Fan in front of HDDs or not?

Hi!
I need 3 relatively cheap PCs with the following requirements:
  • 2 x 6 TB HDDs with ZFS mirror on each PC
  • Operating in rooms without air conditioning (West Europe: ~ 30 °C in summer)
  • Periodic scrubbing and eventually resilvering
Do you think I need a fan in front of HDDs or not?

I've always been told that for mechanical drives too much cooling shortens life span due to worse lubrication: a single fan extracting hot air seems to be enough to keep temperature in the optimal range. But I'm worried it won't be enough in this situation.
 
You can barbeque the whole PC, BUT there's one exception, and it's the most important one: HDD. Keep it way under 40C. I mean, 40C is max (very max!)
Even with multiple fans, it's unlikly to hit below 20C (the whole temperature inside the PC case). Therefore, if you can use fan, use them.
 
The summer temperatures where live are similar, and I recently managed to overheat two disks (after replacing the motherboard and not checking the BIOS settings on the fan controls).

Every day, SMART tools remind me that I have two disks with warranty void!

I have no doubt that, given your summer ambient temperature, the disks will occasionally overheat if not actively cooled.

In that situation I would get two PWM chassis fans. Make sure that the PSU fan blows inwards, and that the chassis fans do the same.

I expect that most modern motherboards would provide two 4-wire (PWM) chassis fan headers. But you should check on that.

I never skimp on fans. Noctua make quality products.
 
If you are interested in reliability, the best answer SEEMS to be keeping them around 30-40, with few and slow fluctuations. Fluctuations from ambient temperatures are hard to avoid unless you have active cooling (with water and stuff like that). But what you should avoid is heat spikes; so getting lots of ambient air to flow around them is good.

In my case (which is small, the size of a shoebox) I use two fans; one in the power supply, and the other on the side. One blows air in, the other removes it from the case; those two working together give me relatively high airflow speed with reasonable noise. And they're redundant: If one fails, I still have some airflow. I think the way I have it set up is that ambient air comes in, goes across motherboard (which has very little power consumption, it's a 1 GHz 32-bit Atom) and the disks, then through the power supply, and out from there. That's because the power supply is the most easily replaced component if it fries, so it gets the warmest air.

Scrubbing and resilvering is good. How often you have to scrub is a question of considerable scientific debate. I do once or twice a week, and that's probably excessive (and perhaps on modern disks even counterproductive).
 
  • Thanks
Reactions: a6h
In my case (which is small, the size of a shoebox) I use two fans; one in the power supply, and the other on the side. One blows air in, the other removes it from the case; those two working together give me relatively high airflow speed with reasonable noise. And they're redundant: If one fails, I still have some airflow.
The traditional issue with some fans pulling and some pushing is that you get a vortex effect, with a direct rush of air through the path of least resistance, and the development of "quiet backwaters" where it can get very hot. I accept that with a small box, the probability of backwaters might be reduced, or even eliminated.
However, the ATX design is for all fans to blow inwards, for positive air pressure to penetrate the entire case, and for the air to exit through a myriad of small holes.
 
There is zero chance a fan could do this unless you live above the Artic Circle.
That's correct.
A few years ago, I was testing mining Monero on a laptop! I put the poor laptop in a typical home refrigerator (0C~3C) and started to mine. I was monitoring different sensors, before and after mining. I can't remember the exact number, but I'm pretty sure it's never reached below 20C. maybe around 25C-30C.

[EDIT/Footnote]:
I looked up my notes (I keep alot of notes for some unknow reason!) Unfortunately I didn't record the sensor stats. But it turned out it was on an Ubuntu machine. Because of CUDA/GT-640M, I couldn't run it on FreeBSD.
 
How cold would it have to be to affect the viscocity of lube?
There is zero chance a fan could do this unless you live above the Artic Circle.
Long time ago, in a giant data center far away (it was actually less than 20 years, and the data center is close enough, I can get there by car from my house in an hour if there is no traffic jam) ...

There was a giant supercomputer, owned by one of those somewhat secretish federal agencies, with tens of thousands of disk drives (that was considered large back then). It was also before people had figured out how to cool data centers efficiently, so really cold air at about 10 degrees C (50F) was everywhere. The vendor turns the machine on, and disk IO is ridiculously slow. Laughably, insanely slow, about a factor of 2 worse than expected. Strange, all the prototypes worked fine in the lab. After much checking, the following was discovered: At low temperatures, the disk drives themselves go into a safety mode, where after every write command, they reread the data just written to check the write succeeded. That means that even the smallest writes take a full disk rotation (and in those days disks were spinning slower), while large writes take two full rotations instead of one.

Why? Because the disk drives themselves have a temperature sensor, and they "knew" that at low temperatures, there was too much thermal inaccuracy in platter size, spindle bearing flutter, and rotation rate.
 
Thank you for your answers. When asking about missing fans in front of disks two different stores told me they are rarely used since:
  1. a correctly ventilated case (as an example one fan on top, one on back to extract air) ensures the optimal HDD temperature;
  2. disks running constantly some degrees below optimal temperature have a higher failure rate.
Searching hardware forums (yes I know, not a scientific method..) seemed to confirm this opinion.

Of course after your answers I'll look deeper into this topic.
 
I put the poor laptop in a typical home refrigerator (0C~3C) and started to mine.
Bad idea though, it can cause condensation (there's typically quite a bit of moisture in the air). Water (droplets) and computer components do not mix well.

a correctly ventilated case (as an example one fan on top, one on back to extract air) ensures the optimal HDD temperature;
Sure, but it assumes there's typically only one to two drives in the case. When you stack a bunch of them, one on top of the other, the cumulative heat can build up. So make sure there's some space between the drives for the air to flow through.
 
I really don't see how case fans can cool HDDs TOO MUCH

However, personally, I use kitchen thermometers (yes, for cooking the meat) to check the temperatures of the internal components.

Industrial infrared thermometers obviously exist but they are more expensive (not the cheap-amazon ones)
 
Sure, but it assumes there's typically only one to two drives in the case. When you stack a bunch of them, one on top of the other, the cumulative heat can build up. So make sure there's some space between the drives for the air to flow through.
Thank you. So I guess it would be a very stupid idea to buy a cheap PC and replace SSD and CD-ROM with two spinning disks!

But if I order a serious case with enough space between disks, do you think it's better to ask for a fan in front of disks or not? For the average home use I would stick with the store suggestion to NON put the front fan since it would probably make HHD temperature go below the optimal range.

As an example I found this study from Google engineers:
According to graphic at page 6 disks operating below optimal temperature are more likely to fail than disks operating slightly over.


But, the idea of a ZFS scrub or resilvering in the middle of the summer scares me.
Maybe an additional fan in front of disks which can be activated during the hotter months?
 
I would stick with the store suggestion to NON put the front fan since it would probably make HHD temperature go below the optimal range.

I really does not think this could happen.
I've never seen an over-cooled magnetic drive with air fans.
And I've really seen a lot of them

It is a question of noise, not of temperature.
If you can handle the noise, put on as many fans as you want.
 
But if I order a serious case with enough space between disks, do you think it's better to ask for a fan in front of disks or not?
It's not going to hurt to add some extra airflow over the drives. Unless you live above the arctic cycle I wouldn't worry too much about getting things too cold.

I usually build systems using external drive trays (so I don't have to take the computer apart to replace a drive), those typically include a fan on the case. Those drive bays usually put 4 3.5 inch disks in a 3 bay 5.25 slot. There are bays with 5 disks in a 3 slot 5.25 bay. If you go that route make sure to get a drive bay that allows you to easily replace that fan. That fan is going to wear out at some point in time, it's going to start to whine or even stop spinning. So you will want to replace it.
 
@fans: I highly suggest Noctua one.
Last for years and years (MTTF ~150.000 hours), 6 years warranty.
 
Hi!
I need 3 relatively cheap PCs with the following requirements:
  • 2 x 6 TB HDDs with ZFS mirror on each PC
  • Operating in rooms without air conditioning (West Europe: ~ 30 °C in summer)
  • Periodic scrubbing and eventually resilvering
Do you think I need a fan in front of HDDs or not?

I've always been told that for mechanical drives too much cooling shortens life span due to worse lubrication: a single fan extracting hot air seems to be enough to keep temperature in the optimal range. But I'm worried it won't be enough in this situation.
What is more problematic for disks (as for any mechanical device) than absolute temperature is the "temperature walk" (the speed of change). Most manufacturers limit this to something like 10 Kelvin per hour.

Then, I personally do not like the idea of chilling the disks. I think they love to run somewhere between 30-45 Celsius. (All which I know are certified up to 55 Celsius, at least.)

Then, what is also important is the mounting of the fan. I have a little one attached to some remote and badly ventilated disks, that auto powers on when SMART reads above 45 Celsius. What it does (besides reducing the temperature) is kick the "raw read error rate" of the neighbouring disks from 0 to significant amounts. Yes, cheap 40mm 1$ fan, but indeed taking care of the vibrations is relevant.
 
  • Thanks
Reactions: a6h
Back
Top