Very strange problem

eX&TriC · 02-11-2002, 03:20 PM

I have a very strange problem with my current WC
I am using a 0.13µ CPU that puts out 45W at default
but when I boot the temp of the CPU is +-42°C
When I overclock to 2.4 @ 1.65V and STRESS the CPU
the temp rises to 46°C

I dont get why the temp at idle is so high...

Its not the radiator or watertemp because they are about
25°C its not the flowrate because I only get a 4°C rise when
overclocking and raising Vcore ... BUT the strangest thing
is that when I put a temp sensor to the block and
CPU I get a reading of 32°C. The waterblock is cold to the
touch.

My first thought was that the temp reading were wrong. So
I put the standard HSF on and here I get temps under 40°C
so the tempsensors are right

I am out of ideas
Some help please ...

----------------------------------------------
Setup :

BeCool waterblock
Eheim 1048
Radiator VW golf
Purple Ice
120mm Enermax
90mm Papst
Artic Silver 2

Th7II Vmod (Vcore/Vmem)
P4 1.8A > 2.4 Ghz @ 1.65V
256 MB PC800 > PC1066 2.75V
...

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

Kevin · 02-11-2002, 03:39 PM

I always hate when people suggest this to me... but are you sure the block is mounted correctly?
-Kev

DigitalChaos · 02-11-2002, 03:40 PM

the answer lies between the cpu and the block...
if the cpu is hot and the block is cold... its inbetween.
are you using thermal paste?
is everything flat?
are you using a shim that is lifting the wateblock up over the cpu?

also.. what are you reading cpu temps with? sounds to me like its a mobo diode that is way off... but still gives the normal delta in idle/max

here is something for you to try... get a cpu cooling program, and if your cpu doesnt drop down to atleast 10c over ambient... THEN you know the diode is off.

eX&TriC · 02-11-2002, 03:52 PM

the block is mounted using a 1cm thick plexi-plate
and 4 screws that are very thight. I am using a
very thin layer of artic silver 2. So that isn't the problem
either. I remounted the block over then times now and
it still stays the same ...

I use both MBmonitor and Winbond Hardware monitor
to get the temps

Kevin · 02-11-2002, 04:37 PM

I DEFINITELY like DigitalChaos's idea about using a CPU software cooling program. Leave it on for an hour or so and report back what happens w/ the CPU temperature temps...

DigitalChaos · 02-11-2002, 04:57 PM

hehe, thanks kevin

im not sure of any p4 cooling progs but im sure there are some.
maybe someone on the board could recommend one for ya

gmat · 02-11-2002, 05:51 PM

Maybe a bad heat transfer due to a slightly warped WB base ? Did U lap it ? And if yes did U make sure it was flat (at least optically, using reflections) ?

Brad · 02-12-2002, 03:31 AM

something to do with transfer between you block and cpu seems to be obvious. Why don't you remove your IHS, and mount the block on top of the cpu, it will help greatly

eX&TriC · 02-12-2002, 06:05 AM

The base of the block is so perfect i can
see myself in it ...
Removing the heatspreader didn't help a
bit with my previous P4 but i'll try

Brad · 02-12-2002, 06:24 AM

the reason why I suggested that is because the ihs is very poorly attached to the core sometimes

eX&TriC · 02-12-2002, 09:03 AM

After removing the HS the temps are :

39°-42° idle
42°-45° stress

the delta of 3° shows that the block is doing its
work fine and the mounting is OK IMHO. But I still
don't get why the idle temp is so high ...

PS : the Northwood core is shiny black instead of
dark blue

EMC2 · 02-12-2002, 12:38 PM

How do you define idle and stressed? (what are you running to get to an idle state and what are you executing to put it under stress)

As I recall almost all P4 motherboards use the internal CPU thermal diode to measure temps. Does yours? Or are these readings from something else?

Are you sure that the processor isn't being throttled back by the system? (some P4 systems implement dynamic power management and will actually throttle back the processor depending upon various factors, in some cases including fan speed).

eX&TriC · 02-12-2002, 12:42 PM

idle = W2K without doing anything
stress = CPU stability test normal

temp is measured by sensor IN the CPU

pHaestus · 02-12-2002, 01:34 PM

I am not too familiar with the register settings for P4 mobos, but many AMD motherboards do not implement the CPU_HLT command that causes the processor to truly idle. This was done for supposedly compatibility and also to avoid huge swings in current demands when the CPU goes under load. You might want to look around for a WPCredit file that is for the i850 (that is what the TH7-II is right?) and see if maybe the CPU is idling high in temp for that reason. 46C under load seems high to me for a P4 anyway; aren't those pretty low heat CPUs?

niner · 02-12-2002, 02:57 PM

Sorry for being off-topic here. But can you tip me on how to remove the heat speader on a P4? I want to learn from someone who has done it before I ruin a $400 CPU.

Thanks,

eX&TriC · 02-12-2002, 05:55 PM

take a sharp exacto knife and put the blade out 2 cm
start at the corner and put the blade against the corner
1 cm above and 1cm below the corner
start moving the the knife making sure the blade is in the
goop between the HS en CPU base. You can do this for
the four corners and then slide the knife from one to
the other. Don't put the knife in too far and make sure
you don't toutch the die

good luck

EMC2 · 02-12-2002, 06:07 PM

I don't know that I would consider W2K doing nothing necessarily as being at idle

If you can find a software cooler for the P4, the Stop-Grant state of a P4 drops the power to almost 1/3 of normal.

For the stress side, try Prime95, Folding, or SETI. (Intel also has a proggy called Maxpower that generates the worst case heat load possible, but I don't remember where they hide it).

Note also that besides the auto-throttle that some motherboards have, the P4 also has an internal throttle circuit that reduces the duty cycle of the processor at a set temp. If you want to be sure this isn't going on, run SiSoft's CPU bench after a cold start and record the info. Then run a stress proggy for 30 mins, then re-run the CPU bench. If the numbers change significantly, then you are being throttled.

pHaestus - yes, relatively low. A P4 at 2.2Ghz and default Vcore is about 55W. At the 1.65 Vcore and 2.4Ghz that eX&TriC said he was running that would go up to about 73W.

Brad · 02-13-2002, 04:43 AM

46C isn't too bad really at full load, I'd never say it's stunning, but nothing to be worried about. Clock throttling comes in at 65 - 75C depending on the model of P4.

Also, I'm sure that win2k idles lower than win9x for example. In my laptop, I can use win2k for hours without the fan turning on, while in win98 the fan will turn on within 10minutes. If under full load the fan will turn on in about 5 minutes.

pHaestus · 02-13-2002, 07:11 AM

For 73W heat, mid 40C is not a particularly good temp for water cooling. It is tolerable, but I would expect more in the neighborhood of 12-14C over room temperature (somewhere around 0.17-0.19 C/W) from messing with the internal diode on my XP.

02-11-2002, 03:20 PM	#1
eX&TriC Cooling Neophyte Join Date: Oct 2001 Location: Belgium (center of the universe) Posts: 51	Very strange problem I have a very strange problem with my current WC I am using a 0.13µ CPU that puts out 45W at default but when I boot the temp of the CPU is +-42°C When I overclock to 2.4 @ 1.65V and STRESS the CPU the temp rises to 46°C I dont get why the temp at idle is so high... Its not the radiator or watertemp because they are about 25°C its not the flowrate because I only get a 4°C rise when overclocking and raising Vcore ... BUT the strangest thing is that when I put a temp sensor to the block and CPU I get a reading of 32°C. The waterblock is cold to the touch. My first thought was that the temp reading were wrong. So I put the standard HSF on and here I get temps under 40°C so the tempsensors are right I am out of ideas Some help please ... ---------------------------------------------- Setup : BeCool waterblock Eheim 1048 Radiator VW golf Purple Ice 120mm Enermax 90mm Papst Artic Silver 2 Th7II Vmod (Vcore/Vmem) P4 1.8A > 2.4 Ghz @ 1.65V 256 MB PC800 > PC1066 2.75V ... ---------------------------------------------

02-11-2002, 03:39 PM	#2
Kevin Slacking more than your weird uncle Join Date: Nov 2000 Location: San Diego, CA (UCSD) / Los Angeles, CA (home) Posts: 1,605	I always hate when people suggest this to me... but are you sure the block is mounted correctly? -Kev __________________ I used to throw hot coffee all over the ass of the horse there, then whip him while he was kickin' at me. Those f***in things are crazy.

02-11-2002, 04:37 PM	#5
Kevin Slacking more than your weird uncle Join Date: Nov 2000 Location: San Diego, CA (UCSD) / Los Angeles, CA (home) Posts: 1,605	I DEFINITELY like DigitalChaos's idea about using a CPU software cooling program. Leave it on for an hour or so and report back what happens w/ the CPU temperature temps... __________________ I used to throw hot coffee all over the ass of the horse there, then whip him while he was kickin' at me. Those f***in things are crazy.

02-12-2002, 03:31 AM	#8
Brad Thermophile Join Date: Oct 2001 Location: Nuu Zeeelin Posts: 3,175	something to do with transfer between you block and cpu seems to be obvious. Why don't you remove your IHS, and mount the block on top of the cpu, it will help greatly __________________ 2x P3 1100's at 1400, Abit VP6, 2x Corsair 256mb PC150 sticks, 20gb 'cuda ATA-III, 2x 40gb 'cuda ATA-IV in raid 0. 20" Trinitron. No fans 2x 2400+ at 2288mhz (16.0 x 143), Iwill MPX2, 2x Kingmax PC-3200 256mb sticks, 4x 20gb 60gxp in Raid 5 on a Promise SX6000. Asus Ti4200 320/630. Cooled by Water

02-12-2002, 06:24 AM	#10
Brad Thermophile Join Date: Oct 2001 Location: Nuu Zeeelin Posts: 3,175	the reason why I suggested that is because the ihs is very poorly attached to the core sometimes __________________ 2x P3 1100's at 1400, Abit VP6, 2x Corsair 256mb PC150 sticks, 20gb 'cuda ATA-III, 2x 40gb 'cuda ATA-IV in raid 0. 20" Trinitron. No fans 2x 2400+ at 2288mhz (16.0 x 143), Iwill MPX2, 2x Kingmax PC-3200 256mb sticks, 4x 20gb 60gxp in Raid 5 on a Promise SX6000. Asus Ti4200 320/630. Cooled by Water

02-11-2002, 03:40 PM	#3
DigitalChaos Cooling Savant Join Date: Jan 2002 Location: Waukesha, Wi Posts: 698	the answer lies between the cpu and the block... if the cpu is hot and the block is cold... its inbetween. are you using thermal paste? is everything flat? are you using a shim that is lifting the wateblock up over the cpu? also.. what are you reading cpu temps with? sounds to me like its a mobo diode that is way off... but still gives the normal delta in idle/max here is something for you to try... get a cpu cooling program, and if your cpu doesnt drop down to atleast 10c over ambient... THEN you know the diode is off.

02-11-2002, 03:52 PM	#4
eX&TriC Cooling Neophyte Join Date: Oct 2001 Location: Belgium (center of the universe) Posts: 51	the block is mounted using a 1cm thick plexi-plate and 4 screws that are very thight. I am using a very thin layer of artic silver 2. So that isn't the problem either. I remounted the block over then times now and it still stays the same ... I use both MBmonitor and Winbond Hardware monitor to get the temps

02-11-2002, 04:57 PM	#6
DigitalChaos Cooling Savant Join Date: Jan 2002 Location: Waukesha, Wi Posts: 698	hehe, thanks kevin im not sure of any p4 cooling progs but im sure there are some. maybe someone on the board could recommend one for ya

02-11-2002, 05:51 PM	#7
gmat Thermophile Join Date: Mar 2001 Location: France Posts: 1,221	Maybe a bad heat transfer due to a slightly warped WB base ? Did U lap it ? And if yes did U make sure it was flat (at least optically, using reflections) ?

02-12-2002, 06:05 AM	#9
eX&TriC Cooling Neophyte Join Date: Oct 2001 Location: Belgium (center of the universe) Posts: 51	The base of the block is so perfect i can see myself in it ... Removing the heatspreader didn't help a bit with my previous P4 but i'll try

02-12-2002, 09:03 AM	#11
eX&TriC Cooling Neophyte Join Date: Oct 2001 Location: Belgium (center of the universe) Posts: 51	After removing the HS the temps are : 39°-42° idle 42°-45° stress the delta of 3° shows that the block is doing its work fine and the mounting is OK IMHO. But I still don't get why the idle temp is so high ... PS : the Northwood core is shiny black instead of dark blue

02-12-2002, 12:38 PM	#12
EMC2 Cooling Savant Join Date: Feb 2002 Location: Dark Side of the Moon Posts: 365	My dumb queries for the day How do you define idle and stressed? (what are you running to get to an idle state and what are you executing to put it under stress) As I recall almost all P4 motherboards use the internal CPU thermal diode to measure temps. Does yours? Or are these readings from something else? Are you sure that the processor isn't being throttled back by the system? (some P4 systems implement dynamic power management and will actually throttle back the processor depending upon various factors, in some cases including fan speed).

02-12-2002, 12:42 PM	#13
eX&TriC Cooling Neophyte Join Date: Oct 2001 Location: Belgium (center of the universe) Posts: 51	idle = W2K without doing anything stress = CPU stability test normal temp is measured by sensor IN the CPU

02-12-2002, 01:34 PM	#14
pHaestus Big Player Making Big Money Join Date: Aug 2001 Location: irc.lostgeek.com #procooling.com Posts: 4,782	I am not too familiar with the register settings for P4 mobos, but many AMD motherboards do not implement the CPU_HLT command that causes the processor to truly idle. This was done for supposedly compatibility and also to avoid huge swings in current demands when the CPU goes under load. You might want to look around for a WPCredit file that is for the i850 (that is what the TH7-II is right?) and see if maybe the CPU is idling high in temp for that reason. 46C under load seems high to me for a P4 anyway; aren't those pretty low heat CPUs?

02-12-2002, 02:57 PM	#15
niner Cooling Neophyte Join Date: Jan 2002 Location: Boston, MA Posts: 7	Sorry for being off-topic here. But can you tip me on how to remove the heat speader on a P4? I want to learn from someone who has done it before I ruin a $400 CPU. Thanks,

02-12-2002, 05:55 PM	#16
eX&TriC Cooling Neophyte Join Date: Oct 2001 Location: Belgium (center of the universe) Posts: 51	take a sharp exacto knife and put the blade out 2 cm start at the corner and put the blade against the corner 1 cm above and 1cm below the corner start moving the the knife making sure the blade is in the goop between the HS en CPU base. You can do this for the four corners and then slide the knife from one to the other. Don't put the knife in too far and make sure you don't toutch the die good luck

02-12-2002, 06:07 PM	#17
EMC2 Cooling Savant Join Date: Feb 2002 Location: Dark Side of the Moon Posts: 365	I don't know that I would consider W2K doing nothing necessarily as being at idle If you can find a software cooler for the P4, the Stop-Grant state of a P4 drops the power to almost 1/3 of normal. For the stress side, try Prime95, Folding, or SETI. (Intel also has a proggy called Maxpower that generates the worst case heat load possible, but I don't remember where they hide it). Note also that besides the auto-throttle that some motherboards have, the P4 also has an internal throttle circuit that reduces the duty cycle of the processor at a set temp. If you want to be sure this isn't going on, run SiSoft's CPU bench after a cold start and record the info. Then run a stress proggy for 30 mins, then re-run the CPU bench. If the numbers change significantly, then you are being throttled. pHaestus - yes, relatively low. A P4 at 2.2Ghz and default Vcore is about 55W. At the 1.65 Vcore and 2.4Ghz that eX&TriC said he was running that would go up to about 73W.

02-13-2002, 04:43 AM	#18
Brad Thermophile Join Date: Oct 2001 Location: Nuu Zeeelin Posts: 3,175	46C isn't too bad really at full load, I'd never say it's stunning, but nothing to be worried about. Clock throttling comes in at 65 - 75C depending on the model of P4. Also, I'm sure that win2k idles lower than win9x for example. In my laptop, I can use win2k for hours without the fan turning on, while in win98 the fan will turn on within 10minutes. If under full load the fan will turn on in about 5 minutes. __________________ 2x P3 1100's at 1400, Abit VP6, 2x Corsair 256mb PC150 sticks, 20gb 'cuda ATA-III, 2x 40gb 'cuda ATA-IV in raid 0. 20" Trinitron. No fans 2x 2400+ at 2288mhz (16.0 x 143), Iwill MPX2, 2x Kingmax PC-3200 256mb sticks, 4x 20gb 60gxp in Raid 5 on a Promise SX6000. Asus Ti4200 320/630. Cooled by Water

02-13-2002, 07:11 AM	#19
pHaestus Big Player Making Big Money Join Date: Aug 2001 Location: irc.lostgeek.com #procooling.com Posts: 4,782	For 73W heat, mid 40C is not a particularly good temp for water cooling. It is tolerable, but I would expect more in the neighborhood of 12-14C over room temperature (somewhere around 0.17-0.19 C/W) from messing with the internal diode on my XP.

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)