Losing Radeon 5870 VRM sensors?

ssateneth

Well-Known Member
Using 1660 on win7 x64. As of late, my Radeon 5870's temperature sensor for the VRM seems to randomly not display anymore. Rebooting often fixes it. The sensor isn't listed in the Sensor Display Config. In fact, all the sensors for the VRM seem to be missing for me atm.. I lost the voltage and amperage sensor too.

Is this a possible fault with the software, or the hardware itself? The thought crossed my mind that another monitoring program is blocking access to the 5870's VRM, but I have no changed any video driver related software, or used any new monitoring software.

I attempted to enable debug mode to get a DBG for you, but hwinfo64 freezes on "Analyzing PCI Bus"
 
Is this a problem of the latest build only (1660) so the previous ones showed the values always ?
Could you please at least attach screenshots of the sensors with the information visible and missing ?
Also, when the values are not displayed and you open sensors again do they appear back ?
 
Martin said:
Is this a problem of the latest build only (1660) so the previous ones showed the values always ?
Could you please at least attach screenshots of the sensors with the information visible and missing ?
Also, when the values are not displayed and you open sensors again do they appear back ?

will update as soon as I can afford getting the info you ask.

Closing and reopening sensors causes the problem to be intermittent..
I opened hwinfo normally (not just sensors) and it showed "Volterra VT1165 @ GPU[#0] ATI/AMD Radeon HD 5870 (CYPRESS XT)" but only GPU VRM Voltage and GPU VRM (doesnt say amperage, but its the amperage). I closed the sensors without closing hwinfo and re-opened sensors and lost the Voltera area entirely. I did this about 8 times to see if it would come back and it didnt. I closed hwinfo entirely and reopened it normally, opened sensors, and all 3 sensors were there. Closing and re-opening sensors couldnt reproduce the bug. I closed and re-opened hwinfo 3 times and repeated this process, couldnt reproduce the bug. Closed it again, and ran it again, only this time I ticked "sensors only" and the entire volterra went missing. After a few more opens and closes, it seems that I lose my Volterra sensors if I do "Sensors only", and they show up if I open hwinfo normally and click sensors after it opens up.

I've encountered this bug rarely in the past, but usually dismissed it because I thought it had something to do with overclocks rendering talking to the card unreliable.

This is with a normal startup and sensors are showing like they should.
allsensors.png


This is with a "Sensors only" startup.
allsensorsmissing.png
 
Thanks for the additional information. I'll need the Debug File from both situations - start with Volterra shown and without.
Are you sure that when you run in Debug Mode it freezes? It might take quite long to start in this mode, especially when you're running it from an SSD for example. If this is the case, then please try to run it from a hard disk or try to wait a bit longer.
 
Martin said:
Thanks for the additional information. I'll need the Debug File from both situations - start with Volterra shown and without.
Are you sure that when you run in Debug Mode it freezes? It might take quite long to start in this mode, especially when you're running it from an SSD for example. If this is the case, then please try to run it from a hard disk or try to wait a bit longer.

I am indeed running the OS (and hwinfo) from an SSD. I'll let it "Analyze" and post back with Debugs when they finish and if they exhibit the bugs.

Here is a DBG file that lost all 3 VRM sensors while clicking "Sensors Only": https://dl.dropbox.com/u/9768004/HWiNFO64dbg.zip
Another DBG with a "Normal" startup, VRM sensors still missing: https://dl.dropbox.com/u/9768004/HWiNFO64dbgfull.zip

I will get DBG files when it successfully detects them during a debug run
edit: I'm having a very hard time having VRM sensors successfully detect under debug mode. I tried doing normal mode with a normal startup (not sensors only) and enabling debug mode after its already open, then opened Sensors, and it appeared to try to detect "special" sensors it wouldnt normally try to detect under normal mode, and it found VRM sensors, but it didn't produce a DBG file. :/
 
Martin said:
Thanks. Please also attach a DBG file when you see the VRM sensors.

So I upgraded to 1665 and my first debug run had the VRM sensors showing. I've since added 2 radeon 5770's to the same PC too, but thats about the only difference. Not sure if the DBG you wanted is relevant anymore
 

Attachments

Thanks for the file.
Is it possible that at the time when HWiNFO didn't report that sensor, the monitor to which this GPU is attached was turned off ?
Anyway, I'll implement a patch for this in the next build.
 
Martin said:
Thanks for the file.
Is it possible that at the time when HWiNFO didn't report that sensor, the monitor to which this GPU is attached was turned off ?
Anyway, I'll implement a patch for this in the next build.

Negative. My monitor is always connected to my 5870. The 2 5770's I added are doing cryptographic calculations, nothing is connected to them.
 
Bumping this to note a similar behavior using 1680. Volterra VT1165 (The voltage regulator of reference radeon 5830/50/70 video cards) temperature sensor seems to be MIA until I open up a program called Sapphire Trixx, a general video card over/under clocking and over/undervolting tool. When Trixx is opening up (before it pops up on screen), the sensors list scrolls up to the top by itself, then the volterra temperature sensor is magically there. The volt and amps sensors are usually already there.

How do you want me to go about capturing this debug file for you to analyze?
 
I implemented some additional patches in v4.00 and I hoped it will fix it. So is the v4.00 behaving the same as before, or do you see any differences ?
 
1680 is the same as v4.00. When I start it up it says HWiNFO64 v4.00-1680. The behavior as before is different as the behavior now. Before it wouldn't pick up the sensors at all, or it would be partially detected. Now it is always partially detected unless Sapphire Trixx was opened (doesn't seem to need to be running at the time of opening, just opened at least once since booting), then it will detect the temperature sensor as well. I can replicate this on 3 of my computers with different motherboards/cpus's, but they all have at least 1 reference 5870 with the Volterra sensors.
 
So now only the temperatures are missing until TRiXX is started ?
And before build 1660 it always reported correctly (as you stated in the 1st post) ?
Please attach a new Debug File and Report File (or sensors screenshot) from the latest version when the problem occurs.
 
Martin said:
And please try this build if it fixes the problem and let me know:
www.hwinfo.com/beta/hw64_401_1684.zip

Sorry, I had to go away for a while. back now, and have the opportunity to test it now (got some 5970's, and noted the same behavior was being exhibited). your link to 1684 is broken, but I see you have a new beta out. I'll try that first, and if it still exhibits, I'll upload debug + report file before sapphire trixx, and debug + reports after sapphire trixx.

Also, do you want me to make a new thread as far as calibrating various sensors on the EVGA nForce 790i SLI Ultra motherboard, or do you want to do it here? Voltages awry here and there (14v for 12v line and other very wrong things).

Edit: All volterra sensors appear to be detected and functioning properly without the need to open Sapphire Trixx first.
edit2: Eh, I'll go ahead and make a new thread since the original problem was solved.
 
Thanks for the feedback. Glad to hear that it works now.
Yes, I removed the intermediate Beta release because I released a new public beta build.
Going to check the other thread...
 
Back
Top