Faulty sensor readings on MSI 990FXA-GD80

doveman

Well-Known Member
Hi

Running HwInfo64 v3.91-1485 on MSI 990FXA-GD80 motherboard under Windows 7 x64 Ultimate.

All the information under Fintek F71889A/879A appears to be correct but under Fintek F75387 it shows:
Motherboard 145c
CPU 145c
Fan1 366 RPM
Fan2 366 RPM

The only fan I have connected is the CPU fan, which correctly reports 1061 RPM under the other sensor and the Motherboard and CPU temps are obviously wrong. There's one other temp listed (Aux) which may be correct at 37c (the Motherboard temp under the other sensor is 35c), and four voltages VIN0-VIN3 but I don't really have any way of knowing what they should be.

Also, whilst the Sensor status page shows the CPU speed at a steady 800Mhz most of the time whilst it's idle, on the Summary page the bars are constantly switching between 800Mhz and 3200Mhz. Perhaps this page is just more sensitive and it is actually doing this? It also shows the NB VID at 1.1v even after I've increased it in the BIOS to 1.3v, but it does correctly show that it's now running at 2800Mhz.

I can't seem to upload an HTM larger than 100k and it won't allow me to upload a TXT file at all, so please advise how you want me to get the report file to you (if you need it).
 
Thanks for the report, the Debug File is sufficient in this case.
It seems that all values under the Fintek F75387 sensor are invalid, I'm not even sure if that sensor is really present/working correctly. I think you can ignore them, you can also click "Hide" on this sensor.
Modern CPUs can very quickly change the CPU speed from idle to full speed even for a little while, so it's normal that you see fluctuations. Depends on particular CPU and configuration how much load is required to raise the clock.
 
Martin said:
Thanks for the report, the Debug File is sufficient in this case.
It seems that all values under the Fintek F75387 sensor are invalid, I'm not even sure if that sensor is really present/working correctly. I think you can ignore them, you can also click "Hide" on this sensor.
Modern CPUs can very quickly change the CPU speed from idle to full speed even for a little while, so it's normal that you see fluctuations. Depends on particular CPU and configuration how much load is required to raise the clock.

Thanks Martin.
 
Hi again,

I'm running 3.91-1490 now and sometimes it lists a phantom sensor, Aux, under F71889A/879A between CPU and Motherboard, showing Min/Max temps in the 100s and a Current temp of 0c.

I haven't managed to grab a debug log when this happens yet, but I'll keep an eye out. Here's a "normal" one anyway, in case it helps.
 
This is most probably a fluctuation on a not connected input and should be ignored.
 
Another issue I've noticed is that when starting it seems to generate random "Max" temps, so I have:

Code:
Sensor   Current   Max
CPU 0      33c     40.9c
CPU        24c     31c
Mboard     32c     35c
T1/D1+     24c     32c

Watching it when it loads, the "Current" temps are at the "Max" levels for a few seconds when it starts, before dropping down. I don't believe the temps are really being pushed to the "Max" readings by HwInfo starting, so this seems to a bug and makes the "Max" readings less useful.
 
This is just the local maximum, the values can be pushed to much higher values when putting high load on the system.
Sometimes the start-up/detection phase of HWiNFO can put some load (depending on system configuration), so that explains the slightly higher values.
 
Martin said:
This is just the local maximum, the values can be pushed to much higher values when putting high load on the system.
Sometimes the start-up/detection phase of HWiNFO can put some load (depending on system configuration), so that explains the slightly higher values.

I'm not sure what you mean by "local maximum".

I cross-checked with Speedfan whilst HwInfo was loading and it does push up the temps whilst it's doing so. It might be handy to have a "reset" button to clear the max temps after it's loaded (or have this happen automatically) so that users are better able to monitor "real" increases after it's loaded.
 
Then it's clearly caused by the sensor detection loop which might put more load on some systems.
The button to reset min/max values is a good idea, I'll implement it.
 
Martin said:
Then it's clearly caused by the sensor detection loop which might put more load on some systems.
The button to reset min/max values is a good idea, I'll implement it.

That'll be great. Thanks.
 
Thanks for the reset button, it's really handy :)

Just to let you know, whilst the two fan sensors under Fintek F73587 might be phantom (Speedfan shows them as 366RPM with nothing connected, the same as HwInfo) there seems to be an additional two sensors on the Fintek F71889A which do seem to work and with nothing connected to them, Speedfan, MSI Control Centre and the BIOS all show them at 0RPM, so if you could possibly get them to display in HWiNFO that would be great.

I was also wondering if you might be able to add VRAM usage for my graphics card (HD6950) as then I wouldn't need to run GPU-Z as well just for monitoring that.
 
Fans which report 0 RPM are considered as not connected and not displayed in HWiNFO. If they report any real value, they will be displayed. I have no intentions to change this, because on several systems it's not possible to distinguish a not connected fan from a fan input that's not routed out at all.
I do plan to add Video Memory usage of ATI/AMD GPUs, however I haven't yet found a reliable way how to determine this. GPU-Z and some other tools currently do this via certain Direct3D queries, but I don't think this is a good method.
 
Martin said:
Fans which report 0 RPM are considered as not connected and not displayed in HWiNFO. If they report any real value, they will be displayed. I have no intentions to change this, because on several systems it's not possible to distinguish a not connected fan from a fan input that's not routed out at all.
I do plan to add Video Memory usage of ATI/AMD GPUs, however I haven't yet found a reliable way how to determine this. GPU-Z and some other tools currently do this via certain Direct3D queries, but I don't think this is a good method.

Sorry, I didn't realise HWiNFO auto-hides the not connected/0 RPM fans. That seems like a good idea and I don't expect it to be changed :)

Good to know you're looking into the VRAM usage monitoring. I'll keep an eye on future releases to see if you find a more acceptable method than using Direct3D queries.
 
Back
Top