Captain Face-Plant Posted October 13, 2010 Share Posted October 13, 2010 OK. I don't know exactly when this happened, but this evening I was keeping an eye on my temps (even though aida64 has been running the whole time) when on the OSD panel, cpu 1 / core #1 started reporting a completely different temp reading from the sensor page within aida64. At first, I thought closing the program and launching again would fix the issue, but it did not. So, I noted how much the OSD Core #1 reading was. About 4-6C cooler than the sensor page. All other cores on the OSD reported the same correct values as the sensor page. I ran a system stability test to see if that might unstick the reading in the OSD, but it continued to report the variance, 4-6 cooler than the sensor page for core #1. So I "watched" it some more, knowing that the actual intake temperature is progressively getting lower. After "watching" the OSD for a bit longer, it seems the OSD really wants core #1 to be 20c when idle, sometimes dropping to 19C or popping up to 21C. Meanwhile, core #1 in the sensor page is bouncing around between 19c and 27c, while the OSD just sits around 20c. I have my sensors set to update every one second in all aspects that can be set within aida64. I have provided a video clip for further assistance. video is 1680x1050 xvid and 1:11 in duration. Sorry for the external link, but the forum only allows for 500K for a file. As with visiting any file hosting site, please have 3rd party ad blocking software running when visiting. Please use adblocker before clicking Quote Link to comment Share on other sites More sharing options...
Fiery Posted October 13, 2010 Share Posted October 13, 2010 1) What processor is it? 2) Do you have C1E or SpeedStep enabled? 3) Do you have Turbo Boost enabled (if your CPU is Turbo-capable)? Thanks, Fiery Quote Link to comment Share on other sites More sharing options...
Captain Face-Plant Posted October 14, 2010 Author Share Posted October 14, 2010 CPU is Intel Core i7 920 C1E and Speedstep disabled TurboBoost on (MB allows for +1x multiplier [boost] in BIOS. Default is 20, is set to 21) I did solve the situation. It seems the aida64.ini seems to be corrupted. I moved the file to my desktop to let aida64 rebuild a new one. New .ini generated has OSD and Sensor Page with all cores reporting the same value for the corresponding cores. (Secretly blames patch tuesday) You may lock this topic and is solved. It seems the solution did not last. Core #1 on the OSD is back to reporting value approx. 5c cooler than SensorPage once again. Quote Link to comment Share on other sites More sharing options...
Fiery Posted October 14, 2010 Share Posted October 14, 2010 Thanks. Maybe a long shot, but can you please try the following: 1) Disable Turbo Boost (you can do it in AIDA64 / CPU Tweakings, I'm sure you know where to find that ), and check if it makes any difference. 2) If #1 doesn't make a difference, then please try to run AIDA64, and manually adjust the CPU affinity (using Task Manager), and so lock AIDA64 to the 2nd logical processor, then the 3rd logical processor. Please let us know if it makes any difference. Thanks, Fiery Quote Link to comment Share on other sites More sharing options...
Captain Face-Plant Posted October 16, 2010 Author Share Posted October 16, 2010 O.K. I spent some time on this, this evening. Disabling TurboBoost did not eliminate the bug. So I left turboboost disabled and went to step two. I started with locking each core separately, 0-7. Here were the results. With processors 4-7 going solo locked, the results remained the same as before with core # showing the temp variance. With processor 0 or 4 locked, the results were the same again. However, with processor 1 locked, the temp variance was still there, however core # 2 readings wanted to play the buggy values game, but not as much as core #1. With processor 2 locked, core #1 still showed variance, but core #2 was showing the variance more than when processor 1 was locked. Quote Link to comment Share on other sites More sharing options...
Fiery Posted October 16, 2010 Share Posted October 16, 2010 Thank you for performing that experiment. Quite frankly, I think it could be normal, and it may come from the fact that AIDA64 uses 2 different threads to measure sensor values (temperatures, voltages, fan speeds). One thread is used exclusively by the Sensor page, and the other one is shared by all the Hardware Monitoring features (Sensor Icons, OSD, Sidebar Gadget, Logitech LCD, External Applications, Logging, Alerting). The two threads usually measure the same when they're running side-by-side (e.g. when you have OSD and also Sensor page opened); however, the Intel DTS core diodes are very sensitive to CPU load, so some variance on the thread that is used by AIDA64 to measure sensor values can occur. When I write "side-by-side", of course I don't mean the 2 sensor threads would measure values at the same time. The two threads are executed one after each other, otherwise their low-level code would collide with each other. When e.g. the thread of the Sensor page is executed first, and it puts a load (utilization) on the CPU core it runs at, the second thread going after that (the Hardware Monitoring thread) may just start from a hotter state of the CPU core, and so it could measure a hotter CPU core temperature than expected. Then, until the next update the CPU core cools down a bit, hence the next thread (Sensor) starts from a cooler state again, hots the core up again, and the next thread (Hardware Monitoring) starts from a hotter state again, etc. etc. Such anomaly could be confusing, I admit, but I believe it's not a program bug, but more like an architectural short-coming of AIDA64 that we could only eliminate by joining the two separate sensor threads, to make sure once a value is measured, the same value is displayed everywhere in AIDA64. We'll work on it. Thanks, Fiery Quote Link to comment Share on other sites More sharing options...
Captain Face-Plant Posted October 19, 2010 Author Share Posted October 19, 2010 Out of curiosity today, i launched the old product, everest. It seems that it now behaves the same way now. Since the old product has the variance now as well, my only conclusion is that M$ patch tuesday has caused this issue. I do not know what could have done that, but since the values became eratic after patch tuesday, I have to assume that is the cause. Both products worked fine that day before updates were installed and both products now appear broke for core #1. Quote Link to comment Share on other sites More sharing options...
Fiery Posted October 19, 2010 Share Posted October 19, 2010 Can you please check if HWMonitor is also affected by this anomaly? Quote Link to comment Share on other sites More sharing options...
Captain Face-Plant Posted October 20, 2010 Author Share Posted October 20, 2010 O.K. Compared HWMonitor vs Sensor Page vs OSD. The OSD and the HWMonitor matched consistently, with a standard variance of +/- 1C. Sensor Page Did not match HWMonitor all the time. It showed the same 4-6 C variance against OSD/HWMonitor when they were not in sync. I decided to poke around some more with aida64 and decided to exam the Statistics page of the System Stability Test versus HWMonitor. I believe the Statistics page in SST is suffering the same issue as Sensor Page in the main part of aida64. As you can see Core#1 (Core #0 in HWM) maximum value have a variance of 4C. Also, forgot one piece of information in my 1st reply: The stepping for the i7 920 CPU is D0. Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.