How are my Tcase/Tunction temps for my 125w AMD Phenom II x4 965?

[Dankk]

Keep in mind that my 965 is the 125-watt revision, and not the 140-watt one. Even still, I realize that this is still a fairly large power consumption compared to most other processors, and I wanted someone to confirm if my temperatures are comparatively good or not good.

I've heard different things about how Tcase should be X many degrees lower than Tjunction, but I'm not sure. Everest reports general CPU temperature (Tcase) as being ~15 degrees lower than the individual cores (Tjunction).

I guess what I really want to know is: Which of these temperatures should I be paying attention to? I'm normally used to looking at the individual core temps, which seem kinda high (44 degrees idle) and under Prime95 they go upwards of 63 degrees.

I just barely got Everest, which is has finally introduced to me the other "general" CPU temperature readout that is much lower.

So again, I just wanted to ask: Do you think my processor is running at an ideal temperature? Should I be worried about the Tjunction temps which are considerably high? Thanks.

Edit: I ran Prime95 again, and although the core temps still top out at 63 degrees, the general/Tcase temps stays absolutely the same (33 degrees). Does this mean something is not calibrated correctly?

 

[Jovec]

Edit: I ran Prime95 again, and although the core temps still top out at 63 degrees, the general/Tcase temps stays absolutely the same (33 degrees). Does this mean something is not calibrated correctly?

I would say so.

AMD CPUs do not have per core temp sensors AFAIK and are merely approximations (and possibly not good ones at that). If the cores are showing 63, the the CPU is likely running hotter than that, and the CPU temp (not the core temp) is supposed be under 62 for Phenom IIs.

Judging from the screenshot, I would say that the motherboard temp is actually the CPU temp. Do some testing between idle and load to see how much it fluctuates and how it relates to CPU core temps.

 

[Dankk]

If the cores are showing 63, the the CPU is likely running hotter than that, and the CPU temp (not the core temp) is supposed be under 62 for Phenom IIs.

Err... that's not good. I can't imagine why it would be running that hot though. The processor and HSF appeared to fit together just fine.

Judging from the screenshot, I would say that the motherboard temp is actually the CPU temp. Do some testing between idle and load to see how much it fluctuates and how it relates to CPU core temps.

Are you saying that the general "Motherboard" and "CPU" temperatures are swapped with each other in Everest? I wouldn't be surprised if it is. That screenshot I posted above displays 52 for the motherboard, but right now it usually sits around 58 or 60.

Indeed, I've realized that on AMD's website, it says that the max temp for my processor is 62 degrees. Is this what Tcase/CPU temperature actually is? Is this the temperature where the processor can actually become unstable, or is it the temperature where it simply takes a performance hit (but not yet unstable)? I'm not sure what this "max temperature" means.

A couple more things:

1) I just barely took out the stock HSF, wiped off the thermal grease and reseated it with Arctic Silver 5. I saw the core temps drop about 4 degrees (so now idling at about 40) which is nice.

2) After that, I turned on the AMD "Cool 'N' Quiet" feature in the BIOS, and the temps dropped even more: down to 35 degrees which is impressive. I'm wondering if I should just keep this feature on because I don't do any overclocking. Hopefully it won't affect any of my gaming or anything.

 

[Jovec]

Are you saying that the general "Motherboard" and "CPU" temperatures are swapped with each other in Everest? I wouldn't be surprised if it is. That screenshot I posted above displays 52 for the motherboard, but right now it usually sits around 58 or 60.

It might be. Or it might not be. Put your CPU under full load with P95 and watch what temps go up to try to deduce if the "motherboard" label is really your CPU temp.

Might want to try AMD Overdrive too. However, use the "Board Status" under "Status Monitor" and not the "CPU Status" (which will show the individual core temps and the stock voltage (regardless of actual voltage)). TMPIN2 might be your CPU temperature - it was with my Gigabyte mobos. My Asus board shows an actual CPU temperature label in AOD.

Regardless, all of my AMD Athlon II/Phenom II CPUs have a CPU temp that is always higher than the individual core temps.

Indeed, I've realized that on AMD's website, it says that the max temp for my processor is 62 degrees. Is this what Tcase/CPU temperature actually is? Is this the temperature where the processor can actually become unstable, or is it the temperature where it simply takes a performance hit (but not yet unstable)? I'm not sure what this "max temperature" means.

I believe that the 62 degree max listed in the specs is the CPU temperature max, not the core max. I doubt it's a hard ceiling cap though, but I wouldn't push it much past it for too long.

A couple more things:

1) I just barely took out the stock HSF, wiped off the thermal grease and reseated it with Arctic Silver 5. I saw the core temps drop about 4 degrees (so now idling at about 40) which is nice.

2) After that, I turned on the AMD "Cool 'N' Quiet" feature in the BIOS, and the temps dropped even more: down to 35 degrees which is impressive. I'm wondering if I should just keep this feature on because I don't do any overclocking. Hopefully it won't affect any of my gaming or anything.

CnQ should drop the CPU voltage down from 1.4v to 1.0v while dropping the clocks from 3.4GHz down to 800MHz. The reduced voltage will noticeably lower your temps. I doubt you'll notice any performance issues with CnQ enabled, but there are a few reports where it might (choppy video playback with it on, smooth with it off, etc). You can always turn it back off. FWIW, I've never had issues gaming with CnQ on.

Since you are not overclocking, try under-volting too. Of course, replacing the stock cooler is also an option.

 

[Dankk]

It might be. Or it might not be. Put your CPU under full load with P95 and watch what temps go up to try to deduce if the "motherboard" label is really your CPU temp.

I ran Prime95 again, while keeping my eye on the supposed "motherboard" temperature in Everest. It barely fluctuated +/- 1 from ~55 degrees. The only way this would be possible with a CPU is if it had a ridiculous idle temperature at 55, and yet the cooling is good enough to keep it exactly the same under full load. I doubt this is the case. As I said, it only went up 1 degree at times, but other times it went down 1 degree.

I'm beginning to wonder what this "Motherboard/TMPIN1" temperature really is. Unless you're willing to convince me that this really is my processor somehow.

Might want to try AMD Overdrive too. However, use the "Board Status" under "Status Monitor" and not the "CPU Status" (which will show the individual core temps and the stock voltage (regardless of actual voltage)). TMPIN2 might be your CPU temperature - it was with my Gigabyte mobos. My Asus board shows an actual CPU temperature label in AOD.

Same results. I've looked across multiple hardware monitoring tools, specifically Everest, AMD Overdrive, and SIW. They all report the same thing. The CPU/TMPIN0 reading still hovers around 33, and the individual core temperatures are in their usual 35-60 range (idle-load).

Another thing I should probably point out is the CPU temperature reading it gives in the BIOS. When loading the BIOS, there is a single CPU temp that reads at around 40, the same temperature that the multiple core temperatures display in other utilities. I'm thinking this is the Tjunc reading, but rather than listing all 4 cores, it just gives an average or something. I don't know if this helps at all, but that's what it shows.

I'm using a Biostar motherboard, specifically the BIOSTAR A785G3. It's an AM3 MicroATX mother board with an AMD 785G chipset.

Regardless, all of my AMD Athlon II/Phenom II CPUs have a CPU temp that is always higher than the individual core temps.

I was searching around, and I found a discussion on the AMD forums about this. There was a guy who said exactly the opposite: He says he looks at the regular CPU temp when running Everest stress tests, which is usually a bit lower than the individual core temps. Again, this is just one guy though.

...And, of course my CPU/TMPIN0 doesn't actually fluctuate the way it should, so that doesn't help.

I'm beginning to wonder if the Tcase sensor on my board is just fubar.

 

[Dankk]

Update: I think I may have discovered which temperature is the true CPU temperature.

Sorry for the double post. I realize we've been throwing around different names/labels for the different temperatures (TMPIN0, TMPIN1, TMPIN2, etc) and it tends to get confusing because different hardware-monitoring utilities display them a bit differently.

For example, SIW displays TMPIN0, TMPIN1, TMPIN2, then the individual cores, the GPU, hard drive, assembly, and so-on.

Everest, however, appears to change things a bit (and even leave out one or two temperatures). It has Motherboard, CPU, individual cores, "Aux", and the hard drive.

Of course, in Everest, the "CPU" readout is the one you would usually look to in order to find the regular CPU temperature. Jovec is right. They are actually mixed up. But it's not the motherboard temperature that it's swapped with. Rather, it's the odd "Aux" readout in Everest that I just mentioned.

In SIW, however, it's labeled as TMPIN0 (I must have been looking at the wrong one in my previous post, because it's not the same temperature. My bad.)

Here is a diagram of both programs to explain it clearly:

The TMPIN0/Aux temperature, assuming it's actually the CPU temperature, falls just a little bit below the core readouts, which would make perfect sense with this statement made at the AMD forums:

I wouldn't use the individual core temps to monitor your CPU temps. As as an example, with Everest, my core temps are at 41c, while my CPU temp reads 37c for my 940. When I run Everest stress tests, I only monitor the CPU reading, not the individual core temps.

However, this disagrees with what you said, Jovec. The other guy says that the regular CPU temp is normally a little below the individual core temps, which is what I'd rather believe, although I'm not sure what kind of setups you're using.

 

[deimos3428]

You have three temp sensors (not counting the individual core temperatures), and Everest guessed wrong about which was which. I believe you have it right now, but let's look at it logically.

1. There's no way during a Prime95 run with the stock cooler that your CPU temp (Tcase) is 33C. So 53C and 61C are plausible options.

2. When you stop Prime95 which falls fastest, "TMPIN0" or "TMPIN1"? The faster-dropping temp is in all likelihood the sensor under the IHS, ie. "TCase", and the one you need to watch. I'm betting on 61C given your core temps and the fact it's the highest.

3. The 53C is probably your NB temp, and it'll tend to increase/decrease slower and not get quite as hot (unless you've got superior CPU cooling).

4. The 33C is likely an onboard sensor for the ambient temperature inside your computer's case.

So it looks like you're at the point you should stop overclocking, or get a better CPU cooling solution.

 

[Dankk]

When you stop Prime95 which falls fastest, "TMPIN0" or "TMPIN1"? The faster-dropping temp is in all likelihood the sensor under the IHS, ie. "TCase", and the one you need to watch. I'm betting on 61C given your core temps and the fact it's the highest.

Yes, I think I found it. The one that shows "61" core temps in the picture above is the one that drops the quickest as well.

So it looks like you're at the point you should stop overclocking, or get a better CPU cooling solution.

Oops, sorry. I neglected to mention that those temps in that last screencap were taken while under 100% load. So, ~61 under load, and about ~33 degrees idle. (If it were truly idling at 61, then yes, something is definitely wrong there and I need a better cooler.)

Thanks.

 

[Jovec]

However, this disagrees with what you said, Jovec. The other guy says that the regular CPU temp is normally a little below the individual core temps, which is what I'd rather believe, although I'm not sure what kind of setups you're using.

It's entirely possible my understandings have been wrong.

The "Core" temperature is read by an on-die sensor in the CPU-NB and is calculated by taking this reading and dividing by 8. Which explains why they are always listed as the same.

I do not think there is another temperature supplied by the CPU.

The "CPU" temperature and the one of "TMPINX" temps probably refers the the BIOS temperature sensor, which is located below the socket.

AMD, to my knowledge, has never officially stated what temperature is subject to the 62 degree max on the Phenom 2. If it is the core reading, then I've been overly safe and have some temperature headroom in my OC.

I also doubt that low 20s idle on my 1090t cores is accurate, especially given the higher idle voltage on the Thubans.

 

[deimos3428]

I neglected to mention that those temps in that last screencap were taken while under 100% load. So, ~61 under load, and about ~33 degrees idle. (If it were truly idling at 61, then yes, something is definitely wrong there and I need a better cooler.)

Yeah, I got that. You could keep using the stock cooler at your current settings without any issue whatsoever, as it's unlikely you'll fully load the CPU in normal use.

The point was that if you are at 61C under full load, it wouldn't be wise to overclock any higher without better cooling. You've hit a thermal wall.

Some other useful discussion/info on the subject of AMD CPU temps:
http://www.overclock.net/amd-cpus/663718-maximum-safe-phenom-ii-955-965-a.html

 

[Dankk]

I know this thread is somewhat old now, but I'd like to point out that since 2-3 weeks ago, my processor temps have dropped by about another 5 degrees. I think it may be a combination of the Arctic Silver 5 setting in, and me rewiring my machine for better airflow.

Main CPU temperature never goes above 56 degrees on full load. (Idles are still in the low 30s.)

While I was initially quite surprised, I saw that both the Tcase AND Tjunction sensors are reporting a few degrees lower, which can't be a coincidence. I'm pretty happy right now.

Edit: Nevermind, I tried it again and it still idles at about 60. I think it was particularly cold last night when I tried it...