Erroneous temp readings ! ?

[B1G]

I wasn't sure where to post this. It's more disk related then unRaid related, but since this is the only computer forum I like I'll ask here

 

I spin down all disks last night before going to bed. This morning I log on to the server, spin up all disk and took a temp reading of the drives ( I want to log temp reading in different circumstances. {Spinning with fans on, spinning fans off etc...}) Well the initial readings on most disk was 23c (WD GREEN), which sound perfect since the room was at 22c. The parity disk (WD BLACK) was at 25c. Still sound ok for a BLACK. But then I have two old disks. The 250g WD (No idea of it's color) was at 58c. And the 120g MAXTOR was at 10c.

 

I opened up the case and put my finger on the 250g and it wasn't hot !

 

Can s.m.a.r.t reading be off like that ? Is it somthing I should worry about ?

 

Thanks for your input.

[SSD]

This is very unusual. Likely nothing to worry about. You might check the temps very soon after powering up after computer has been off overnight. If you are getting similar results, it confirms the issues is not a boiling or freezing disk.

[B1G]

Yeah. I'll add that to my log tests. I was just curious to know if this was something others had experienced. 

[rich.bingham]

I would suspect that neither the measurement method nor the sensor is very standardized across brands or even between generations within a brand.  Your WD Cavair drives are reporting close together which makes sense, same manufacturer, roughly the same age so likely using the same design and implementation of temperature sensor.  I'm guessing the 250GB WD is an older design so having a value so out of whack could be explained by using a different temp sensor design or measurement method (i.e. could be sensing temperature of one of the onboard chips instead of temperature of air within the drive).  The Maxtor drive is probably a whole different can of worms.

 

That said, while absolute measurement may not be accurate, relative measurement usually is.  If you take readings under load you might see that the ambient to load temperature delta is comparable between drives.

 

edit:  Of course this is complete speculation.  I'm not familiar with the spec (if any) that defines how this tempearture is measured and/or reported.

[B1G]

I would suspect that neither the measurement method nor the sensor is very standardized across brands or even between generations within a brand.  Your WD Cavair drives are reporting close together which makes sense, same manufacturer, roughly the same age so likely using the same design and implementation of temperature sensor.  I'm guessing the 250GB WD is an older design so having a value so out of whack could be explained by using a different temp sensor design or measurement method (i.e. could be sensing temperature of one of the onboard chips instead of temperature of air within the drive).  The Maxtor drive is probably a whole different can of worms.

 

That said, while absolute measurement may not be accurate, relative measurement usually is.  If you take readings under load you might see that the ambient to load temperature delta is comparable between drives.

 

edit:  Of course this is complete speculation.  I'm not familiar with the spec (if any) that defines how this tempearture is measured and/or reported.

 

I get what you mean. But the variance is ALOT. From one drive saying 23C to another saying 58c! And all of that at startup! (within 1 minute of startup)

 

Anyways... MORE POWER TO standardization

 

I've always hated having one salesman tell me "This amp is 100W RMS" and then having another salesman tell me "Oh but no, that's not a real 100W RMS"

[SSD]

edit:  Of course this is complete speculation.  I'm not familiar with the spec (if any) that defines how this tempearture is measured and/or reported.

 

Speculation and not true. 

 

While it is true the different manufactures may embed their temperature sensors in slightly differnt places, the reporting of temperature data from the drives has been very consistent, at least going back to the 200G IDE drives that I first used with unRAID.  There were some reports that Samsung was purposely putting their sensors away from the hottest action and resulted in it reporting slightly cooler temperatures than WD and Seagate were reported here, and their was a heated debate on the topic, but we were only talking about a few degress C.  Not 20 or so like here.

 

This is a malfunction IMO.  Just weird you'd have 2 malfunctions related to the temperature sensors on the same machine.  Unless your controller is mishandling the data coming back from the drive, however, I think it is just a coincidence.

[rich.bingham]

A thought on the 250GB WD.  The "good" drives report 23°C and 25°C while the 250GB drive reports 58°C, the offset between these readings is ~32°C (32=2^5), makes me wonder if there's a bad bit in the register that is storing temp value.  To test this you could kick the room temp up 5-10°F and read the temperatures.  If for instance the "good" WD's read 26-28°C afterwards does the 250GB WD read in the low 60°C range?  Not necessary, but could be interesting to find out. 

 

As bjp999 said, it's a malfunction but nothing to worry about especially since you confirmed that the drive is cool to the touch and not actually 58°C. 

 

With regards to the SMART spec, I was able to find a more recent document (2005) which indicates that "the placement, accuracy, and granularity of temperature sensors to support Table 19 are vendor specific."  Table 19 spells out the meaning of the various bytes of the HDA Absolute Temperature SMART reading.  There may be a more recent revision to this part of the spec, but it would explain slight variations in powerup drive temps between vendors (not to the extent reported here though).

 

Speculation and not true. 

Yep, you're correct there.  I found some volume data in Google's paper "Failure Trends in a Large Disk Drive Population" which indicates the temperature readings are generally consistent.  Page 6 shows a fairly normal distribution of average drive temperatures recorded using SMART none of which seem to fall below 15 and few above 50.  The data are a product of drives, drive utilization and environment, but if there were big differences we'd expect a multi-modal distribution (assuming they were looking at drives from multiple vendors).