Hard drive temperature vs. power consumption

[pjc]

This is probably a silly question, but why wouldn't a drive with a lower power consumption run cooler than one with higher power consumption regardless of RPM?

It seems like the thermal output would be proportional to the power input. But I'm seeing 7200 RPM drives that idle at 40C vs. 5400 RPM drives that idle at 30C, even though their power consumption is identical (5W idle, 7.5W r/w).

Why would that be?

 

[Bidule0hm]

Because the power consumption stated in the specs is a maximum, not an average. The average will be higher on the 7200 RPM drive than on the 5400 one.

 

[BigDave]

But I'm seeing 7200 RPM drives that idle at 40C vs. 5400 RPM drives that idle at 30C, even though their power consumption is identical (5W idle, 7.5W r/w).

I would think the higher RPM motor would generate more heat, regardless of power consumption.

 

[Chris Moore]

It has been my observation that some brands of drive run hotter than others even at the same 'rated' RPM. For example, I had a batch of servers we received where some were filled with Western Digital Re drives and others were filled with HGST drives and others were filled with Seagate Constellation drives. The HGST drives were hottest, followed by the Western Digital and the Seagate drives were coolest of all even though they were all rated at 7200 RPM and the chassis were the same brand and model. It took some time, but I got all the HGST and WD drives changed out for Seagate drives because they ran (on average) 10 degrees cooler.

 

[pjc]

@Bidule0hm, that's an interesting idea, but the idle specifications are for the average, not peak. Likewise the r/w specifications are for average (assuming ~40 io/s) instead of peak (peak can be up to 25W!)

@BigDave, that would seem intuitive, but the first law of thermodynamics says otherwise. Essentially all of the electricity consumed by a drive is converted to waste heat (it has to go somewhere). So if you consume 5W electricity, you produce 5W heat. Electricity -> motion -> friction -> heat is just a longer path to the same result as heat loss in the circuitry.

@Chris Moore, you may have a partial answer. The Seagates are running much cooler than HGST. But that would still require the HGST to consume more power (to convert to heat) than the Segates. But as mentioned above, mine are rated as having the same power consumption.

 

[pjc]

Ew...I think I may have figured it out.

According to SMART, all the drives that are running cooler have APM enabled...and obscenely high load counts to prove it...

So they're probably pulling less than their active idle power.

But the FreeNAS UI says APM is disabled on all of them(?!) And changing it in the UI doesn't seem to have any effect. Changing it via smartctl --set apm,off /dev/daX does seem to have worked.

(Now to deal with the drive that I've just now realized is at 2x its rated load cycle count...)

 

[pjc]

Confirmed: power consumption determines temperature (along with environment). RPM is only relevant to the extent it increases the power consumption.

Back in the day, 7200 RPM drives reliably used more power than 5400 RPM drives. Now there are 7200 RPM drives that require far less power than 5400 RPM drives from a few years ago. And, if you believe the spec sheet, WD's Red 10TB drives use the same amount of power for both 7200 (Red Pro) and 5400 (Red), so their thermal characteristics should be the same.


...But the spec sheets can be a bit funny. The one-pagers are not necessarily accurate, I think in part because they don't get regularly updated. If you look up the user manual for the specific drive (and revision), you may see very different numbers.

So in my case, the spec sheet for the Seagate 5400 RPM drives advertised the same power consumption as the HGST 7200 RPM drives' spec sheet. But when I found the full manuals / OEM specifications for both (documenting all the characteristics, ATA commands, how they measured average power, etc.), it turns out the more recent models of the Seagate use considerably less power: 4W idle and 5.6W r/w (average).

I also measured the power consumption on an individual drive (error probably within 1W) and found the Seagate used 4W idle and 7W r/w and the HGST used 6W idle and 8W r/w.

So 20%-50% more power at idle for the HGST, which would explain the higher temperature.


I'm curious about the 10TB Red/Red Pro, though. I wasn't able to find a detailed spec document for them. Just the one-pagers that say they use 2.8W idle and 5.7W r/w, which are the lowest absolute values I've seen for any NAS-rated drive 4TB and up, and the best W/TB rating as well. But they're not cheap.

 

[Chris Moore]

So in my case, the spec sheet for the Seagate 5400 RPM drives advertised the same power consumption as the HGST 7200 RPM drives' spec sheet. But when I found the full manuals / OEM specifications for both (documenting all the characteristics, ATA commands, how they measured average power, etc.), it turns out the more recent models of the Seagate use considerably less power: 4W idle and 5.6W r/w (average).

The manufacturers do themselves a real disservice when they make it so difficult to find that information.