FX8120 and his temp

[EarthDog]

ok.. lets not wake sleeping dogs manu.. All I want is to get to the bottom of this once and for all, like the other thread about the amount of water.

The AMD guy may be right, but Im just not sure why Coretemp would seemingly be rendered useless and to use the other applications instead.

 

[manu2b]

I am sure you will get the same kind of output...

 

[Bubba-Hotepp]

Thats the problem with applications like Speedfan and Hwmonitor as opposed to Coretemp. In Coretemp there is one temp which is the only one we need to care about frankly. We have all been fine using Coretemp for years with great success. I have seen the emails, and done my best to decipher them, but am coming up empty. Honestly, you are the only person I have seen, across three forums, ever try to bunk the readings from Coretemp... as I said, its been the bible for years.

I will look at the emails again and contact the maker of Coretemp to see if he can help get to the bottom of this.

The problem with Core Temp is that it was designed for Intel CPU's which is why it lists only the core temps (which don't apply to AMD) and "TJunction" which is what Intel uses for it's "Max Operating Temp" while for AMD it's "TCase".

Edit - For AMD CPU's TJunction is the thermistor in the socket which of course is not going to be accurate which is why they use TCase. HWMonitor is a good program but it's only going to list what it receives from the bios/uefi which is why on a gigabyte MB you'll see TMPIN instead of "CPU Temp" like you would on an Asrock (and other MB's use different labels).

ok.. lets not wake sleeping dogs manu.. All I want is to get to the bottom of this once and for all, like the other thread about the amount of water.

The AMD guy may be right, but Im just not sure why Coretemp would seemingly be rendered useless and to use the other applications instead.

It's not that "CoreTemp" is rendered useless as so much as it's just designed to report "TCore" and "TJunction". It's directly relevant for an Intel CPU which uses Tjunction in a different way than AMD.

 

[EarthDog]

Bubba, I PM'd the creator of Coretemp in hopes he will clarify the concerns.

 

[Bubba-Hotepp]

Bubba, I PM'd the creator of Coretemp in hopes he will clarify the concerns.

Good deal. I'm trying to dig up the "white paper" on AMD that lists max operating temps for the CPU series of Phenom (it goes all the way to FX). I just have to remember which paper it was.

http://support.amd.com/us/Processor_TechDocs/43375.pdf

Page 17 Table 7.

 

[The Coolest]

After K8, AMD changed the way their internal temperature reporting mechanism worked.
With Phenom Core Temp (and HWMonitor) users were introduced to very strange readings.
In AMD's technical documentation these readings are described as non-absolute values, that are used to thermally control cooling solutions, and monitor the temperature to prevent overheating.
Reading the reply that Bubba-Hotepp got from AMD it makes sense.
Basically neither of the values we get are actually 100% (or close enough) reliable. As we are still using chips on the motherboard to read and interpret the values given off the TCase sensors, and we're still not getting accurate values from the 'TCore' readings as well.

It's really hard for me to say which readings should be trusted, it's obvious that the readings given by Core Temp at default settings are too low, as they very often dip below ambient temperature. With many samples, it became apparent that in most cases the temperature difference between TCase and TCore readings is between 10 and 15C.
My personal opinion on this issue is simply to use whatever you prefer. For TCore readings it's simply suggested to set a +15C offset.

 

[EarthDog]

Thank you for the explanation.

I would personally stick with Coretemp +15C as to not have to sit and wonder which is what on Hwmonitor and Speedfan. To each their own... and thanks bubba, and the coolest for posting up great information.

 

[Bubba-Hotepp]

After K8, AMD changed the way their internal temperature reporting mechanism worked.
With Phenom Core Temp (and HWMonitor) users were introduced to very strange readings.
In AMD's technical documentation these readings are described as non-absolute values, that are used to thermally control cooling solutions, and monitor the temperature to prevent overheating.
Reading the reply that Bubba-Hotepp got from AMD it makes sense.
Basically neither of the values we get are actually 100% (or close enough) reliable. As we are still using chips on the motherboard to read and interpret the values given off the TCase sensors, and we're still not getting accurate values from the 'TCore' readings as well.

It's really hard for me to say which readings should be trusted, it's obvious that the readings given by Core Temp at default settings are too low, as they very often dip below ambient temperature. With many samples, it became apparent that in most cases the temperature difference between TCase and TCore readings is between 10 and 15C.
My personal opinion on this issue is simply to use whatever you prefer. For TCore readings it's simply suggested to set a +15C offset.

I have to give kudos to the AMD Tech Rep since he went above and beyond to actually get the embedded guys to actually give up what little information they did.

 

[The Coolest]

I agree.
I've tried contacting them in the early days of Phenom, but my attempts to get any useful information were futile.

 

[Bubba-Hotepp]

I agree.
I've tried contacting them in the early days of Phenom, but my attempts to get any useful information were futile.

I think they're a little paranoid because of all that's gone on between AMD and Intel.

 

[WhitehawkEQ]

Here is 3 screen shots of my PII 1090T, FX-8120 and FX-8150 systems running F@H,

Phenom II 1090T at 3.2GHZ


FX-8120 at 3.4GHZ


FX-8150 at 3.655GHZ



all using the V6 Cooler by CoolerMaster

 

[Bubba-Hotepp]

Thank you for the explanation.

I would personally stick with Coretemp +15C as to not have to sit and wonder which is what on Hwmonitor and Speedfan. To each their own... and thanks bubba, and the coolest for posting up great information.

Honestly that will give a close enough idea of your temps. I ran HWMonitor, Speedfan, and Core Temp side by side and got this -



Uploaded with ImageShack.us

As you can see Asrock makes it easy by their UEFI labeling it "CPUTIN". Core temps were 14 to 16C lower than CPU Temp so it falls right in line with what Coolest said. My only concern is that as information passes from one to another and then onward the message gets garbled and people don't learn about the +15C factor and then you have someone running their system with cores at 57C and their TCase Max is 62C for their chip and they think they're within that limit when actually their TCase is around 72C and they're actually well past it. Other than that using Core Temp should be fine with the +15C factor kept in mind as Coolest said. (Why oh why do these companies have to make things soooo freakin difficult )

Edit - BTW ambient was between 25.5 and 26C in the computer room today. Was even worse yesterday

 

[trents]

After K8, AMD changed the way their internal temperature reporting mechanism worked.
With Phenom Core Temp (and HWMonitor) users were introduced to very strange readings.
In AMD's technical documentation these readings are described as non-absolute values, that are used to thermally control cooling solutions, and monitor the temperature to prevent overheating.
Reading the reply that Bubba-Hotepp got from AMD it makes sense.
Basically neither of the values we get are actually 100% (or close enough) reliable. As we are still using chips on the motherboard to read and interpret the values given off the TCase sensors, and we're still not getting accurate values from the 'TCore' readings as well.

It's really hard for me to say which readings should be trusted, it's obvious that the readings given by Core Temp at default settings are too low, as they very often dip below ambient temperature. With many samples, it became apparent that in most cases the temperature difference between TCase and TCore readings is between 10 and 15C.
My personal opinion on this issue is simply to use whatever you prefer. For TCore readings it's simply suggested to set a +15C offset.

Well, I'm not sure if this relates but I've found that the delta between CPU socket temps and core temps with good air cooling is commonly about 10c under load, with the socket temp being the higher. I get uneasy about the accuracy of the temp sensor's calibration if it's much more than that, which it often is with the Thubans and the FX CPUs.

 

[Bubba-Hotepp]

Well, I'm not sure if this relates but I've found that the delta between CPU socket temps and core temps with good air cooling is commonly about 10c under load, with the socket temp being the higher. I get uneasy about the accuracy of the temp sensor's calibration if it's much more than that, which it often is with the Thubans and the FX CPUs.

Remember that the socket thermistor on an AMD motherboard outputs the "TJunction" reading. When you see "CPU Temp" in bios that's actually TCase. TCore are the individual core temps which are derived by a formula from the TCase thermistors on the die package. TJunction (as it was explained to me by the Tech Rep) is basically a fallback reading in case the bios/uefi doesn't receive a temp reading from the CPU for whatever reason.

 

[trents]

. . . TJunction (as it was explained to me by the Tech Rep) is basically a fallback reading in case the bios/uefi doesn't receive a temp reading from the CPU for whatever reason.

"From the CPU" meaning TCore or TCase? I ask because remember part of the problem we're dealing with here is the vague terminology we commonly use and you seem to fall back into that in this statement. Sometimes in HWMonitor we see CPUTIN or it may be missing and we see only TMPINx lines when certain CPU model lines (for instance, the Athlon IIs) are being employed. I've always assumed that was because the program wasn't able to specifically identify the sensor, i.e. that the socket sensor wasn't in its database. On the FX CPUs, however, we always see a CPU temp line.

 

[Bubba-Hotepp]

"From the CPU" meaning TCore or TCase? I ask because remember part of the problem we're dealing with here is the vague terminology we commonly use and you seem to fall back into that in this statement. Sometimes in HWMonitor we see CPUTIN or it may be missing and we see only TMPINx lines when certain CPU model lines (for instance, the Athlon IIs) are being employed. I've always assumed that was because the program wasn't able to specifically identify the sensor, i.e. that the socket sensor wasn't in its database. On the FX CPUs, however, we always see a CPU temp line.

The way it works is that the CPU sends the temp signals (for AMD the Tcase, and Tcore which comes from a formula using the Tcase readings to approximate the cores) to the BIOS/UEFI. The BIOS/UEFI is what then adds the labels that you see in programs like HWMonitor which is why for the same CPU, you will see labels like "TMPIN" on a Gigabyte board vs a labels like "CPUTIN, AUXTIN" etc. on an Asrock board. That's what causes the confusion because the MB companies and companies like Award etc, haven't gotten together to "standardize" that part of it. A program like HWMonitor doesn't interpret the various labels, it just lists them as is. I get the feeling that coders like cooler don't really want to go through the hassle of verifying with each MB manufacturer which labels are what and then code that all into their program. Easiest way to check which is the CPU Temp in HWMonitor is to either look at the readings while in the "health" section of BIOS or run the monitoring program that some MB's come with (or can be downloaded from the manufacturers site) and compare those with the readings found in HWMonitor to see which is which. With an FX CPU in a Gigabyte MB you will still see the "TMPIN" readings as you still see the various labels that other MB's output. I've only seen Asrock use "CPU" in the label but there might be others I'm not sure.

 

[trents]

Good explanation. Apparently Asus uses "CPU" label as well.

One of the conclusions I have arrived at lately is that I think it wise to look not only at core temps but the CPU/TMPINx lines in HWMonitor as well since it seems there is so much inconsistency in all this. I say that because it seems folks are getting instability well below the 55c core temp max we so often preach around here as the stability wall. I suspect that the core temps are often being under reported in some of the newer AMD product lines. I'm not sure if the motherboard/bios is partly responsible for this or what but it is my observation.

 

[Bubba-Hotepp]

Good explanation. Apparently Asus uses "CPU" label as well.

One of the conclusions I have arrived at lately is that I think it wise to look not only at core temps but the CPU/TMPINx lines in HWMonitor as well since it seems there is so much inconsistency in all this. I say that because it seems folks are getting instability well below the 55c core temp max we so often preach around here as the stability wall. I suspect that the core temps are often being under reported in some of the newer AMD product lines. I'm not sure if the motherboard/bios is partly responsible for this or what but it is my observation.

And that's why the explanation provided by AMD makes so much sense. In all of their documentation they list "Tcase max" as the maximum temperature (see the link for the thermal guide I posted from AMD). Why would they list Tcase max if Tcore were so important? Wouldn't it be "Tcore max"? It was those two questions that got me to email AMD T.S. in the first place. Now you look at how they explained core temperatures and that it's not an actual reading but is basically a "guesstimate" using the Tcase sensors and a formula and it's easy to see why people report varying levels of core temps when they hit instability. Not to mention the danger posed by thinking your cores should be able to hit the max temp (for example 62C) and the frustration people have when they become unstable in the 50s wondering what they're doing wrong or if they got a "bad chip". I don't think it's that the "cores" are being underreported at all. Think about it this way. In essence Tcase is the sum (averaged out) of all the heat buildup inside of the CPU which is comprised of more than just the cores. The most important part of the thermal guidelines is the removal of that heat using cooling methods. It makes sense that you would want to know the total temperature of the CPU. What good is a reading of 50C on the core if the onboard NB is at 70C and vice versa. Not to mention that as he said putting a thermistor inside the core itself is extremely difficult and by removing it frees up real estate on the core itself. Die space = money. So they put the thermistors on the package (from how he described it I'm assuming just under the die on the wafer) and use all of them averaged out for a "CPU Temp" and of course to give the "approximate" "core temp". The motherboard's BIOS is only taking the readings from the CPU, adding the vendor specific "labels" and using those readings for the BIOS, OS, & various software.

 

[caddi daddi]

cloudheaven, please lower your "cpu voltage" to 1.25 volts.
lets mount a fan in every fan port you have on your case.
lets make sure your cpu cooler fan exausts in the direction of a case exaust fan.
when you have fans on all the fan ports les close up the case.
please try to make sure your airflow is front to rear.

earthdog, thanks for the review of the coolermaster hafx case, your write up and pictures proved to be much more help than the little coolermaster supplied with the case or on the colermaster site.
I just moved the pig from an antec 200 case to the haf x 942, great case to build, a loser for water cooling.
how can they call this high air flow? i added a second 200mm fan on top and it still just moves a little air.

 

[Bubba-Hotepp]

Speedfan?!?

EDIT: I apologize. I just learned it has been designed by a NASA thermodynamics engineer!

Hmmmm...seems to be really quiet now.

cloudheaven, please lower your "cpu voltage" to 1.25 volts.
lets mount a fan in every fan port you have on your case.
lets make sure your cpu cooler fan exausts in the direction of a case exaust fan.
when you have fans on all the fan ports les close up the case.
please try to make sure your airflow is front to rear.

earthdog, thanks for the review of the coolermaster hafx case, your write up and pictures proved to be much more help than the little coolermaster supplied with the case or on the colermaster site.
I just moved the pig from an antec 200 case to the haf x 942, great case to build, a loser for water cooling.
how can they call this high air flow? i added a second 200mm fan on top and it still just moves a little air.

That's weird....sorry to hear that. I have the HAF 922 and I have a ton of airflow through it.

Edit - I'm still waiting for cloudmaster to post all of his settings so we can look and see what's going on.