View Issue Details
| ID | Project | Category | View Status | Date Submitted | Last Update |
|---|---|---|---|---|---|
| 0000996 | SpeedFan | Hardware support | public | 2007-10-10 12:12 | 2007-10-11 19:30 |
| Reporter | norman02us | Assigned To | alfredo | ||
| Priority | normal | Severity | minor | Reproducibility | always |
| Status | resolved | Resolution | fixed | ||
| Platform | Desktop PC | OS | XP 32bit | OS Version | 2600 |
| Summary | 0000996: CoreTemp change with 4.34 beta 23 | ||||
| Description | Hello alfredo, i know the core temp for Core CPUs is a difficult subject, but the change in beta 23 must be wrong. If it was correct, there would be a temp difference between CPU Temp and Core Temps of 20 | ||||
| Tags | No tags attached. | ||||
| Motherboard Model | |||||
| Video Card Model | |||||
|
|
If you send me a SEND REPORT from SpeedFan's INFO tab I will look into this. I don't like to add things to SpeedFan that are not on a datasheet, but this CORE temperature thing was causing so much talk that I chose to include it. I had to include a lot of CPU identification stuff, but now I can set Tjmax based on the die revision. The change I applied is in line with what almost every user out there has decided to accept as correct. |
|
|
I know that for early Core 2 Duos there was a registered that is supposed to indicate whether the Tjmax is 85 or 100 |
|
|
The problem with the PECI value is that it work with an inverse logic. I can support it only if you agree to see a negative value. In fact Intel states that the PECI reading is a negative value. |
|
|
I know PECI is negative, but it's the only safe way to show coretemps. All temperatures deducted from an assumed Tjmax might be wrong. If Speedfan would use PECI value, no specific adjustments to new models would be necessary and the value would always be correct. I don't know if other users have the same opinion about it. The PECI value might confuse many users but on the other hand there wouldn't be any more questions like: "Is my Core Temp to high?" or "Does .... show the correct Core Temp?" With the PECI value there is one simply explanation. It always shows you how far you are from the maximum operating temperature. That's actually very easy to understand. @edit In Intel's document it says (p.92): "Fan speed control solutions based on PECI uses a TCONTROL value stored in the processor IA32_TEMPERATURE_TARGET MSR. The TCONTROL MSR uses the same offset temperature format as PECI though it contains no sign bit. Thermal management devices should infer the TCONTROL value as negative. Thermal management algorithms should use the relative temperature value delivered over PECI in conjunction with the TCONTROL MSR value to control or optimize fan speeds. Figure 28 shows a conceptual fan control diagram using PECI temperatures. The relative temperature value reported over PECI represents the delta below the onset of thermal control circuit (TCC) activation as indicated by PROCHOT# assertions. As the temperature approaches TCC activation, the PECI value approaches zero. TCC activates at a PECI count of zero." The picture below shows a difference of 10 |
|
|
I think that if SpeedFan showed a negative CORE temperature I would get many more support requests. But a negative value is the only one that would not break all the other assumptions about how temperatures work (higher=worse). |
|
|
Why is that? -10 is higher than -20 :) I know what you mean, but if there is no way to correctly identify the Tjmax than it might be better to use the correct delta to Tjmax. |
|
|
Yes, -10 is higher than -20. But the point was that, IMHO, if SpeedFan publishes a negative temperature, a lot of users would start asking what's that. |
|
|
In the documentation at tomshardware it says that Tjmax depends on the stepping. The question is where did they get this information from? Do you know any Intel document that verifies this statement? I know that the PECI value might be confusing to many users, but what's the alternative? We just GUESS the temperature? 'Cause as long as there is no document that says: Stepping ... has a Tjmax of ... one cannot be sure to have correct core temp values. @edit Intel: "Note Tj is not a fixed value and the IA32_TEMPERATURE_TARGET[15:8] value can vary from part to part. Tj is also not software readable" |
|
|
My position is not very different. Look at all the people asking why SpeedFan doesn't always show the correct voltages (I've answered in the online FAQ). I was simply getting too many reports from people telling me that SpeedFan was showing the wrong temperature. It took less to me to implement what users wanted than answer all of those emails :-) |
|
|
But when Tjmax isn't a fixed value, as assumed by Everest, CoreTemp and Speedfan, then the PECI value is the only reliable value. I know this is a very difficult situation and all Intel's fault :), but I don't see any use in showing a possibly wrong temp in contrast to a confusing but correct value. What do you think? @edit Maybe just leave it as it is now :). The people who know the problem also know how to evaluate the temps shown and the ones that don't know the problem won't be confused. If everybody believes strongly that these temps are correct, then they are :) @edit2 pp. 520-531 are very interesting: http://developer.intel.com/design/processor/manuals/253668.pdf |
|
|
I fully agree with what you say. I don't think that the correct name would be "PECI". I think it should be "DTS". PECI is the one-wire interface used to read the value returned by the DTS. |
|
|
But doesn't the software read the value from the PECI Interface? I think this issue can be marked as "resolved". I don't know anything else to say. Subject is clear, and unless Intel improves documentation there is nothing that could be done to truly resolve the issue. |
|
|
No, the software reads it from the CPU specific MSR. The PECI interface allows external devices to read it. I just created SpeedFan 4.34 beta 24 which includes an ADVANCED option to select "DTS interpretation" between ABSOLUTE and RELATIVE. Everybody can now choose the preferred reading type. |
|
|
Great, I think that's a very goog way. @edit Just saw the new beta in my email account. I'll try it right away :) |
|
|
Did you try the beta I sent to you? |
|
|
New Beta works perfectly :) |
|
|
SpeedFan 4.34 beta 25 will add support for ABSOLUTE / RELATIVE temperature reading for Intel Core. The option will be in CONFIGURE / ADVANCED. |
| Date Modified | Username | Field | Change |
|---|---|---|---|
| 2007-10-10 12:12 | norman02us | New Issue | |
| 2007-10-10 12:12 | norman02us | Status | new => assigned |
| 2007-10-10 12:12 | norman02us | Assigned To | => alfredo |
| 2007-10-10 12:53 | alfredo | Note Added: 0003042 | |
| 2007-10-10 12:53 | alfredo | Status | assigned => acknowledged |
| 2007-10-10 13:05 | norman02us | Note Added: 0003043 | |
| 2007-10-10 13:06 | norman02us | Note Edited: 0003043 | |
| 2007-10-10 13:13 | norman02us | Note Edited: 0003043 | |
| 2007-10-10 13:14 | alfredo | Note Added: 0003044 | |
| 2007-10-10 13:23 | norman02us | Note Added: 0003045 | |
| 2007-10-10 13:24 | norman02us | Note Edited: 0003045 | |
| 2007-10-10 13:27 | norman02us | Note Edited: 0003045 | |
| 2007-10-10 13:28 | norman02us | Note Edited: 0003045 | |
| 2007-10-10 13:38 | alfredo | Note Added: 0003046 | |
| 2007-10-10 13:48 | norman02us | Note Added: 0003047 | |
| 2007-10-10 13:56 | alfredo | Note Added: 0003048 | |
| 2007-10-10 14:09 | norman02us | Note Added: 0003049 | |
| 2007-10-10 15:09 | norman02us | Note Edited: 0003049 | |
| 2007-10-10 15:11 | alfredo | Note Added: 0003050 | |
| 2007-10-10 15:15 | norman02us | Note Added: 0003051 | |
| 2007-10-10 15:18 | norman02us | Note Edited: 0003051 | |
| 2007-10-10 15:18 | norman02us | Note Edited: 0003051 | |
| 2007-10-10 15:30 | norman02us | Note Edited: 0003051 | |
| 2007-10-10 15:56 | alfredo | Note Added: 0003052 | |
| 2007-10-10 16:27 | norman02us | Note Added: 0003053 | |
| 2007-10-10 16:35 | alfredo | Note Added: 0003054 | |
| 2007-10-10 16:52 | norman02us | Note Added: 0003055 | |
| 2007-10-10 16:53 | norman02us | Note Edited: 0003055 | |
| 2007-10-10 16:54 | alfredo | Note Added: 0003056 | |
| 2007-10-10 17:03 | norman02us | Note Added: 0003057 | |
| 2007-10-10 17:15 | norman02us | Note Edited: 0003057 | |
| 2007-10-10 17:20 | norman02us | Note Edited: 0003057 | |
| 2007-10-11 19:30 | alfredo | Note Added: 0003066 | |
| 2007-10-11 19:30 | alfredo | Status | acknowledged => resolved |
| 2007-10-11 19:30 | alfredo | Resolution | open => fixed |
