PWMIC temps: broken censor? 90C+

[aboutblank]

So I've been overclocking my Opteron 165 and have been trying to find my max CPU frequency and have been slowly upping the HTT. I started up Prime95 and SuperPI 16M test and RTHRIBL at the same time and saw my PWMIC temp on MBM 5 shoot up to 92C!! Of course I was like :eek: :eek2: :eek: and shut down my computer. 92C is nearly 200F. I had looked at the PWMIC temps before, and had been slightly worried about them.

 

Let me talk about my cooling for a second. I have a watercooling setup with TDX, maze4 blocks with a DD D5 pump and a Black Ice Pro II radiator. The rad has two Sunon 120x120x35mm fans (these babies push air HARD). I have one 120mm fan in the back of my case pushing air out. Pics of the rig can be found here though there are no pics of what I did just a few minutes ago.

 

I understand that there are 2 rows of 3 for the PWMICs. The right 3 have heat sinks on them from stock. When the computer had shut down I unplugged the power and gently touched all 6 of the PWMIC circuits. They were hot, but it wasn't burning my flesh off (could hold my thumb on for about 2 seconds). I just got my OCZ BGA ramsinks so I put 3 of those on the 3 that weren't sinked. I also wedged an 80mm fan in my case and aimed it at the sinks (about 3 inches away).

 

I booted up again at the same settings (this is at 2.9 ghz on opty 165) and temps are a bit lower. I fired up SuperPi, Prime95 and RTHRIBL and started writing this post. At this point (about 10 mins in) MBM 5 reports the PWMIC temps to be at 66C. I can touch all the sinks and they are warm, but I can hold my thumb on them.

 

So, understanding all this, do you think I have a bugged temperature sensor that is reporting temps higher than true ones? I know some people argue that any temp sensor is unreliable due to manufacturers not calibrating the sensors or sensor quality. Do I have something to worry about? At 90C, wouldn't the ICs burn off/catch on fire? Thanks for the support, I really appreciate it.

[Thran]

I booted up again at the same settings (this is at 2.9 ghz on opty 165) and temps are a bit lower. I fired up SuperPi, Prime95 and RTHRIBL and started writing this post. At this point (about 10 mins in) MBM 5 reports the PWMIC temps to be at 66C. I can touch all the sinks and they are warm, but I can hold my thumb on them.

 

So, understanding all this, do you think I have a bugged temperature sensor that is reporting temps higher than true ones? I know some people argue that any temp sensor is unreliable due to manufacturers not calibrating the sensors or sensor quality. Do I have something to worry about? At 90C, wouldn't the ICs burn off/catch on fire? Thanks for the support, I really appreciate it.

 

 

Pretty sure that's a faulty sensor..however, my x1900xt could get up to 90C overclocked on stock and it didn't catch fire, so I don't know anymore. The best way to test is to get a thermometer for electronics and attach the sticky sensor pad to the PWMIC and see what temprature you get.

 

Mine is currently at 43C and I have only 3 intake fans and two outtakes that go over my primary radiator, I get very low airflow through the case as everything is on one of my two watercooling loops. So I doubt your's could even be 66C.

[red930]

The temp diode measures the temp in the general vicinity of the MOSFETS. The diode is mounted on the motherboard PCB not in the actual PWMIC. You'd be better served by simply aiming a fan in that direction.

 

BTW The MOSFETS are good to 110C.

[EdLSmith]

One problem that most find when changing from air to water is PWMIC temps. These need some air. I run a fan for these and my ram.

[supershanks]

see soundx98 - What is PWMIC. PWMIC Location shows the sensor position as ExRoadie describes.

[M_M]

Overclocking X2 or Opteron DC with 1.5V+ and 2.8GHz+ with (almost) no airflow over mosfet area will lead to 80+ degrees celsius reading on PWMIC under full load on both cores.

 

I have a X2-4400, and when I OC to 2.7GHz @ 1.5V, with decent airflow over mosfet area generated by 60CFM 120mm fan on SI-120, I still get 65C reading under high load. When I had A64-3700+, OCed to same 2.7GHz @ 1.5V, I had PMWIC reading around 55C under high load (in both cases case temp is around 33-35C). This is because dual core CPUs are using more power, which is comming to the CPU through mosfets.

 

Note, at 2.9GHz @ 1.55V, your Opteron165 will use up to 110W*(1.55/1.30)^2*(2.9/2.4)=190W of power!!! (assuming that DC cpus have TDP=110W @ 2.4GHz @ 1.3V)

 

My advice, if your intention is to keep the CPU at 2.9GHz, is to mount some quiet 80 or 92mm fan to blow into the mosfets, trying to keep PWMIC temp reading below 70C.

[supershanks]

@M_M

You mean lias in :-

aboutblank Post #1

I just got my OCZ BGA ramsinks so I put 3 of those on the 3 that weren't sinked. I also wedged an 80mm fan in my case and aimed it at the sinks (about 3 inches away).

 

M_M

My advice, if your intention is to keep the CPU at 2.9GHz is to mount some quiet 80 or 92mm fan to blow into the mosfets

[red930]

My advice, if your intention is to keep the CPU at 2.9GHz, is to mount some quiet 80 or 92mm fan to blow into the mosfets, trying to keep PWMIC temp reading below 70C.

I prefer a big 120mm fan that turns around 1,200rpm. It keeps the air stirred up with little to no fan noise.

[toddm27]

I was having temps up near 90 to 95c when I was pushing 1.65v and trying to stabilize 3.1 on my opteron 165, that was on water with a 120 blowing over the top, I've since redone it with a 60mm fan tyed into place and pointed towards the mosfets lowering the temp to a max of 65c. that was on an ultra d btw

[M_M]

I prefer a big 120mm fan that turns around 1,200rpm. It keeps the air stirred up with little to no fan noise.

 

I prefer 120mm as well (having 4 of them in total in my case, all between 1000-1200rpm under normal load), but when modding the inside the case it's usually easier to fit a 80 or 92mm fan (speeds up to 1600rpm for low noise).

 

BTW, one question for community:

 

Are there slight variations in the power consumption beween "same" CPUs on same frequencies and under same voltage (on same motherboard) and under same application load (for example Prime95)? There shouldn't be, right? I somehow I have a feeling that some CPU, just different steppings (or production date), generate slightly more or slightly less heat then others... Just wondering have someone else got the same feeling? I have this hunch based on the temperature reading of CPU & PWMIC, measured with two different X2-4400 CPU's (steppings 0525UPMW and 0534VPMW, both E6 cores), everything exactly the same (including the application load), and the temperatures (both CPU and PWMIC) are consistently 2-3C lower with one then with another?

 

Just wondering could this be some measurement or sensors fault, or have someone else spotted this as well? I know, that ultimately I would have to measure the how many amps actually goes into the CPU itself but I don't have the equipment to do so. Maybe someone did this aleady?

[toddm27]

I prefer 120mm as well (having 4 of them in total in my case, all between 1000-1200rpm under normal load), but when modding the inside the case it's usually easier to fit a 80 or 92mm fan (speeds up to 1600rpm for low noise).

 

BTW, one question for community:

 

Are there slight variations in the power consumption beween "same" CPUs on same frequencies and under same voltage (on same motherboard) and under same application load (for example Prime95)? There shouldn't be, right? I somehow I have a feeling that some CPU, just different steppings (or production date), generate slightly more or slightly less heat then others... Just wondering have someone else got the same feeling? I have this hunch based on the temperature reading of CPU & PWMIC, measured with two different X2-4400 CPU's (steppings 0525UPMW and 0534VPMW, both E6 cores), everything exactly the same (including the application load), and the temperatures (both CPU and PWMIC) are consistently 2-3C lower with one then with another?

 

Just wondering could this be some measurement or sensors fault, or have someone else spotted this as well? I know, that ultimately I would have to measure the how many amps actually goes into the CPU itself but I don't have the equipment to do so. Maybe someone did this aleady?

I would say if you look at the max t case of different processors, some opteron 165 cpu's have a wattage rating of ccbbe 0610 75w,ccb1e 0608 105w,ccb1e 0606 110w and I believe my first 165 had a rating of ccbwe 0534 65w. so yes I would imagine it makes a difference depending on the wattage rating, affecting the pwmic and temp of the cpu is dependant on how well the ihs was attached

[M_M]

I would say if you look at the max t case of different processors, some opteron 165 cpu's have a wattage rating of ccbbe 0610 75w,ccb1e 0608 105w,ccb1e 0606 110w and I believe my first 165 had a rating of ccbwe 0534 65w. so yes I would imagine it makes a difference depending on the wattage rating, affecting the pwmic and temp of the cpu is dependant on how well the ihs was attached

 

I agree that CPU temperature depends on how good the IHS is attached, but CPU consumption should not be affected by this, right?

 

But, as you suggested, there might be something in different max t case specs, which vary from the particular CPU to CPU (all same type and stepping). How are they (AMD?) detemining this? What it actually means? Is it better (for OC) this spec to be lower or higher?