Sheesh, could it just be my chip? (Temperature)

[essenbe]

Kbrady is right. Today most chips will throttle down or completely shut down if the max CPU temp is reached. However, I would not go there. Sorry, but I just don't trust such things. If mine gets close to max, I will stop it.

 

[Britton30]

Odd statement about AMD, I have a Phenom, well two of them, and I can see temps for each core, one 4 and one 6 core.

Erick, I have got my Intel CPU to 100C and it had throttled itself at about 75C. That was 1C below TJmax but I can't run it at as high a clock as when it was new. I also have a 16C difference between the cores when stressing it.

 

[ICIT2LOL]

Yes Steve I know what you mean and I did some further reading last might and it would appear that the thermo diode is actually part of the silicon chip structure.

Yes KB that throttling effect is not only for stopping a "melt down" but also if you read that article I posted monitors the point at which the chip stops functioning properly (transistors, resistors, caps - circuitry wise) to put it as simple as that.

This I found when working with dedicated transistors years ago - there comes a point where the temp changes the characteristics of the transistor function for example whether it conducts or not and I had many explode on me so hence I guess why they include those diodes into the chips.

Problem is the info is very hard to find.

 

[essenbe]

from what I have read, the info comes from those diodes in raw format, meaning nothing unless you know the algorithm to use. For some strange reason, neither Intel or AMD will tell the temp monitoring companies what the algorithm is. So, they have to figure it out themselves, which is not easy and also why I don't completely trust any of them. From everything I have read, no matter what temp monitoring program you use, the idle temps are usually way off but the higher the temp goes the closer to being correct it is. I presume the reason for that is even if you know the algorithm, the diode info only tells the distance to Max temp.

The thing about CPUs and temps is they act completely different under cold temps. When overclocking with ln2, Sandy Bridge would not boot unless the FSB (or whatever they call it now) was over 106, Ivy Bridge is supposedly better. Ram acts different too. And the CPU's under normal cooling function better under lower temps and suffer less leakage. No matter what voltage you set in bios, the CPU never sees that much, there is supposedly leakage where a small part of that voltage goes to other components. But the lower the temps, the lower the leakage. Many high end overclockers say that with sandy and Ivy Bridge, the CPU voltage does not matter, only temps matter.

@Gary, that was a quote from the FAQ at core temp.

 

[Britton30]

It's all above my pay grade I am now thinking.
Not trying to be an ass or argumentative, but here's Core Temp fomr one AMd machine. Maybe they meant the 1st gen Phenom.

 

[essenbe]

Gary, I am only reporting what Core Temp said, I didn't say it. If you look at your screenshot, it is showing frequency for each core but only 1 temp.

 

[Britton30]

yup, you're right. Actually I had noticed before it only showed the Core 0 temp. I'm getting too old for thinking.

 

[essenbe]

That's why I just report what I read. I'm too old to think. Besides, it hurts!

 

[ICIT2LOL]

Well it is going a bit over my head too Gary if that is of any consolation -mind you it doesn't have far to go LOL!!

I was just thinking maybe the diode is embedded in the centre of the cores - after all we are talking extremely small structures here eh?

Like anything else with any of the hardware stuff one could go on forever and either get no real answer or consensus as to what is happening I suppose after all is said and done the thing is going to shut down hopefully before it melts eh?

 

[essenbe]

I've always thought (there I go again) that with one diode it would measure Tcase (the temp in the center of the CPU) With all cores showing temps it would be showing TJ Max (T Junction, the temp of each core). Of course, when I think I usually get in trouble.

 

[Layback Bear]

Gary my Intel chips never show more than 3 degc difference between cores, idle or under load. I would suggest new thermal past and making sure you cpu cooler is mounted correctly and tight.
If you still get the spread of temps it could be the temp monitoring program or just the chip you have. It could also be the cpu cooler you are using.
I use Real Temp and HWMonitor.

 

[Britton30]

Jack, I just think I cooked the CPU by going to 100C once. When new the differential was 5C-7C. each one has its quirks.

I have re-re-applied the paste and lapped the base of the cooler, Hyper212 EVO, with 2 fans in push-pull.

 

[kbrady1979]

That one temp it is measuring might be a general CPU temp, or CPU Package .......which will never be the temp of any of the cores. I think I have lost the purpose of this thread lol. Is he thinking he needs a new cooling solution?

 

[Britton30]

He is wondering why his temps are hotter.

 

[kbrady1979]

Thank you........I was confused for a minute.

I don't know if this is applicable, or helpful....but here ya go. Shown here is a snippet of AIDA64 and RealTemp running a stress test for about 30-40 minutes. After roughly 5-10 minutes, my CPU Cores are just about as hot as they will ever get. That is how long it takes to register the shock of 100% load. Notice the temps it is monitoring: Motherboard, CPU, Core 1-4, and CPU Package. CPU and CPU Package will sometimes have a value all their own(even though CPU Package in this snippet looks to be similar to the core temp). I don't really know what the CPU temp is measuring, and from what I have read, CPU Package may indicate the temperature of the socket.

I know this is an Intel, and likely may have no relevance to this thread, but it is showing that there is a big difference between CPU and Core, and sometimes CPU package temps. Also, all of these programs that measure temps, may be measuring something other than CPU Core temps. Just a thought.

 

[Erick Aguilar]

CPU package temps are ridiculous when I test and look at HWmonitor, listing my package temp at 105 celcius.
First time I saw though.. I went crazy.

I decided to purchase a new power supply since I managed to get my chip to 4.4GHZ but not stable, I am guessing it has to do with voltage. Temperatures are fine, and system has not acted up due temperatures. Core temperature reached 65 at 4.4GZ with 1.46 volts

 

[ICIT2LOL]

Hum Erick has got me thinking and I know that has got a few panicking probably but I am just wondering if perhaps the PSU is "over volting" the CPU.

Now that maybe a faulty sensing issue on the core/s or the board.

 

[Erick Aguilar]

Hum Erick has got me thinking and I know that has got a few panicking probably but I am just wondering if perhaps the PSU is "over volting" the CPU.

Now that maybe a faulty sensing issue on the core/s or the board.

I believe it's actually undervolting. I set 1.46 on the BIOS, but tests show 1.42 as top voltage.

 

[ICIT2LOL]

Hum Erick has got me thinking and I know that has got a few panicking probably but I am just wondering if perhaps the PSU is "over volting" the CPU.

Now that maybe a faulty sensing issue on the core/s or the board.

I believe it's actually undervolting. I set 1.46 on the BIOS, but tests show 1.42 as top voltage.

Ok Erick as I don't know of anyway to test the mobo or core sensing devices I suppose one will have to rely on readouts of the machine. I am assuming you have already tested the output from the PSU so that is probably out as a cause for that under volting.

It almost makes you think that the CPU is trying to compensate for it somehow via the motherboard.

 

[Erick Aguilar]

Hum Erick has got me thinking and I know that has got a few panicking probably but I am just wondering if perhaps the PSU is "over volting" the CPU.

Now that maybe a faulty sensing issue on the core/s or the board.

I believe it's actually undervolting. I set 1.46 on the BIOS, but tests show 1.42 as top voltage.

Ok Erick as I don't know of anyway to test the mobo or core sensing devices I suppose one will have to rely on readouts of the machine. I am assuming you have already tested the output from the PSU so that is probably out as a cause for that under volting.

It almost makes you think that the CPU is trying to compensate for it somehow via the motherboard.

And Just like that I was able to hit 4.4GHZ somewhat stably at 1.45 real volts. But a little too close to the TJ max I set as 74. I will invest in a Noctua NH-D14. Not only for overclocking purposes, but for the upcoming summer.