Various problems with RPi Quad SATA Hat

Noctua NF-A4x10 (https://www.amazon.fr/gp/product/B00NEMGCIA/ref=ppx_yo_dt_b_asin_title_o04_s00?ie=UTF8&psc=1)

2 Likes

my only gripe with the noctua fans is that it is too thick…10mm… it causes the oled pcb to be warped, which isnt good on the long run… once the stock fans die, i’ll prob buy noctua and sand it down to 8mm or 7mm if possible for better clearance…

Agreeing with you HAT PCB is little under pressure, but not too enough. For LCD not cramped and if you unscrew top cover you can release the presure.

Thanks for the info CrunchyDoodle and StephaneP! Fan on order.

You have to make some adjustment for installation:
1/ Be careful with OmniJoin : cable from SATA HAT are more thick than Noctua :slight_smile: Push both cables very far inside OmniJoin and be sure to close OminJoin firmly (I’ve cut original fan cable’s at 50 % very short, from my POV : cut then very close to original fan)
2/ For Anti-vibration Pads, you need to remove them on opposite at SATA HAT board. But keep those are in contact with board.
3/ Original screws from SATA HAT are not compatible (too small).
4/ For Anti-vibration mount you must fix them firmly and cut very very very very near Fan frame.
5/ When mounting you SATA HAT board : do not block on screw : put you cover, adjust for button and after block you board screws.

How exactly do you turn on Power Management on the your hard drives? Did you use hdparm?

Yes but I’ve set at 96 to have medium APM level

I have the Noctua fan in place and running. It seems that it does not change speed. The fan on the CPU does but the Noctua fan seems to be at %100 all the time. It’s a bit loud and I may seek out a resister to slow the RPMs to help the noise. It does seem to drop the temps 9-10 degree F, which is a plus.

i’m seeing the same behavior with the noctua fan, full speed no matter what I set up in the config, even when I do a shutdown of the PI, the fan is still at ful rpm :expressionless:

if you have WD drives, please do not set APM on hdparm… WD drives have the intellipark nonsense that will keep adding to load_cycle_count every 10s and degrade the drive prematurely… setting the APM to 255 on hdparm for WD drives will solve the load_cycle_count issues, but generates more heat… its a double edge sword…

I tried swapping the top fan over to suck fresh air into the chasis with a ā€œO-ringā€ that i hack together from a strip of plastic to ensure that the surround air in the chasis dont get sucked in and recirculated… will report tomorrow on whats the stable temps…

1 Like

Ok, I let the system run overnight… apparently running with the tophat fans flipped to suck fresh air instead of blow hot air actually increased by idle temps by 2-3deg…

But the plus side is that my drives temps are lower by 7-8deg… so it’s a give and take and adjustments to the fan curve to hit the sweet spot to prevent the fans from keep kicking up too high…

It’s not actually true that a WD drive will add to load_cycle_count every 10 seconds. it depends on workload of the device and many other things. the drive will attempt to spin down every 8 seconds, the 10 seconds people talk about i is you have something that keeps waking the drive up every 10 seconds.

Not all WD have it and if they do you can change the timeout through hdparm using the -J flag.

"GET/SET Western Digital ā€œidle3ā€ time out for a WSC ā€œgreen driveā€

Its not a ā€œnonsenseā€ feature it was developed for windows desktops where it makes sense due to the disk workload, Drives with it are not recommended by WD for linux RAID/NAS. It has its right use case where it makes perfect sense for energy savings. Which is end user devices,

if you tried to use hdparm -J on the quad sata hat, it will throw errors back at you, so you cant change the timeout flag no matter what… the only way to kill it is to set hdparm -B 255 which disables APM altogether…

you should probably try it out and also do a smartctl -a and check the load_cycle_count…

before i disabled APM, the load_cycle_count was increasing at a rate of 1 per 10-15s… after disabling APM, i only get a +1 to load_cycle_count when the disks spin up… you can read it here Killing your drive by trying to save milliwatts

1 Like

eskimo is right, on the newer drives you can’t disable the idle timer anymore, i saw the exact same behavior before i set it to 255, i have wd blue drives, i used hd-idle instead, works great

From my POV and my setup (my RPI-NAS will be store in my house tech room :slight_smile: then noise of noctua at 100% is not a problem. And I’m thinking about to remove the case (unfortunaly :frowning:) -10°. But before do that i will try add 2 fans for bottom of case.

Afer disks or RPI at 65° is not a risk is most ot time.

To be frank, i’m not concerned about the temps of the RPi, its designed to operate in heat… i’m more concerned about the HDD… heat is the HDD’s main enemy… the problem of removing the case is likely dust accumulation and accidental damage to components…

the main problem of my solution

You guys can try using the Jmicron updater tool and run the following command to set the HDD idle sleep…

sudo ./JMS561FwUpdate -d /dev/sdX -t Y

where X is the disk, and Y is the idle time before spin down in minutes…

1 Like

As soon as I switched mine on for the first time, I thought ā€œI may replace that top fan with a Noctuaā€. Later I came to these forums and discovered your post, so ordered one, performed the mod and it works a charm. I have other issues but that’s unrelated to this mod. Many thanks!

1 Like

https://www.youtube.com/watch?v=cwWkUQFGdm8 is a really handy video on how to configure a RockPi Penta Hat.

I’m missing all the screws to mount the hat to the standoffs (the top 4, not the actual standoffs).

1 Like