High Idle temp but small Delta T

[alexwai]

I think it is a quite weird problem and I am not comfortable with it. The story is that I was using Barton 2500+ OC 2400Mhz @ 1.85V. I started to record the CPU idle temp after turning on the PC 30 mintues later. It was suprisingly high which was about 37.2C comparing with the ambient temp of 30.8C. However, after running 1 hour Prime95, the full load temp was 39C making the Delta T between the idle and full load quite small IMO. It seems that the high idle temp is completely due to the bad efficiency of my rig. WTF, it can get a small Delta T. So, are there something wrong on my setup or else? Any expects like yours can explain to me.

thanks


PS: all my temp measured through the external sensor which I stick it as close as possible to CPU die.

My setup:

Custom-made water block
Iwaki MD-20RZM water pump
Abit NF7-S v2.0
Thermochill HE-120.2 with the custom made shroud
1/2' ID tubing except the 3/4' on suction line with T.

[jaydee]

If we had the water temp we might be able to help. Kind of a critical temp for trouble shooting. Also are you sure your system is at a true idle?

(P.S. load of 9C over room temp is impossible with that setup)

[bigben2k]

Ambient temp of 30.8 would be typical in a mediterranean country, with no AC. Drop that measure, and use a real thermometer, to measure the *real* room temp, ok?

Your small delta T would be typical of a temp probe that's not setup properly. I'm just guessing here, but I believe that your temp probe isn't quite close enough to the CPU die.

[alexwai]

Originally Posted by jaydee116

If we had the water temp we might be able to help. Kind of a critical temp for trouble shooting.

I will check it tonight.

Also are you sure your system is at a true idle?

As I know, there are total 33 programs running at the background. I am not sure they can account for the high idle temp. In fact, according to the CPU resources consumed on the Task Bar menu, it is very small.

[alexwai]

Originally Posted by bigben2k

Your small delta T would be typical of a temp probe that's not setup properly. I'm just guessing here, but I believe that your temp probe isn't quite close enough to the CPU die.

Sorry for my bad quality pic. It si not my one, I use to explain my situation.

[Cathar]

My god. Can I grab some Arctic Silver from the system in that picture? I see enough for about 5 mounts there...

[alexwai]

Originally Posted by Cathar

My god. Can I grab some Arctic Silver from the system in that picture? I see enough for about 5 mounts there...

LOL, yes it is. Mine one is not the same though.

[bigben2k]

Boy, I sure hope that that's not your CPU: way too much thermal paste!

The temp probe has to touch the cpu die. In this case, if the mylar (yellow) extends past the probe, you will not get a reliable temperature measurement: you're measuring the ambient air at ~1mm (or more) away from the CPU die.

Ref thermocouples.

[alexwai]

Originally Posted by bigben2k

Boy, I sure hope that that's not your CPU: way too much thermal paste!

The temp probe has to touch the cpu die. In this case, if the mylar (yellow) extends past the probe, you will not get a reliable temperature measurement: you're measuring the ambient air at ~1mm (or more) away from the CPU die.

Ref thermocouples.

In this case, if I can put the temp probe just like you said, will it reach more higher CPU temp because of better contact between the CPU die and temp probe? Let say 12C over the ambient temp.

[HAL-9000]

Originally Posted by bigben2k

Ambient temp of 30.8 would be typical in a mediterranean country, with no AC. Drop that measure, and use a real thermometer, to measure the *real* room temp, ok?

Your small delta T would be typical of a temp probe that's not setup properly. I'm just guessing here, but I believe that your temp probe isn't quite close enough to the CPU die.

Thirty celsius I think could be considered a probable ambient temperature for Hong Kong in June. Quite humid there as well, very sticky! Only been there twice, but it is a muggy place.

[alexwai]

Originally Posted by HAL-9000

Thirty celsius I think could be considered a probable ambient temperature for Hong Kong in June. Quite humid there as well, very sticky! Only been there twice, but it is a muggy place.

Yes, it is the case. very humid and hot. The average daytime temp is around 30 C.

[HAL-9000]

I think there is a bit too much goo on the chip. Looks like there is thick goop on the sides as well, which doesn't do a whole lot of good!

The temp probe where you mounted it is reporting the temperature of the silicon package more than the die I think, and since silicon is such a high heat capacity material, the small DT between idle and max ops seems reasonable.

What I do when I apply thermal grease (I use Arctic Silver) is to just put a small drop on the die. Then I use a piece of something that is very thin and chemically inert, like packaging cardboard or the like, and "spread" the drop over the die. Its kind of analagous to taping sheetrock, but on a small scale. You end up with a die that has a opaque grey coating, but you can still pretty much see the die through the goo. Its not totally obscured like in your sample pic.

Try that, and I bet your temps will adjust to "normal" measurements. Also read Phaestus's article on test measurements. Your insturments could be reporting at either end of their sampling error, which could add up to a experimental deviation of some four or five degrees C.

[freeloadingbum]

Which side is the cache on? Maybe the cache doesn't produce as much heat as the rest of the cpu.

[alexwai]

Originally Posted by HAL-9000

I think there is a bit too much goo on the chip. Looks like there is thick goop on the sides as well, which doesn't do a whole lot of good!

The temp probe where you mounted it is reporting the temperature of the silicon package more than the die I think, and since silicon is such a high heat capacity material, the small DT between idle and max ops seems reasonable.

What I do when I apply thermal grease (I use Arctic Silver) is to just put a small drop on the die. Then I use a piece of something that is very thin and chemically inert, like packaging cardboard or the like, and "spread" the drop over the die. Its kind of analagous to taping sheetrock, but on a small scale. You end up with a die that has a opaque grey coating, but you can still pretty much see the die through the goo. Its not totally obscured like in your sample pic.

Thanks for you reply. Infact, it is only the sample pic which is not mine one. For applying the thermal paste, I did exactly what you have mentioned.

Try that, and I bet your temps will adjust to "normal" measurements. Also read Phaestus's article on test measurements. Your insturments could be reporting at either end of their sampling error, which could add up to a experimental deviation of some four or five degrees C.

I agree that I may get a bad mount of my waterblock. Before re-mounting the WB, I guess I may check the water temp as suggested by Jaydee116 to see whether it is normal or not. Do you think it is logical?

[HAL-9000]

Originally Posted by alexwai

Thanks for you reply. Infact, it is only the sample pic which is not mine one. For applying the thermal paste, I did exactly what you have mentioned.


I agree that I may get a bad mount of my waterblock. Before re-mounting the WB, I guess I may check the water temp as suggested by Jaydee116 to see whether it is normal or not. Do you think it is logical?

You for sure want the water in temp. That is the actual ambient of your block, because that is the active fluid cooling your setup.

Also, OK...that isn't a pic of your CPU on this thread. But, is the temp probe in the photo representative of the same place where you have mounted your temp probe? Because if it is, you are getting a better measurement of the ceramic package silicon's temp than the CPU die. Silicon has a high specific heat. If you want to find out yourself, since you live in HK, head down to New Territories and get yourself some sand off the beach. Then, put a blowtorch to it, and watch how long it takes to heat it up and cool down when the blowtorch is removed. I think that is a big factor in your high initial temp/low delta issue, at least if that is the way your thermal probe is mounted.

[alexwai]

Originally Posted by HAL-9000

Also, OK...that isn't a pic of your CPU on this thread. But, is the temp probe in the photo representative of the same place where you have mounted your temp probe? Because if it is, you are getting a better measurement of the ceramic package silicon's temp than the CPU die.

Yes, same place I put my temp probe on. The only difference is no thermal paste sticking on the probe.

[Jabo]

Gnerally, if all is installed as it should be, dT is low and temps are high it means thatheat dissipation part of your setup is not up to the job. Change radiator or install better fans (producing higher back pressure in case of restricitve rad design or more flow in the opposite case).
Did you post specs for your fans and how you insalled them and where did you palce your rad/fans assy?

[bigben2k]

Originally Posted by alexwai

In this case, if I can put the temp probe just like you said, will it reach more higher CPU temp because of better contact between the CPU die and temp probe? Let say 12C over the ambient temp.

Yeah, the temps will be higher, as they should be.

The temp probe you're using is meant to take measurements from the side, not the tip.

[Ruiner]

Some AMD chipsets don't allow true idle by default, even when cpu utilization reads 0% in systray. The newest bios for my shuttle has a C1 disconnect, IIRC, that, when forced on, drops idle temps a degree or two. I believe the correct register in H.Oda's Wpcredit's utility will do the same.

[alexwai]

Just run a test using the previously setting (Prime 95 1 hour), Room temp is still on 30~ C. I put the thermal probe into the T line, the water temp is 36C. WTF, too high, right! As my understanding that the water temp will not be that high in similar setup. What does it imply? Well, I have 2 conclusion.

1) Will the pump putting too much heat into the water?
2) Just like Jabo said, the radiator is not up to the job.

I am not too sure these conclusion are right or not. I think it is a better idea to provide as much as information so that all of your expects can help me to figure it out what is going on?

Re Jabo

I place my radiator on the top of my case. Certainly, helped by my friend, I opened 2 x 12cm fans holes using the jigsaw. What fans i am using is 2 Delta EFB1212SHE-F00 (during the test, the fans is at 12V) which can provide 141.96CFM, 13.2mmH2O . IMO, they are quite decent and able to do the job well. I put them in between the radiator and the top of the case so that the cooler air can be pushed from the area outside the case through the radiator , then to the case.

That is all I can tell ATM.

[starbuck3733t]

Now that you have no restrictions on your intake and exhaust of the radiator (you cut holes in the top and are running w\o side panel... right?) it should have lowevered itself. Have you examined the fins of the radiator itself for debris? How about flow rate?

[alexwai]

Originally Posted by starbuck3733t

Now that you have no restrictions on your intake and exhaust of the radiator (you cut holes in the top and are running w\o side panel... right?) it should have lowevered itself. Have you examined the fins of the radiator itself for debris? How about flow rate?

The fin is quite clean as I regularly use the vacuum cleaner to clean out the dirt or debris on weekly basic.

Regarding to the flow rate, I don't have flow meter. but, you can imagine that it will not be too small in normal one CPU block setting using Iwaki MD-20RZM.

Here is my pic.

[freeloadingbum]

Originally Posted by alexwai

Just run a test using the previously setting (Prime 95 1 hour), Room temp is still on 30~ C. I put the thermal probe into the T line, the water temp is 37C. WTF, too high, right! As my understanding that the water temp will not be that high in similar setup.

Try using the same sensor (if possible) to measure the rad fan intake temp. Your water temp should be around 3-4C above fan intake temp with that rad and fan setup (considering sensor error)

[alexwai]

Originally Posted by freeloadingbum

Try using the same sensor (if possible) to measure the rad fan intake temp. Your water temp should be around 3-4C above fan intake temp with that rad and fan setup (considering sensor error)

Rad fan intake temp? It is the same as the ambient temp I mentioned before. What you mean is within the custom made shroud?

Actually, If I place the heat sensor within the shroud, the temp is about 3C higher than the ambient temp.

[Ruiner]

Correct. Ambient should be nearly the same as 'rad intake temp' given your design. In your rig, air temp in the shroud has been heated by the rad.

Is your setup "pump--->radiator------>block----->pump" ?

The cascade likes being 1st after the pump, being a 'jet impingement' design.

Speaking of that, your setup has like a potential case airflow problem. I assume your case fans are all set to exhaust, such that air flow will be the opposite of heat convection in the case.