Temperature Monitors wrong?

[Lightstylez]

So i finally decided to upgrade my Amd Phenom II Black Edition X6 1090T from stock cooler to Corsair H80 water-cooling, which arrived earlier today. When using stock i hadn't oc'd my CPU at all but both my CPU and mobo (Asrock 890FX Deluxe4) supported Turbocore technology which made the 3.2ghz stock sometimes made half the cores go up to as high as 3.8ghz if lucky although usually hovering around 3.6ghz average. I never really monitored my CPU temps because i assumed i wouldn't need to unless i OC'd it, but i found out that my CPU temps were going between 60-70 C under high load.

Anyway i installed the H80 earlier today & did a lot of tests with prime95, it improved cooling massively without tweaking anything it went from 70-72c in prime95 torture test to 48c maximum. Now for some reason when turbocore is enabled i cant see the CPU CORE temperatures. Anyway, since my cooling was decent i Disabled turbocore & overclocked my CPU to 3.85ghz and ran a torture test for 1 hr with no problems, its still working as i type this so all is well. But after doing more prime95 tests and using multiple monitoring apps im not sure the temperatures are correct?

The highest my CPU went with the new OC in prime95 torture was 60c, and yet the cores were only 35c , the cores usually around 20-23c as idle, and CPU temp 32-35 idle, are the cores really suppose to be that much lower than cpu temp. I heard that there is problems with some mobos not accurately reading x6 CPU's? also Speedfan and OC tuner had a whole 8c cpu temp difference at one point with both programs running idle, which is weird to say the least. I've been using monitoring programs such as ASRock OC Tuner, Speedfan,CPUID HWMonitorPro, CPU-Z, AMD overdrive and others just to make sure.

All of them show that CPU temps are definitely a hell of a lot higher than my Core temps both idle and under load. Is this normal?

Also If i were to overclock even more whats the max temp i should stop at ?

 

[trents]

Add about 15c to the core temps as a realistic offset. The Thuban core temp sensor is no very accurate. Yes, it is common with good aftermarket cooling that core temps are lower than CPU socket temps, typically by about 10c. There are some good reasons for this. Mostly, there is a seam between the processor and the socket that interferes with transfer of heat between the socket and the processor and since the cooler is making direct contact with the processor it is able to efficiently remove heat from that component but the seam slows the transfer of heat from the socket to the processor. Also there are some very hot running components around the socket, namely the capacitors and the VRMs that are contributing heat to that component.

 

[Lightstylez]

This is just after PC boot, at idle temps. Easier to see if you click to view full size - sometimes Speedfan and OC Tuner have a 4-8c difference between the CPU monitors?

After using Torture test-Blend in Prime95 for 20 mins

Temps 2 mins after Blend test

For some reason it seems my CPU temps are quite high considering i have my H80 On max cooling, i've seen a lot of people using the h80 or similarly cooling efficiently hardware (I have seen cooling hardware comparisons) and they seem to be able to get a lot lower temp even with 4ghz+ with the 1090T BE processor

 

[RGone]

Since the 'same' questions are asked again and again, I keep the answers saved so don't have to type them over and over. Here is where we most generally are about AMD temps and the way too many programs for displaying some sort of temps.

Let us just say a couple of things about AMD cpu temps. It is best to read what is said here and not what you think or supposed or something you heard, that is rumored.

1. They are not calculated/measured/derived the same as Intel cpus.

2. The 'closer' you get to a real AMD cpu temperature, the 'closer' you get to proprietary/make that secret information that will NOT be released to the general public. Period. This non-release of information includes 3rd party program writers. In the end they too have to make a highly educated guess at how AMD is measuing/reporting temperatures of their AMD processors.

3. Any piece of software has a learning curve. M$ Office, Win 7, Adobe products etc are all installed and at some point you will have learn how that piece of software will do what you want it to do within the designs of the software itself. Someone super skilled in WordPerfect is not going to be able to tell you as much about M$ Office as someone that is skilled in M$ Office. The certainly should make sense.

4. So based on the logic of #3 above, users that you see actually 'helping' AMD people most of the time have programs/applications that THEY are MOST familiar with. In the AMD sections where you see people like "trents", who is helping out ALL THE TIME, uses HWMonitor by CPUID dot com. He does not use What's My TemP, How's My CPU doing now?, CoreTemp or anything else but HWMonitor by CPUID dot com. So get in step with the software he and most of us are used to dealing with. That is really where you need to be if going to ask us for help. We don't use other temp/mobo monitors.

5. Then if YOU want to learn another temp/mobo monitoring software, you can do it at your own leisure, because in general we who try to help are not off somewhere with different software.

6. I hope that makes the situation a little more clear. In general we use HWMonitor and have learned HOW IT works more, make that much more, often than not.

 

[trents]

To keep it simple and to get the maximum information out of one program interface, as RGone suggested, I would recommend using HWMonitor. It gives loads of information about PSU/CPU/GPU voltages and temps and records maximum/minimum and current values. Sometimes HWMonitor gives a generic label to the CPU socket temp like "TMPIN_" so to check it against OC Tuner for that particular value may be helpful in figuring out which one of those TMPIN_ lines corresponds to CPU socket temp.

Those temps do look high for a H80 cooler. What is your CPU voltage set to in bios? AMDOD is a very unreliable program for reporting some voltages, especially when they have been taken off of stock. AMDOD tends to report some things like that at whatever the factory stock setting is, regardless of what you change it to in bios. Please download and install CPU-z and post pics of these three tabs: "CPU", "Memory" and "SPD". It will give us a more accurate report of the CPU core voltage.

HWMonitor and CPU-z should be all you need and will simply your monitoring.

Also, please don't attach pics of the whole desktop when you post as even when you enlarge the image it is difficult to read the values in the software interfaces, at least on my 19" monitor. They are just too small. Please crop and attach each one as a separate image. Snipping tool in Windows accessories is great for this. Yeah, I know, it's a little more work.

 

[Lightstylez]

To keep it simple and to get the maximum information out of one program interface, as RGone suggested, I would recommend using HWMonitor. It gives loads of information about PSU/CPU/GPU voltages and temps and records maximum/minimum and current values. Sometimes HWMonitor gives a generic label to the CUP socket temp like "TMPIN_" so to check it against OC Tuner for that particular value may be helpful in figuring out which one of those TMPIN_ lines corresponds to CPU socket temp.

Yes i have that program also, but it is buggy for me it doesn't display all of the info, it has the words TRIAL on half of my temps and things like fan speed are obviously displayed wrong. It showed me CPU fan was going from 2000rpm to 10000rpm?
I showed multiple programs that work best for me + any time that HWmonitor works for certain temps/speeds its usually the same as the other software im using just not more in-depth. In any case im unsure why it HAS to be HWmonitor, for the info to be more accurate in comparison to other people on overclockers sure, but i just want an opinion on if my CPU is hotter than it should be and if it would be a bad thing if i overclocked it more according to what i have displayed.

Sorry if its hard to see the info, its what i personally have been using to help record statistics - my CPU voltage is 1.44 i believe. I will try to re-do some tests with CPUZ + HWMonitor but really im unsure how anyone could help me, my voltages are a lot lower than other peoples who have OC'D to similar speeds. I have made sure they are the lowest possible stable voltages.

 

[Lightstylez]

CPU

Memory

SPD

HWMonitor

I will start running prime95 and relay the info, i have a feeling my fans on the H80 fan controller are not working as efficiently as possible. It does not sound like its getting faster when changing between the 3 speed settings, i have tried holding down the button for 5 secs forcing a reset, the fans then sound like a turbine going fast and then 2 seconds after it goes back to normal speed, all programs i have used say the CPU_FAN speed is 2190rpm and it doesn't seem to change no matter what i try. even so, the 3 setttings low -1500rpm, medium 2000rpm, high - 2500rpm its not exactly low anyway, even when under 100% load for 20 mins the fan speed never changes both cool n quest and turbocore are disabled in the bios.

 

[Lightstylez]

OK 20 mins Prime95 Blend test -

I took the screenshot a little late but the max voltages/temps are shown just before prime95 stopped.

The max temps are pretty high, the H80 was on max fan setting and although the fan speed monitor doesn't necessarily show the fan going much faster (it was on 2163 speed most of the 20 mins) it definitely sounded as though the fans were going a lot faster, i think my board is not monitoring it properly. It felt as though my pc was really giving off a loooot of heat and im pretty sure my CPU voltage is around 1.44/1.45 and yet HWmonitor is showing upto 1.53v for some reason. is that normal?

UPDATE: INFACT i was just able to to email my bios screenshots from my phone, cpu voltage only at 1.387v

 

[trents]

Go with the "non Pro" version of HWMonitor unless you want to buy the pro product. I think that's why you're getting "trial" messages.

 

[trents]

Your max CPU vcore is higher than what you have it set to in bios because of LLC (Load Line Calibration), a vcore supplementation technology that many motherboards have either built in or as an option in bios. It helps with stability under load while maintaining lower voltages at idle.

 

[Lightstylez]

Your max CPU vcore is higher than what you have it set to in bios because of LLC (Load Line Calibration), a vcore supplementation technology that many motherboards have either built in or as an option in bios. It helps with stability under load while maintaining lower voltages at idle.

Yeah thanks i just got the Non-pro version of HWMonitor, also is it good that LLC is enabled? i mean even while idle its still higher voltage. Unless i needed that much to boot anyway and 1.387v is too low in the first place.
Also a max of 1.53v sounds really high i heard 1.55 is the max you should even go to for the 1090T usually with a lot higher clock then i have currently.
Can anyone comment on these screenshots any idea why its as hot as it is? I spent £80 on the H80 cooling, its not the best but from performance reviews i thought it was a good price. i was hoping to get at least 4.0ghz maybe 4.2/4.3 which i have seen a lot of people on the overclockers forums get to with the 1090T.

 

[trents]

LLC will also give a smaller vcore boost in idle. Do you have a bios line item for LLC? If so, it might just give choices like Enable or Disable or it might give finer adjustments. The thing you need to realize is that it will take the same max vcore to keep the system stable whether or not you do it with LLC. Yes, 1.55 is the max I would feel comfortable with.

Tell us about your case fans and ventilation. Antec 900 should have pretty good ventilation if the fans are oriented correctly.

Have you added any voltage to your CPUNB? That usually helps with stability and may enable you to get by without quite so much vcore.

 

[Lightstylez]

Well i looked around a lot of different forums to see what the best orientation was for the h80, most people said it would be better to have both fans pointed towards the inside of the case so that cool air from outside cools the radiator, although this would make the inside of the case hotter, but i really don't have any problem with GPU/mobo heat.

Even on the high-end games my GPU never goes past 50c and i usually only keep the fan on 50% (its like an aeroplane on 100%) my top fan is taking air out the case and my 2 front fans taking air in, all fans are on high setting.

Now i looked around a long time for the best side case fan(which i was going to order with the H80), i saw ones up to 5000 rpm but people said they were obnoxiously loud. I opted for a Coolermaster Excalibur, said to have amazing airflow and more than decent cooling without too much noise. What i didn't realise is the H80 was so wide it wouldn't allow me to put my sidepanel on even though i saw people saying they fitted the H80 in their Antec 900 before i bought it, they didn't seem to mention this problem.

so both the front and back are taking in air and my top fan is sucking air out but with no sidepanel, although the open side of my pc is near my window which is almost always open. Its kinda of a bad time to of bought this stuff atm because where i live we started getting a massive heatwave temps upto 23c, usually its more average of 15c where i live. still i doubt that would be a massive contributing factor. Im sure that i did increase CPUNB a small amount of voltage, i suppose i could keep decreasing voltage of CPU+CPUNB and testing to see how low i can go while still being stable - although this would take a lot of time and im unsure how much this would even help.

EDIT: http://www.overclockers.com/forums/showthread.php?t=679606&page=4 - 2nd post down this guy has 6c more minimum temp and 13c more maximum temp than me and yet this guy is only using stock cooling + he's been running Prime95 1 hr 4 mins, while i get my temps in 20 mins, im sure it would be a lot closer if i actually ran it that long.

 

[trents]

My only other suggestion would be to remove the water block from the CPU, thoroughly clean the two surfaces, reapply the TIM and reinstall the block. If you do this, take careful not of how good the TIM coverage was when you remove the block. Poorer than expected cooling performance can be caused by incorrect TIM application or a poorly seated water block.

 

[Lightstylez]

My only other suggestion would be to remove the water block from the CPU, thoroughly clean the two surfaces, reapply the TIM and reinstall the block. If you do this, take careful not of how good the TIM coverage was when you remove the block. Poorer than expected cooling performance can be caused by incorrect TIM application or a poorly seated water block.

Yeah well i bought this http://www.amazon.co.uk/gp/product/B003VRBPAI/ref=oh_details_o00_s00_i00
Thermal material remover and purifier which i used before attaching the H80, there was even a guide on the site saying how much paste to apply according to the CPU, it told me to put a small dot in the middle for mine.

What i didn't know is the H80 came with a pre-applied self adhesive, i found someone who already asked whether to use the pre-applied stuff or to clean it and use a thermal paste - most people said to just use the adhesive and it would still be very good.

One thing i will try when i have time tomorrow is to take the 2 CPU fan headers out of the block and plug them into the mobo, i heard some people got about a 6c decrease in temp because the fan only increases in speed when the coolant inside the H80 gets to a certain temp, even if the cpu gets really hot it wont go higher it only triggers by the coolant. So by plugging them directly to the mobo i should just be able to choose the fan speeds with Speedfan program and monitor them more accurately that way, i will see how this goes and then i guess i will try to clean/reapply TIM and then if both don't help enough then i may turn the fan orientation, although this is a last resort, it really took me a long time installing this with the limited space inside my case atm.

Also is there anything else in the BIOS that needs changing/disabling? C1E is currently enabled (ive seen mixed views on whether to keep this on or off) and there are a few options in the bios which i dont understand things like c-state support, l3 cache - auto/bsp only/all cores, cpu thermal throttle e.t.c

What temps would you expect me to be actually getting compared to what im getting now, is it really that bad? i haven't really OC'd before, but i have custom built a lot of PC's

 

[trents]

Did you use both the self-adhesive TIM that came with the water block and extra TIM on top of that? That's what it sound like from your description and I would think that would be excessive and create an insulating condition. Maybe I misunderstood what you said.

What is your CPUNB voltage set to? It may be just called "NB" in your bios but don't get it confused with the chipset NB.

 

[Lightstylez]

Did you use both the self-adhesive TIM that came with the water block and extra TIM on top of that? That's what it sound like from your description and I would think that would be excessive and create an insulating condition. Maybe I misunderstood what you said.

What is your CPUNB voltage set to? It may be just called "NB" in your bios but don't get it confused with the chipset NB.

Erm no, like i said i looked up what was best, i didn't apply the paste that came with the cleaners, i just cleaned the CPU up from the use my last heat-sink i then seated the H80 with the self-adhesive i still have all of my ordered paste left, althought im not certain how good Arctic silver 5 is i've heard it conducts well with everything so you have to be really careful

Well in the bios picture above - my CPU NB is 1.1375v but there is also NB voltage below that 1.145v
I havent changed anything since i made this post all my screenshots are still valid, i wont be changing any options yet

 

[trents]

I would try bumping your CPUNB voltage up to 1.25-1.25 and the CPUNB frequency to 2600.

 

[Lightstylez]

I would try bumping your CPUNB voltage up to 1.25-1.25 and the CPUNB frequency to 2600.

Ok i will do that soon (EDIT: 1.25-1.25? is that a mistake?), i just finished doing a prime95 10 minute blend test - both with the fan headers into the h80, and also with the headers plugged into the mobo instead.
The results were pretty much the exact same for all 3 tests - idle temps, 10 minutes into P95 & also 45 seconds after stopping prime95 to see how fast they cool.
Also HWMonitor had ALL of my fan speeds wrong, going from 0 to 49000 on all of them for some weird reason, but Speedfan did show while plugged into mobo they are actually running at 2500rpm like advertised so im sure that the H80 is definitely doing its job.

Also i saw this before i went to bed off a forum

is it normal for the cpu socket temp to be 15-18 degrees C higher than core temp? when i prime95 my system my max core temp is 42 while my cpu socket temp is 61. on idle its 31 core temp and 45 cpu socket temp.

Most likely a stuck sensor. I wouldn't worry about it. The core temps are what you should be worried about.

yes socket will be higher then core, the socket dosent have a heat sink where your core has a heat sink on top of it

Thats not how it works. Heat transfers though the socket, into the chip, which then goes into the heatsink. Having direct contact with the Heatsink doesnt make a difference with the Socket.

OP, usually the socket is 5-10 degrees warmer. Like i just stated, its because the volts go through it into the chip. Not all the heat can go through the CPU, so some may transfer back to the socket.

I found multiple posts on multiple site people saying that their CPU temps are indeed also 12-18 degrees hotter than the cores

Like the guy said saying a stuck sensor? he was talking to someone else but it may be the case for me aswell
In this IMG the CPU actually spiked from 60c to 127c for 1 second, then again could just be a program error also.

 

[trents]

Do you have multiple CPU fans running off a splitter from the same motherboard header? If so, HWMonitor will show erratic fan speeds at times but if you watch it it will for a while it will also intermittently show the correct speed.

Yes, the differential you see between core temp and CPU temp is normal with Thubans. Add about 10c to the core temp for an offset. I'm not sure the sensor is stuck it's just poorly calibrated.