Temperature reading discrepancy

[Y0tsuya]

I noticed that when HD Sentinel GUI is first started, the temperature reading matches that of SpeedFan and 3Ware's 3DM2. But after a while, usually overnight, the reading would "drift" higher. I first noticed the discrepancy yesterday, and re-launched the GUI which fixed the reading. But this morning I had to do it again. HD Sentinel would report 40 degrees C while 3DM2 and Speedfan both would report 29 deg. Re-launching the GUI again brought HD Sentinel back inline with the other two. I think the drift happens over a long period time because the temperature graph only goes back a few hours but I can see a definite upward trend even when the drives are definitely idle at 28-29 deg C. I'll see if I can reproduce this over the next few days and maybe get a screenshot or something.

[hdsentinel]

The difference can be caused by different detection periods. I'm not sure how often other software ask the drives to provide updated information, but you can check (and configure) it in Hard Disk Sentinel at the Configuration -> Advanced Options.
Higher disk usage (even a background activity like defragmentation, scheduled virus scan, depending on the OS and other software) may increase the temperature with 10-15 Celsius or even more quickyl (as you can see on the second image at http://www.hdsentinel.com/screenshots.php )

It is possible that other software ask for information less frequently thus does not follow temperature changes so quickly.

It may be important that different software may list the drive(s) in different order. So if you see (for example) higher temperature for the first drive in Sentinel, it may not be the same as the first drive displayed in the other software.

To diagnose this issue, please use the "Report" menu "Send test report to developer" option when the system started and send a different one some hours later, preferably during higher disk activity and/or if you feel the temperature value may not be correct. This way I can examine the details and compare the differences between the "idle" and "activity" reports to check how the temperature, power on time and other attributes change - to verify if the drives reported all data properly and these are displayed in the software as should.

[Y0tsuya]

I managed to catch HDSentinel reporting higher-than-expected temperature just now. But after I opened 3ware 3DM2 to get its reading HDSentinel shifted the temperature reading back down. The temperature history still works however and you can see the sudden shift back in line with other readings. The readings from the same drive is circled in red. HDD#5 is connected to port 0 on the 3ware controller. Here's a screencap to prove it's not just my imagination. I think it also happened once at 10:28am. I'm pretty sure the drives were spun down most of the early morning so the reading should have stayed down at 28-30 instead of 36. Then the array spun up at 10:28am to do something and HDSentinel also noticed the change and updated the reading, hence the sudden shift from 36 to 29.

I still can't nail down exactly what caused the elevated readings and what triggered the resets. But I'll also try to set things up later to generate a report.

[Y0tsuya]

Alright I submitted 2 reports, one after re-starting HDSentinel. That one looks right. The second was was submitted after pressing "Update" button the reading shot up by 4 degrees. Attached another image showing 5 degree difference in temp readings.

[hdsentinel]

It seems that the drive may provides a "cached" temperature value after it was spinning down. The software constantly reads updated information but did not read/write the surface so the drive could remain in sleep mode and provided the last measured (slight higher) temperature.

When you restart the software or start other software which accesses the drive, the drive spins up and provides the current (actual) temperature. The same happens if you press the "Update" button because it does a complete re-initialisation (like if you were restart Hard Disk Sentinel).

[hdsentinel]

After the "Update" used, the actual temperature should appear. As I see, the temperature level then changed, increased slightly - so I feel the software reads the proper temperature because it followed the change.
What I feel interesting that the other software still reports 30 Celisus, so I feel they did not update the information as should.

Can you please try other software, for example Everest also? Just to see the value(s) detected by a 4th tool.

[Y0tsuya]

Yes I agree it might have something to do with the HDD's powerdown mode. However I've also noticed that, although updating shows increased temperature, it quickly dropped back down afterwards to match the reading from other tools. Doesn't happen everytime so not easy to reproduce. In the meantime the other tools are in agreement and show stable temperature. One of my concerns is that sharp temp drops displayed by HDSentinel shouldn't be possible since the drives are packed tightly into a 3U enclosure and air cooling can only manage a slow cool-down. Even when drives are not tightly packed I haven't seen drive temp drop so fast with simple forced-air cooling. Also I'm not sure the drive platter has to spin back up in order to provide SMART data, especially with the newer "green" drives.

Let's go back to the first picture with the steep temperature drop and subsequent slow rise. After the sharp drop, HDSentinel is in agreement with Speedfan and 3DM2 with the subsequent slow ramp up. But then I believe the drives spin down and temp dropped, which was reflected in Speedfan and 3DM2 readings but HDSentinel stayed high. Speedfan can chart temp over time so I'll try to confirm this later. I'll have to wait until things line up correctly to take a snapshot.

[hdsentinel]

After the disk spun down, the temperature reading can still remain a bit higher (as you can see) because of the cache-effect of the drive, which provides the higher temperature for the software. But the drive itself is stopped and cooling down slightly. The result will be a cooler drive - which is reported in Hard Disk Sentinel when the drive is spinning up again.
So when you see the "drop", it is really the new reading, after the drive totally cooled down. Because of the spin down (and the cache effect) you cannot see the slight decrease.

All software detects temperatures at different times, different intervals. So it is possible that when you see a temperature increase/decrease (in any software), it is not immediately reflected in an other software.

[Y0tsuya]

While I agree the powerdown may be messing up the reading, the fact that two tools agree and one doesn't suggests the other one is wrong. BTW to get SMART readout with 3DM2 on 3ware you have to press a button to get the latest refreshed reading (since it's static HTML). So just before I did the latest screencap the last action I performed was to get 3DM's reading.

The latest screencap shows HDSentinel and SpeedFan both logging increased temperature. But while SpeedFan logged a temperature "hump" after the drive spun down, HDSentinel reading remained constant. As always, 3DM2 agrees with SpeedFan. So my interpretation is that SpeedFan and 3DM2 are not affected by spindown but HDSentinel is.

[hdsentinel]

So my interpretation is that SpeedFan and 3DM2 are not affected by spindown but HDSentinel is.

Yes - this is definitely the case.
For safety, if Hard Disk Sentinel detects that a drive is sleeping, it does not try to detect further information to prevent unwanted wake-up of the drive. This is why you can still see the latest measured values (cached) during inactive periods. When the drive spins up again by any means (including restart of Hard Disk Sentinel, because on startup it must detect complete disk information), it immediately updates the temperature and other information.

This safety feature was introduced on all drives and controllers to maximize power savings if HDD power down features are enabled on the drives because many controllers and drives wake up from sleeping state just to provide health and temperature information.
It seems 3DM2 and SpeedFan knew that it is not required on this 3ware controller because this 3ware controller does not wake up the connected disks just to get the updated SMART information.

(The "Information" page in Hard Disk Sentinel confirms the above, you can see "Power State: Sleeping (since 10/14/2009 7:02:06 PM)" on that page for the particular disks).

Thanks for the details, soon I'll send you some more information in e-mail and will provide a test-version which may work differently in this situation.

[Y0tsuya]

Sounds good. I verified the sleep state of the HDD's and everything matches up as we discussed.

[Emmaking]

Thanks for your post.it is so great







----------------------------------------------------------------------------------------------------------------
http://www.reliabletop.com

[nimbystripes]

I agree that this is very helpful.