extreme heat issue - a8n sli + a64 3500+

[dmfcomputing]

Hey, im new posting here but have been reading these forums for advice for a while now
My hardware is in my sig, my problem is with the CPU's temps. Theyre insane.
Since i got my new RAM ive been trying to OC somewhat, cause thats why i paid $40 extra for a winnie. The RAM is rated for 2.5-4-4-7 1T at DDR500 for up to 3.1v
I was able to boot fine with 240 HTT, 3x at 1:1 (DDR400 setting)
What i noticed with Asus Probe was, my temps were somewhat high for a winnie considering i was using as5 and a good HSF, idling at about 45C and hitting 57C prime95. prime95 errored out with some roudning error after about 5 min, subsequent testing showed it wasnt the RAM. i upped the voltage to 1.45 and later 1.475 (which turned out to just give me 1.5). The first lengthened my prime95 run to about an hour, the second one hit 66C and just scared the crap out of me so i killed it. I took off the HSF, cleaned it well with arcticlean, put the small dab of as5 on and did the twist thing with it. with no OC, my temps are currently 40C idle and 48c prime95. Thats still much higher then what i think it should be... anyways i just want a way to hit 2.6ghz and run 1:1 with my RAM, and not kill the thing. Ill go up to 1.5v, no further, i have to keep this thing for a while. but im afraid to try 1.5 again as i dont think 66C was a good thing (i bought a winchester, not a prescott)
any advice on how i can keep a stable 2.6 without cooking this damned thing?
-D

 

[fstime]

lol, well 66C is VERY HOT, max die is 65, I wouldnt reccomend those temps unless you want a new cpu soon. What hs/f are you using, I reccomend a xp-90 because if you run iy 66C with the max die being 65, your killing it quick.

 

[StuffedMonkey]

Are these Prime runs all off cold boots? MSI has a problem calculating temps properly off cold boots and Asus may have the same. Off a cold boot at 100% usage my cpu temp reports 59-60C, but immediately rebooting at 100% usage again reports 47-48C. Poor contact with the heat sink could be another. You could try a piece of paper between the HS and CPU and see if it comes out easily. If so, you have a contact problem. Another problem could be poor contact with the TIM within the chip itself, although that is very rare.

 

[dmfcomputing]

as is in my signature, HSF is a zalman 7000b-alcu, which accoding to reviews i looked at before buying was about as good as an xp-90, but i like the quiet (its pretty silent at full speed)
i redid the installation to make sure such contact problems were fixed if they existed, i couldnt get it any tighter then it is now though, and it has th recommended size of a5 on it. i havent really tried ocing again since, i will soon and see if i get a repeat performance... though the current idle/load temps seem too high. and there is really no such thing as a cold boot on my machine as its always on I think i may just avoid changing voltage for now... how high can most winchesters get on 1.4v?
Fo the paper idea, do you mean i should reinstall the HSF with a piece of paper insde and then see if it comes out, then take it off again and put it back on? id need a new tube of as5 nd new articlean for that, i dont see what i wuld accomplsh tho.
Is it possible i just have a bad chip? or does this mean i must have done something wrong.... twice... :\

 

[Duvie]

Originally posted by: StuffedMonkey
Are these Prime runs all off cold boots? MSI has a problem calculating temps properly off cold boots and Asus may have the same. Off a cold boot at 100% usage my cpu temp reports 59-60C, but immediately rebooting at 100% usage again reports 47-48C. Poor contact with the heat sink could be another. You could try a piece of paper between the HS and CPU and see if it comes out easily. If so, you have a contact problem. Another problem could be poor contact with the TIM within the chip itself, although that is very rare.

bullSHIT!!! I tested this on my MSI baord and it is not true...I tested a week ago for another AT memebr and warm boot or cold boot no differencein reported idle or load temps.....

MSI has no problems it is likely idiots using older bioses with issues...1.37mod and 1.41mod do not have this issue....


Beware some ppl dont know what they are talking about....It appears this may have been more of a issue with newcastles and other MSI non NEO2 boards....

 

[steamnputer]

Typically when I OC I would read higher then actually were by about 4-5 degrees, i think that has to do with the voltage change as TC's are voltage biased(as they read input voltage changes so does the temp value) this was read by affixing a TC to the top the heatsink and geussing it was about 3 degreess hotter. Goodluck trying to get a piece of paper between the heatsink and the die, If you can you got some bad problems. The AS5 took about 2 weeks of 12 hours aday to set in now I run stable temps of 49-50 at stock speeds and HSF all the way down with p95 on toruture test. My HSF should cool just slightly better then yours at slow speed but here is an idea of what i got my last p95 run.
was clockin at 2200mhz 1000%load
vcore was 1.4 ish
with C&q engaged after 11.5 hours
cpu via asus probe and confirmed with sistrand =49C
mobo via asus probe and confirmed with sustrand =38C
chipset via TT TC was 51.3 on the topside.
case ambient was 31C
C&Q only clocks back when there is no load for the cpu, so it was effectly stock.
steamnputer

 

[Duvie]

I used the paste that came with my TT cooler and the temps have not budged more then 1-2c since I tested initially after installing it...My max temps are about the same...

In the paste when I have used AS5 I usually ran prime95 for 24 hours and FH for additional days and the temps were usually fully set within 3 days....never lower after that....with my P4 it usually started to climb weeks after that as the system(fans) needed to be cleaned constantly to keep temps acceptable

 

[steamnputer]

duvie just out of curosity are you booting off the seagate or the maxtor?

steamnputer

 

[fstime]

I would suggest burning in the cpu at the lowest posible voltage setting for 2 weeks then... Should allow you to push without to many volts.

EDIT: Fold or use CPU Burn.

 

[PetNorth]

Originally posted by: BouZouki
lol, well 66C is VERY HOT, max die is 65, I wouldnt reccomend those temps unless you want a new cpu soon. What hs/f are you using, I reccomend a xp-90 because if you run iy 66C with the max die being 65, your killing it quick.

No, it isn't. 65º is Tº Case not Tº Die.


# dmfcomputing: about temp. Or board reports it wrong or heatsink isn't installed properly. Winchester is a cold CPU not a hot CPU. Period.

 

[fstime]

Originally posted by: PetNorth

Originally posted by: BouZouki
lol, well 66C is VERY HOT, max die is 65, I wouldnt reccomend those temps unless you want a new cpu soon. What hs/f are you using, I reccomend a xp-90 because if you run iy 66C with the max die being 65, your killing it quick.

No, it isn't. 65º is Tº Case not Tº Die.

What?

 

[dmfcomputing]

heatsink is installed properly... but i cant proceed thinking thet the board is reporting it wrong, what if im wrong
ive been using this CPU for about 2 months now btw, i see no reason its not burned in yet
another question: in order to get up to around 2.6ghz like many winnies do, how much voltage is typically needed?

 

[Duvie]

Originally posted by: steamnputer
Typically when I OC I would read higher then actually were by about 4-5 degrees, i think that has to do with the voltage change as TC's are voltage biased(as they read input voltage changes so does the temp value) this was read by affixing a TC to the top the heatsink and geussing it was about 3 degreess hotter. Goodluck trying to get a piece of paper between the heatsink and the die, If you can you got some bad problems. The AS5 took about 2 weeks of 12 hours aday to set in now I run stable temps of 49-50 at stock speeds and HSF all the way down with p95 on toruture test. My HSF should cool just slightly better then yours at slow speed but here is an idea of what i got my last p95 run.
was clockin at 2200mhz 1000%load
vcore was 1.4 ish
with C&q engaged after 11.5 hours
cpu via asus probe and confirmed with sistrand =49C
mobo via asus probe and confirmed with sustrand =38C
chipset via TT TC was 51.3 on the topside.
case ambient was 31C
C&Q only clocks back when there is no load for the cpu, so it was effectly stock.
steamnputer

I boot off of the raptor drive..not a maxtor....

BOth raptor drives are off of the onboard ports 3-4...the seagate is off of channel 1 of a SATA PCI card....All HDDs are SATA...

 

[PetNorth]

Originally posted by: BouZouki

Originally posted by: PetNorth

Originally posted by: BouZouki
lol, well 66C is VERY HOT, max die is 65, I wouldnt reccomend those temps unless you want a new cpu soon. What hs/f are you using, I reccomend a xp-90 because if you run iy 66C with the max die being 65, your killing it quick.

No, it isn't. 65º is Tº Case not Tº Die.

What?

What?...what? lol

http://www.amd.com/us-en/assets/content...e/white_papers_and_tech_docs/30430.pdf

 

[steamnputer]

[/quote]

I boot off of the raptor drive..not a maxtor....

BOth raptor drives are off of the onboard ports 3-4...the seagate is off of channel 1 of a SATA PCI card....All HDDs are SATA...[/quote]

err meant raptor, sorry not enough sleep just drove 900 miles, at any rate did you have any problems with these at all? and have you gone RAID (are going to use the native NV or the SI)yet. Thought about buying a couple of these and going to RAID 0 or 5(is that mirror and stripe? right...) on those they seem to boot up real quick. the seagate is kinda slow. I want to do it before I get any farther with the system since I will be starting from scratch again. I take it you are running CD drives off the other two ports then.- or are you avoiding the lock on 1&2 if there is one.
steamnputer

 

[htaddict]

Originally posted by: Duvie

bullSHIT!!! I tested this on my MSI baord and it is not true...I tested a week ago for another AT memebr and warm boot or cold boot no differencein reported idle or load temps.....

MSI has no problems it is likely idiots using older bioses with issues...1.37mod and 1.41mod do not have this issue....


Beware some ppl dont know what they are talking about....It appears this may have been more of a issue with newcastles and other MSI non NEO2 boards....

:frown:

Duvie,

I have done my research.... have you (checking your MB alone isn't research).
Be careful who you call an idiot and what you call BS.

I have found references to the temperature offset issue on the Abit, Asus, MSI and AMD forums for both Newcastle and Winchester cores.

I made available a utility to check cpu offset temperature calibration as set by the MB/Bios. Maybe your board doesn't see the varability that others do... but many others experience the problem.

I have tried both the 1.4 and 1.5 officially released bioses from MSI on my NEO2 Platinum board and have the same problem with both bioses. I have not found a reference anywhere to a bios that has this problem fixed for those of us with the problem. I will try the 1.41mod bios and post DATA.

Did you actually try running the utility I provided and confirm that the offset is identical on boots when the CPU is THERMALLY cold and when the CPU is thermally warm.

If so please provide the data (at least I provided data).

 

[Duvie]

Originally posted by: htaddict

Originally posted by: Duvie

bullSHIT!!! I tested this on my MSI baord and it is not true...I tested a week ago for another AT memebr and warm boot or cold boot no differencein reported idle or load temps.....

MSI has no problems it is likely idiots using older bioses with issues...1.37mod and 1.41mod do not have this issue....


Beware some ppl dont know what they are talking about....It appears this may have been more of a issue with newcastles and other MSI non NEO2 boards....

:frown:

Duvie,

I have done my research.... have you (checking your MB alone isn't research).
Be careful who you call an idiot and what you call BS.

I have found references to the temperature offset issue on the Abit, Asus, MSI and AMD forums for both Newcastle and Winchester cores.

I made available a utility to check cpu offset temperature calibration as set by the MB/Bios. Maybe your board doesn't see the varability that others do... but many others experience the problem.

I have tried both the 1.4 and 1.5 officially released bioses from MSI on my NEO2 Platinum board and have the same problem with both bioses. I have not found a reference anywhere to a bios that has this problem fixed for those of us with the problem. I will try the 1.41mod bios and post DATA.

Did you actually try running the utility I provided and confirm that the offset is identical on boots when the CPU is THERMALLY cold and when the CPU is thermally warm.

If so please provide the data (at least I provided data).

That is fine, but you have been throwing this around as being fact....It doesn't appear to be on my board through 3 different bioses and I had a relatively old mobo versus most....I aslo asked a few friends who purchased the neo2 and they have no issues...

In most all the threads I ask what bios ppl use...I usually get no response....Is there a chance this is a bios specific problem??? seems like it can be easily....

I have noticed that the worst bioses are the mSI official ones....

Also the program you gave us says for another board altogether....Whats up with that???


The program runs the black screen but it flashes and wont stay up on the screen...I read the readme file and it said not for this mobo and beware or use at own risk on other boards......

I did this a week back for another member where I restarted the system from warm boots and cold boots...I varied the cold boots from 5 min. to 20 min...no difference....I get nothing like you are suggesting or I have seen in a few threads......


Like normalk ppl with week science and little real proof, back large claims across the board without trying to find out if this is a bios issue.....You need to do some more work!!!


Edit: get me a fvcking program that works and I will get back to wityh that data...

Other then that I can redo my warm and cold starts and take screenies for you!!!! I automatically launch FH on my system each time so I get to a good load temp all the time and in the past few weeks I haven't seen more then 1-2c difference.....

 

[htaddict]

Duvie,

get me a ..... program that works and I will get back to wityh that data...

The program is dos based... try opening a command window first and then try running the program (i.e don't doubleclick on it).

Try following the instructions I put in my previous post on how to use it...

1. Open up a command shell (Start/Run, type "cmd" in the open box)
2. Change directory into where you unzipped it (e.g. cd c:\cputmpset)
3. run it by typing "cpu_sensor_offset"
4. It should tell you what your current offset is.
5. run it again to set the sensor offset (if you want 56C offset "cpu_sensor_offset 56")
6. Start up corecenter/speedfan/mbm5 and it should show lower temps

The program works for the Winbond W83697HF and Winbond W83627HF hardware sensors. The MSI K8N Neo2 Platinum uses the W83627HF sensor.

I flashed to the 1.41 Mod Bios to see if the problem still exists... it does.

Test Sequence:

1) System allowed to cool to ambient for 2 hours (it will still retain residual heat for 5-20 minutes)

2) System booted into windows and allowed to remain idle for 3 minutes then rebooted

3) System booted into windows and Prime95 Large FFT ran for 5 minutes then system rebooted after short wait

4) System booted into windows and Prime95 Large FFT ran for 10 minutes then system rebooted immediately


Here are my results corresponding to the above test sequence:

Date, Time, CPU Offset (Deg C)
3/22/2005, 3:56 PM, 61
3/22/2005, 4:00 PM, 58
3/22/2005, 4:12 PM, 55
3/22/2005, 4:23 PM, 54

Now if you run your system 24 hours a day and never let it cool off between reboots you won't see a large fluctuation in cpu offset.

-----------------------------------------

If anyone else would like to check out the large number of threads about temp issues with the Neo boards on the MSI forum go to the Search utility, select AMD64 nVidia Based board only and search for "temp reboot"

MSI Forum Search Utility

 

[Duvie]

I did have it off for overnight 2 nights ago cause I was cleaning out the fans and removing one that was going bad...I never turned it back on until the morning and immediately started Folding again...I dont remeber any differencein reported load temps.....Like I said I have had this since novemeber and have been in the 2.6-2.66ghz range for 90% of it and have used 1.47-1.57v and I have only seen 4c max difference in my prime95 load temps.....49-53c.....consisitency of 49-51 has been there for weeks since I always have a temp program open and visible and have been Folding pretty hard for weeks as well....

I have had it turned off for more then 2 hours on many occasions.....

I will turn it off tonight then start it in the morning to insure it gets cool.....

I didn't do 1) obviously....but I have done 2,3,and 4


I agree there are temp issues on many boards....Are they all still there?? I dont know....Did a bios possiby fix it??? I dont know...Does it appear to be an issue on my board??? Doesn't look like it from my 4 months of use......


many do not have this, and in that case one should start spending more time trying top figure out what is the difference.....Why do some have 6c swings?? why do some have 10c swings?? Why do soem have less then 6c??? This is where we can see what the issue is, and how it can be fixed. Too many ppl dont have this issue, to make one wonder....


Sidenote: What are most ambient case temps??? Mine rarely are morre then 2-3c over room temp...I have a great case and a few fans and my exhaust fans blow out rather cool air. It is not hot in my case, period.....I have herad of issues of cpu temps beiung set relative to case or system temps being monitored elsewhere on the board.....

In that stuff I pm'd you about....what were the 2 second set of numbers for???

 

[Duvie]

Your test results do not match your sequence...

1) where is the cpu idle temps at start and after 3 minutes??? 60-58c are way too high for idle temps
2) the time between results dont match since with your loading procedures if you did not start loading immediately you couldn't have had that temp at 4 minutes according to step 3...

Basically what do those results at the bottom signify and how do they relate to the steps you listed... I dont see it!!!

 

[steamnputer]

Now what would be the point of adjusting the offset, you are basicly just changing the reading to make it look better, even though it may not be. This would be painting a rose red picture of a black rose.

ISO procdures at my plant requires that you use an outside source that of course is calibrated to a master to adjust offsets on anything that controls temps in a process(We run 130 foot long ovens). Adjusting the temp controls to read 150F in my case would be easy should I decide to do it that way, they would always be at 150F. Now if I adjusted the offset or the bias to read 150F(from say 160F) because I think its 150F(65.6C), even though it may be 156F would be a problem should someone actually check the temp becasue there proscess was not quite right.

I would say that you are messing with things that you should not, if you have no way to know for sure what the temps are without something from the outside to confirm what you are changing the offset to(and thats not real easy on the amd64 on my board). Now as for differencs in the temps people are seeing most all electronics on a good day have what about 5% variance, and most standards require 10% tolerance. I would be willing to bet if you built 10 of the same computers exactly 7 of the ten would be the same 2 would be a degree higher and 1 would be a degree lower. There are some many other factors like case ambients, and construction that it would pointless to say any two are alike. BTW 2 degrees above ambient, I want your case Duvie..... 10C above my room temp at the MOBO at least, and thats with 2 fans blowing in fromt the side(not to mention the others) of the case, the radient heat alone must make it 5-10 hotter. When I had the ambient i was reading earlier I was suspending the TC near the nforce chip about 30mm from anything.

steamnputer

 

[dmfcomputing]

ok, its an interesting argument, but it doesnt help my problem... assuming my HSF is in right and my temp monitor is just fine, what else can i do about this issue?
and to get a winnchester 3500+ to about 2.6ghz, how much voltage should i have to use?

 

[htaddict]

Sorry... Clarification of Test Sequence:

1) System allowed to cool to ambient for 2 hours (it will still retain residual heat for 5-20 minutes). System booted into windows. CPU offset measured.

2) System allowed to remain idle for 3 minutes then rebooted. CPU offset measured.

3) Setup and ran Prime95 Large FFT for 5 minutes then system rebooted after short wait. CPU Offset measured.

4) Prime95 Large FFT ran for 10 minutes then system rebooted immediately. CPU offset measured.

The results posted above were with the 1.41 mod Bios. I flashed specifically for this run, since it was said that it had no temp issues.

-----------------------------------------------

The CPU offset is NOT the reported CPU temperature.

We need to start with a basic (SIMPLIFIED for discussion purposes) lesson on how the Winbond hardware sensor chip works with the Temperature Diode in the AMD Athlon CPU.

The Winbond sensor chip is connected to the CPU Diode using a circuit like the one below:

Example of Winbond circuit

The Winbond sensor chip reads an analog voltage (T2) which it converts to a temperature.
In order to convert the voltage to a temperature the chip uses a linear equation like the following:

Temperature = (slope * voltage) + offset

The temperature diode voltage (T2) from the above circuit varies linearly with temperature. For example the voltage increases 1.0 volts for every 10 degrees C.

This relationship between temperature CHANGE and voltage is useful, but it does not get us an actual temperature. To get the actual temperature, we need to know the temperature when the measured T2 voltage is at a specific voltage. This is where the offset comes in. The offset allows us to CALIBRATE the voltage read by the Winbond sensor chip into an actual temperature.

Since the Winbond sensor chip is located remotely from the CPU there will be different absolute voltages due to the length of the circuit and other electrical characteristics of the motherboard. This does vary manufacturer to manufacturer AND from motherboard to motherboard. MSI attempts to determine the offset by performing a calibration routine at bootup.

Once the offset is determined, then the temperature can be calculated by the Winbond sensor chip using the measured voltage, the change in voltage relative to temperature (slope) and the CPU offset.

Example:

Voltage T2 = 1.5 volts
Slope = 10 degrees C per volt
Offset = 30 degrees C

Using the above values the Winbond sensor chip would report the CPU temperature as 45 degrees C.

Calculation: Temp = (1.5*10) +30 = 15 + 30 = 45

So if the calculated cpu offset changes then the reported CPU temperature will change on a degree by degree basis along with the cpu offset.

The utility I posted reads the cpu offset value from the Winbond sensor chip (it can also change it).

-------------------------------------------------------
Q&A

Why might some people not SEE the effect? (Just because a certain person doesn't SEE it doesn't mean it doesn't exist).

They keep their system on 24/7 or they generally only boot once a day. Either way the cpu offset calculated by the Bios does not vary much because the temperature of the CPU at bootup is similar from bootup to bootup. Remember the calibration routine is executed only once at bootup and the offset remains unchanged until the next bootup.


Why do some have 6c swings?? why do some have 10c swings? 1c swings?

I don't know.... All I know is that it does vary.

but here is MY BEST GUESS...

The cpu offset calculated by the calibration routine is sensitive to the temperature of the CPU. Those with highly effective cooling solutions probably see a smaller swing in the cpu offset calculated by the calibration routine, because the CPU temperature at bootup is more consistent (closer to ambient) even when the CPU has been on for some time.

Maybe it has to do with the the electrical characteristics of the Winbond sensor chip/CPU diode circuit specific to their MB.


In that stuff I pm'd you about....what were the 2 second set of numbers for?

Just to set the record straight... I didn't write the utility and I don't claim any credit for it.

The utility first outputs your cpu offset followed by examples on how to pass paramters on the command line to the application. The second set of numbers are just examples on how to use the utility to set the offset.

----------------------------------------------------------------------------------

Now I may still be an idiot... been called worse... but I am a well researched idiot.

For me the problem does exist, even if I don't have a real fix for it.

 

[Duvie]

well you may have hit the nail on the head with what I was alluding to above...My ambient temp in my case is realtively close to my room temp all the time...very good case cooling...this rig was originally built for a p4 2.4c@3.5....Airflow is tremendous with little interference and well place fan and direction....I mean this thing exhaust air that feels as cool as the room air...My legs actually get cold from the air blwing down there, but that is a factor of windchill and not a degree of actual temperature....

I think My cpu reaches that point well before the 2 hours you list as a directive


OP...sorry to hijack thread....However this may be a real result of some of your temp anamolies....You maybe should try a few of these test..cant hurt....

 

[htaddict]

Originally posted by: steamnputer
Now what would be the point of adjusting the offset, you are basicly just changing the reading to make it look better, even though it may not be. This would be painting a rose red picture of a black rose.

ISO procdures at my plant requires that you use an outside source that of course is calibrated to a master to adjust offsets on anything that controls temps in a process(We run 130 foot long ovens). Adjusting the temp controls to read 150F in my case would be easy should I decide to do it that way, they would always be at 150F. Now if I adjusted the offset or the bias to read 150F(from say 160F) because I think its 150F(65.6C), even though it may be 156F would be a problem should someone actually check the temp becasue there proscess was not quite right.

I would say that you are messing with things that you should not, if you have no way to know for sure what the temps are without something from the outside to confirm what you are changing the offset to(and thats not real easy on the amd64 on my board). Now as for differencs in the temps people are seeing most all electronics on a good day have what about 5% variance, and most standards require 10% tolerance. I would be willing to bet if you built 10 of the same computers exactly 7 of the ten would be the same 2 would be a degree higher and 1 would be a degree lower. There are some many other factors like case ambients, and construction that it would pointless to say any two are alike. BTW 2 degrees above ambient, I want your case Duvie..... 10C above my room temp at the MOBO at least, and thats with 2 fans blowing in fromt the side(not to mention the others) of the case, the radient heat alone must make it 5-10 hotter. When I had the ambient i was reading earlier I was suspending the TC near the nforce chip about 30mm from anything.

steamnputer

I agree that without outside confirmation the number is just that a freaking guess... So the real fix needs to come from MSI (never said otherwise).

The problem is that every time I reboot the system reports a different cpu offset. So which one is right? If I cold boot and the offset is 61 then my temps will consistently report 6 degrees higher than if I reboot after a few minutes when the system is warm (offset of 56).

If I remove the variability of the offset then my idle and load temperature are consistent from boot to boot. I can then use them as a RELATIVE reference (e.g. is my CPU hotter than it was before since I raised the voltage to get to 2.7 Ghz)

I used a batch file to log my offsets. Over the course of 3 weeks and 178 reboots I found the following:

Cold boots always between 60-63 (average 61).
Warm boots always between 51 -58 (average 56)