Help! Cooling mystery

Graystar · 04-08-2003, 06:18 PM

Take a look at these pics...

The fan in the second picture is nothing special. Just an ordinary 80mm fan. You can't even hear it when it's running.

When I have the fan, the temperature of the two blocks is about 40/42 (I know for a fact that I messed up the thermal paste on the second processor, hence the temp difference. Gotta fix that...)

When I remove the fan, the temperature rises to 48/51!

That's a HUGE increase in the temperature reading.

Also, the water temperature remains the same during all of this.

Here's what you're looking at:
Tyan Tiger MPX S2466N-4M motherboard
AMD 2200+ model 6.8.0. X2
Antec 1000SX case
Enermax 431w PS

Has anyone experienced anything like this before? Have any of you guys turned all your fans off in your case to see what happens to the CPU temps?

Axle · 04-08-2003, 07:23 PM

You're the guy with the "Whatblock?" block, correct? My best guess would be the water can't remove the heat fast enough, so some dissipates right into the block. I'd try a slighty more powerfull pump and see if that makes a difference.

On the other hand, I'm probably wrong...

Alchemy · 04-08-2003, 07:35 PM

Graystar, figure out how hot different parts of your block are. My guess is that either the top of the block is really hot or the sides of your CPU dye are exposed so that more circulation cools it down.

I'd take this as something to be concerned about if your barbs are hot - the tubing will fail if it gets too warm.

Alchemy

Graystar · 04-08-2003, 07:39 PM

Quote:

Originally posted by Axle
On the other hand, I'm probably wrong...

LOL!! Who isn't?

Yes, I made the WhatBlock? I forgot to mention that I'm using a Danner Supreme Mag-Drive model 2. It's about the same capacity as a Eheim 1250, so it's got a good bit of power. The rad is an Everco Heater core for a Pinto. It 's 7 3/4" X 6" X 2". Finally, I actually use a Vornado 510 fan to blow air through the heater core.

Seems strange...

pHaestus · 04-08-2003, 07:43 PM

I mentioned to you before that you think your motherboard is reading from the diode but it isnt. There are surface mount thermistors inside the socket, and when you point a fan at the area you are cooling the PCB and the socket area, moving air around, and getting lower temps. I could make the "CPU" temp of my A7V-133 (also a surface mount resistor) fluctuate with a fan as well. The temp never changed with a homemade diode reader however.

TerraMex · 04-08-2003, 07:47 PM

Where are you reading those temps? Insocket sensor? If so, then i've read that the air movement created by the fan is refreshing the air near the sensor, and thus giving a lower temp . With no air movement whatsoever the sensor reads the temperature of the hot air trapped under the cpu.

... And you're cooling the pin organic bed (or whatever that's called) with the fan. Or...er...hmmm.

Graystar · 04-08-2003, 09:29 PM

Quote:

Originally posted by Alchemy
Graystar, figure out how hot different parts of your block are. My guess is that either the top of the block is really hot or the sides of your CPU dye are exposed so that more circulation cools it down.

I'd take this as something to be concerned about if your barbs are hot - the tubing will fail if it gets too warm.

Alchemy

When I touch the blocks I can't really discern any temperature difference. I tried to use the sensor on my thermometer but I couldn't get a good enough contact with the block.

Since87 · 04-08-2003, 10:13 PM

Quote:

Originally posted by Graystar
When I touch the blocks I can't really discern any temperature difference. I tried to use the sensor on my thermometer but I couldn't get a good enough contact with the block.

A glob of thermal paste on the probe helps immensely, but it would leave you with a messy waterblock.

Graystar · 04-08-2003, 10:22 PM

Quote:

Originally posted by Since87
A glob of thermal paste on the probe helps immensely, but it would leave you with a messy waterblock.

Yeah, I tried that but my blocks are sooooo small that I would get paste all over the top of the CPU.

UnloadeD · 04-09-2003, 12:20 AM

pH hit the nail on the head. Also points out how hard it really is to compare a HSF to watercooling even in the same computer. I always suspect HSF's are showing better temps than whats really happening. If the fan is quiet and you don't mind the looks, you should leave it. If nothing else it will help cool the mosfets and such, I doubt your cpu's are any cooler though.

peace.
unloaded

Graystar · 04-09-2003, 12:51 AM

Quote:

Originally posted by UnloadeD
pH hit the nail on the head. Also points out how hard it really is to compare a HSF to watercooling even in the same computer. I always suspect HSF's are showing better temps than whats really happening. If the fan is quiet and you don't mind the looks, you should leave it. If nothing else it will help cool the mosfets and such, I doubt your cpu's are any cooler though.

Well, it's always been the case that when I replaced a HSF with a waterblock on a single processor that the temperature of that processor when down. This was also the case on another case-less motherboard that showed a temperature drop. In this case, the HSF was the only fan. Once removed, there was no forced air movement around the motherboard.

It was only when I replace both HSF's with waterblocks and removed all the fans did this issue show itself. I've always known that my readings were only good relative to this board. However, I didn't realize just how sensitive the reading was to other conditions in the case. I can still perform valid comparitive measures...I just need to be extra careful about the conditions.

BillA · 04-09-2003, 01:07 AM

Graystar

as the geek said (and that SHOULD hurt BAD), you have a lot to learn

The Overclocker's Mantra:

all temps are crap, but Watts are worse
oooohhhhhhmmmmmm

but no, you never heard of secondary path losses
try moving the fans around
christ you're dense
people are posting and you don't catch a bit of it

Les · 04-09-2003, 10:45 AM

Quote:

Originally posted by pHaestus
The temp never changed with a homemade diode reader however.

Would expect a small effect.
Maybe* this type of behaviour for a 100watt** CPU with a very good waterblock( "C/W"= Rwb= 0.2c/w) :-
Cwb= Wt*Rwb - (Rwb/Rmb)*Cmb

* From this arithmetic :-
Wt=Wwb + Wmb, Rmb = Cmb/Wmb, Rwb= Cwb/Wwb
Where Wt=Total Heat output, Wwb=Heat Dissipated Thro wb, Wmb=Heat Dissipated Thro Secondary Paths(probably motherboard(mb)), Rwb=Thermal Resistance of wb(inc TIM), Rmb= Thermal Resistance of Secondary Paths(probably mb), Cmb= Temp Difference between Die and Secondary Heatsink, and Cwb= Temp Difference between Die and Water..

Clumsy maninipulation and substitution[Wt/Cwb = Wwb/Cwb + Wmb/Cwb , Wt/Cwb = Wwb/Cwb + Cmb/Rmb x 1/Cwb = 1/Rwb + Cmb/Cwb x 1/Rmb, Cwb = Wt/(Wwb/Cwb + Cmb/Rmb x Cwb, Cwb= Wt/(1/Rwb + Cmb/Cwb(1/Rmb), Cwb/Rwb + Cmb/Rmb = Wt ] gives:-
Cwb= Wt*Rwb - (Rwb/Rmb)*Cmb

** I am still open minded the apparent discrepancy between CPU heat calculations and reality. Do not think it is anywhere near established how much is due to "crap sums and crap CPU stress programs" and how much is due to secondary heat-paths.

Graystar · 04-09-2003, 03:43 PM

So I need make a diode reader then. Where can I get the Maxim IC? My usual source for such things (Jameco) doesn't seem to have it.

JSimmons · 04-09-2003, 04:41 PM

Graystar:

Are you running 100% lucite water blocks? I can see the whole CPU under them.

bigben2k · 04-09-2003, 04:49 PM

Quote:

Originally posted by JSimmons
Graystar:

Are you running 100% lucite water blocks? I can see the whole CPU under them.

No, he's got a solid copper block, it's just very tiny!

Graystar · 04-09-2003, 05:02 PM

Quote:

Originally posted by JSimmons
Are you running 100% lucite water blocks? I can see the whole CPU under them.

LOL!!!

Yes.....yes I am

Noooo...just joking. It's tiny...but it's copper

JSimmons · 04-09-2003, 05:58 PM

Why so small? Wouldn't larger blocks cool better?

Graystar · 04-09-2003, 06:01 PM

Quote:

Originally posted by JSimmons

Why so small? Wouldn't larger blocks cool better?

Well, one of the goals was to find that out. Also, the block was fairly easy to make and very inexpensive. I use about $1.50 of copper. I also have eight processors to cool and the idea was to make blocks myself to save money.

Althornin · 04-09-2003, 06:48 PM

Quote:

Originally posted by Graystar
So I need make a diode reader then. Where can I get the Maxim IC? My usual source for such things (Jameco) doesn't seem to have it.

actually, just ask Maxim to send you one.
They usually will

Arcturius · 04-09-2003, 08:29 PM

Actually, if the Tiger MP you're using is similar enough to the Thunder K7 that preceded it, I wouldn't trust the onboard monitors very much at all.
It seems Tyan did a pretty poor job with the layout of sensors (close counts, apparently), and the hardware monitoring aspects of the board in general--I've basically given up on the ones we run at work.

Essentially what I'm saying, is that the fan is probably either blowing on the thermistor, or is doing a good job of cooling the PCB right next to the thermistor. I found a diagram of the 2462 once upon a time that showed where the thermistors are, but I've lost it. If you want accurate temps, you're going to have to do some soldering...

Check this out if you want to see the details of what I'm talking about. Check out the referenced tickets, as well.

04-08-2003, 06:18 PM	#1
Graystar Registered User Join Date: Mar 2003 Location: Brooklyn, NY Posts: 112	Help! Cooling mystery Take a look at these pics... The fan in the second picture is nothing special. Just an ordinary 80mm fan. You can't even hear it when it's running. When I have the fan, the temperature of the two blocks is about 40/42 (I know for a fact that I messed up the thermal paste on the second processor, hence the temp difference. Gotta fix that...) When I remove the fan, the temperature rises to 48/51! That's a HUGE increase in the temperature reading. Also, the water temperature remains the same during all of this. Here's what you're looking at: Tyan Tiger MPX S2466N-4M motherboard AMD 2200+ model 6.8.0. X2 Antec 1000SX case Enermax 431w PS Has anyone experienced anything like this before? Have any of you guys turned all your fans off in your case to see what happens to the CPU temps?

04-08-2003, 07:23 PM	#2
Axle Cooling Savant Join Date: Aug 2002 Location: Wind Gap, PA Posts: 112	You're the guy with the "Whatblock?" block, correct? My best guess would be the water can't remove the heat fast enough, so some dissipates right into the block. I'd try a slighty more powerfull pump and see if that makes a difference. On the other hand, I'm probably wrong... __________________ "Once we've got those bugs flattened out, we'll be running on flat bugs"

04-08-2003, 07:47 PM	#6
TerraMex Cooling Savant Join Date: Nov 2002 Location: Portugal, Europe Posts: 870	my 2 cents. Where are you reading those temps? Insocket sensor? If so, then i've read that the air movement created by the fan is refreshing the air near the sensor, and thus giving a lower temp . With no air movement whatsoever the sensor reads the temperature of the hot air trapped under the cpu. ... And you're cooling the pin organic bed (or whatever that's called) with the fan. Or...er...hmmm. __________________ "we need more cowbell."

04-08-2003, 07:35 PM	#3
Alchemy Cooling Savant Join Date: Oct 2002 Location: Boston Posts: 238	Graystar, figure out how hot different parts of your block are. My guess is that either the top of the block is really hot or the sides of your CPU dye are exposed so that more circulation cools it down. I'd take this as something to be concerned about if your barbs are hot - the tubing will fail if it gets too warm. Alchemy

04-08-2003, 07:43 PM	#5
pHaestus Big Player Making Big Money Join Date: Aug 2001 Location: irc.lostgeek.com #procooling.com Posts: 4,782	I mentioned to you before that you think your motherboard is reading from the diode but it isnt. There are surface mount thermistors inside the socket, and when you point a fan at the area you are cooling the PCB and the socket area, moving air around, and getting lower temps. I could make the "CPU" temp of my A7V-133 (also a surface mount resistor) fluctuate with a fan as well. The temp never changed with a homemade diode reader however.

04-09-2003, 12:20 AM	#10
UnloadeD Cooling Savant Join Date: Oct 2002 Location: MidWest USA Posts: 176	pH hit the nail on the head. Also points out how hard it really is to compare a HSF to watercooling even in the same computer. I always suspect HSF's are showing better temps than whats really happening. If the fan is quiet and you don't mind the looks, you should leave it. If nothing else it will help cool the mosfets and such, I doubt your cpu's are any cooler though. peace. unloaded

04-09-2003, 01:07 AM	#12
BillA CoolingWorks Tech Guy Formerly "Unregistered" Join Date: Dec 2000 Location: Posts: 2,371.493,106 Posts: 4,440	Graystar as the geek said (and that SHOULD hurt BAD), you have a lot to learn The Overclocker's Mantra: all temps are crap, but Watts are worse oooohhhhhhmmmmmm but no, you never heard of secondary path losses try moving the fans around christ you're dense people are posting and you don't catch a bit of it

04-09-2003, 03:43 PM	#14
Graystar Registered User Join Date: Mar 2003 Location: Brooklyn, NY Posts: 112	So I need make a diode reader then. Where can I get the Maxim IC? My usual source for such things (Jameco) doesn't seem to have it.

04-09-2003, 04:41 PM	#15
JSimmons Cooling Savant Join Date: Apr 2003 Location: 4-sided room with an exit going east, and an exit going south Posts: 392	Graystar: Are you running 100% lucite water blocks? I can see the whole CPU under them.

04-09-2003, 05:58 PM	#18
JSimmons Cooling Savant Join Date: Apr 2003 Location: 4-sided room with an exit going east, and an exit going south Posts: 392	Why so small? Wouldn't larger blocks cool better?

04-09-2003, 08:29 PM	#21
Arcturius Cooling Savant Join Date: Aug 2001 Location: Chippewa Falls, Wisconsin Posts: 294	Actually, if the Tiger MP you're using is similar enough to the Thunder K7 that preceded it, I wouldn't trust the onboard monitors very much at all. It seems Tyan did a pretty poor job with the layout of sensors (close counts, apparently), and the hardware monitoring aspects of the board in general--I've basically given up on the ones we run at work. Essentially what I'm saying, is that the fan is probably either blowing on the thermistor, or is doing a good job of cooling the PCB right next to the thermistor. I found a diagram of the 2462 once upon a time that showed where the thermistors are, but I've lost it. If you want accurate temps, you're going to have to do some soldering... Check this out if you want to see the details of what I'm talking about. Check out the referenced tickets, as well.

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