculator we get I^2 * R = (1.66^2)*50 = 180W. The insulation on the LNA will melt. Or the resistor will fry. Or some other bad thing will happen. Anyway: don’t use an LNA with high power fans. This math is incorrect - it doesn’t take into account that the fan will draw much less power with an LNA, lowering the current through the resistor. I have no easy way to calculate that.

I’d get that Noctua controller and a powered splitter like AquaComputer SPLITTY9.

Or get something like Aquacomputer OCTO - it does up to 25W per channel and 100W total, and is powered via molex. The software is Windows only, but if you use a temp probe connected to the controller, it can work the curves independently. A hwmon driver for it was merged in 5.18-rc1, unfortunately it’s only for monitoring. The software isn’t entirely free - you only get half a year of updates with the OCTO. You can still download an older version, but just no updates.

2 Likes

I have successfully used these for this purpose on some servers in the past. Just be careful not to lower it to the point where the fans never engage as the mobo will think they are on and running regardless. It’s a little too easy to dial them down past that threshold IIRC.

2 Likes

I was about to mention that one as well. I have three of those in my backup 4U chassis to quiet down 3 Nidec Ultra Flo 80mm fans (just under 1 Amp each) when the machine is running. They work great and can handle up to 3 amps.

1 Like

I may be cheap, but i see this as easily fixed with 10 cents worth of resistors and a soldering iron.

Unless my math above is very wrong, I’d be very wary of resistors on such a high power fan.

Also, you assume OP has a soldering iron and knows how to solder. Electric tape doesn’t do well with higher temperatures, so you’d need some heat shrink tubing to isolate it too.

1 Like

They make resistors that can work for different wattage ranges, the second part is a variable though.

Yeah, but 10W are somewhat bulky. Depends whether you care about it.

…I don’t know why I didn’t get notifications for these responses sooner. I already bought three of those Noctua controllers. These are perfect.

Results:

At 20 percent power, via the IPMI’s fan controller, the front blastomax 9000 fans ran at 2200 or so RPM, and sounded like a small jet trying to lift off. I could be in the same room with it without damaging my hearing, but it wouldn’t be too much fun to live with.

So, I hooked up a fan controller to each, and set each to half power. (NGL, I love just being able to dial them in. Retro.)
I expected ~1100 RPM, but they sit at ~1300 RPM. These fans are powerful.

CPU and VRM were still idling in the 30s.

I booted into Windows, downloaded Cinebench 23, and set it to run a continuous 30 minute “Stability Test.”

Reported temps after 28 minutes:

Temps hit 60-63 C pretty quickly and stayed in that range the whole test.

I’m new to fan tuning and temperature control–this is my first server–but my understanding is that a 30 minute Cinebench run is a worst case scenario for a machine that will be used as a Proxmox server and won’t be doing any sort of GPU work at all–it doesn’t even have a GPU.

Does that sound reasonable, or am I missing something important?

(I feel like I might be able to push the fans even further, down to 1000 RPM each, but OTOH, I don’t want to push my luck. I still haven’t figured out a good way to monitor the SSD temps when the whole system is under load, but the SSDs are all in the front of the chassis, directly behind the fans, and should get the coldest air of any component.)

Notes on Fan Controllers:

Yes, I know I could theoretically split them all of a single cable, and use a single controller, but I’m wary of drawing that much power through a single fan header. These are 3x0.58A 12v fans, and even at half power, they’re still blowing at ~1300 RPM.

The chassis and power supply are also ~11 years old, and all I’ve got to power things inside the case are EPS and MOLEX. No SATA power at all.* Everything is working, but it’s a mess of adapters, including more MOLEX to SATA adapters than I really want to think about to power a couple of 8x2.5 SAS SSD bays.

Getting a separate controller for each front fan did cost more than I wanted to spend, but it gives me individual control of each, it was easy, and it just worked.

*So, this is actually super annoying. I have 8 SATA data ports on this board and no SATA power off the PSU. It’s not a huge problem since I’m almost entirely using SAS drives with two backplanes, but I don’t even have the option to use anything else, which is annoying. This board was released in, I think, 2020. I don’t know why they gave us so many SATA ports and only 3 PCIe 4.0 slots (x16, x8, and x1–I’m not even convinced useful PCIe 4.0x1 cards exist). I’d have gladly given up some of those SATA ports for at least an x4 slot.

1 Like

That’s why I suggested a powered splitter. What’s done is done.

Yeah. If I had it to do over again, I’d do it differently.

At least I have extras now, and this isn’t the worst inefficient budget moment I’ve had while learning how to build this thing.

How do the thermals look after the 30 minute stress test? I think they’re fine, but am I missing something?

According to Noctua’s site it’s rated up to 3A max current. Mine’s arriving in a few days, I’ll hook it up to 2* 12V 0.8A 80mm fans using the SATA power adapter for power and see if anything blows up.

2 Likes

Good news, fan controller arrived and I hooked everything up. So far nothing blew up yet, but I’ve also turned both fans way down, so I don’t think both fans combined are even drawing 1.6A of current.

3 Likes

This topic was automatically closed 273 days after the last reply. New replies are no longer allowed.