View Full Version : CPU+GPU same loop, 10C difference
Spoobs
01-11-2010, 07:23 AM
i got a 4870 with a MCW60 waterclock in the same loop as a e6600 with a DTEK fuzion installed.
When checking the temps the GPU (33C) was idling about 10C cooler than that of my processor, and the processor core temps jumped alot between 38C-48C. but primarily stick around 44C
are these symptoms of a bad installation of a cpu waterblock? perhaps too much/bad thermal paste?
also, should i look into lapping the cpu for optimal heat conductivity?
Hondacity
01-11-2010, 07:31 AM
thermocouples or sensors from different manufacturers are DIFFERENT
Spoobs
01-11-2010, 08:05 AM
thermocouples or sensors from different manufacturers are DIFFERENT
as we all would have suspected....... thanks for the Billy Mays version of the word "different"
however 10C difference seems a bit of a gap to boil down to manufacture differences unless one is actually faulty or made from crapco. however considering both products are produced by reputable companies id think to think those senors are fairly accurate to about 1C-3C.
before the Dtek was installed the core temps never used to jump as rapid as they do, i just want to know if this is normal in water cooling for jumpy readings and such a big gap in between two items being cooled in the same loop.
Your GPU and CPU are totally different components, having each their own TDP and surface dissipation area etc. so why would you expect any close temperatures between them ? A hotter component will be hotter... Doesnt matter if they are in the same loop or not, same loop means having same water flow around your components, and nothing else.
Alexandr0s
01-11-2010, 08:37 AM
Yeah, your GPU produces much less heat than your CPU, hence it's less hot.
Conumdrum
01-11-2010, 08:38 AM
Your GPU and CPU are totally different components, having each their own TDP and surface dissipation area etc. so why would you expect any close temperatures between them ? A hotter component will be hotter... Doesnt matter if they are in the same loop or not, same loop means having same water flow around your components, and nothing else.
+1^:up:
Hondacity
01-11-2010, 08:38 AM
cpu temps
1. tcore
2. tcase
...again different thermocouples...and different locations...
gpu temp...is that tcase? tcore?
Spoobs
01-11-2010, 08:43 AM
Your GPU and CPU are totally different components, having each their own TDP and surface dissipation area etc. so why would you expect any close temperatures between them ? A hotter component will be hotter... Doesnt matter if they are in the same loop or not, same loop means having same water flow around your components, and nothing else.
this is my first loop, i would have expected them to be within 5C of each other because like you said they have the same water flowing through the loop. Since the water would have somewhat the same temp all around the loop, my hypothesis was that both components would have roughly the same temp as the water being carried to the heat conductors minus losses in imperfections and error.
so it appears my hypothesis is wrong and i thank you for the input.
this still leaves one question unanswered, when i had a tuniq tower installed the core temps never fluctuated like they do now, is a 10C rapid jump within a second or two normal for new installations of water blocks?
Conumdrum
01-11-2010, 08:47 AM
Sure, a 10C jump from idle to load is common. You could try a remount. What paste? Not all pastes need curing time.
Nice to know whats actually in your loop besides what you told us. Make a Sig? I see x x, linkie is broken I guess.
Spoobs
01-11-2010, 08:58 AM
Sure, a 10C jump from idle to load is common. You could try a remount. What paste? Not all pastes need curing time.
Nice to know whats actually in your loop besides what you told us. Make a Sig? I see x x, linkie is broken I guess.
updated sig
I have not applied to load yet, these fluctuations are at desktop with minimal applications running.. if you dont count firefox. thats my mistake for not mentioning that.
@Hondacity : Ill try and get the ambients asap, at work ATM and just comparing core temps i took last night atm
try both
1. pump>GPU>CPU>rad
2. pump>CPU>GPU>rad
do things that you do most to torture your system => find out which of above ways is best
this is my first loop, i would have expected them to be within 5C of each other because like you said they have the same water flowing through the loop. Since the water would have somewhat the same temp all around the loop, my hypothesis was that both components would have roughly the same temp as the water being carried to the heat conductors minus losses in imperfections and error.
You are confusing heat transfer with temperatures.
Lets imagine you have the same waterblocks and same surface contact on 2 different chips in your loop (lets say, 2 different GPUs for example). But one GPU has a TDP of 150W, the other has a TDP of 50W. Obviously the 150W one will be a lot hotter... Here you have 2 different chips, with 2 very different waterblocks which have very different heat transfer capacities, and even 2 different set of sensors...
prava
01-11-2010, 09:44 AM
Since the water would have somewhat the same temp all around the loop, my hypothesis was that both components would have roughly the same temp as the water being carried to the heat conductors minus losses in imperfections and error.
so it appears my hypothesis is wrong and i thank you for the input.
Your hypothesis failed because you thought about temperature and not heat. Water dissipates heat per surface unit (then water velocity and block design takes into place, of course). Now, think how many parts you are dissipating in a VGA and how many parts you are dissipating in a CPU: yea, a CPU may have less power consumption but its cooling surface is rather small compared to a VGA which is why this happens.
Why do you see so many temperatura variations? Well, thats because the water temperature is more or less constant and the temperature of your heatsink is not, touch it while idle and touch it again while you've been stressing the cpu for long: the HS gets hotter so gets worse to disipate the heat. Thats why you see less variation with a HS and a lot more with water.
Hondacity
01-11-2010, 02:18 PM
You are confusing heat transfer with temperatures.
Lets imagine you have the same waterblocks and same surface contact on 2 different chips in your loop (lets say, 2 different GPUs for example). But one GPU has a TDP of 150W, the other has a TDP of 50W. Obviously the 150W one will be a lot hotter... Here you have 2 different chips, with 2 very different waterblocks which have very different heat transfer capacities, and even 2 different set of sensors...
you can talk about tdp all day long....but as quoted below..the surface area will change how much heat is dissipated/transferred...
Your hypothesis failed because you thought about temperature and not heat. Water dissipates heat per surface unit (then water velocity and block design takes into place, of course). Now, think how many parts you are dissipating in a VGA and how many parts you are dissipating in a CPU: yea, a CPU may have less power consumption but its cooling surface is rather small compared to a VGA which is why this happens.
Why do you see so many temperatura variations? Well, thats because the water temperature is more or less constant and the temperature of your heatsink is not, touch it while idle and touch it again while you've been stressing the cpu for long: the HS gets hotter so gets worse to disipate the heat. Thats why you see less variation with a HS and a lot more with water.
temperature and heat .....they do have a relationship...thus - not fail
NaeKuh
01-11-2010, 02:54 PM
i got a 4870 with a MCW60 waterclock in the same loop as a e6600 with a DTEK fuzion installed.
When checking the temps the GPU (33C) was idling about 10C cooler than that of my processor, and the processor core temps jumped alot between 38C-48C. but primarily stick around 44C
are these symptoms of a bad installation of a cpu waterblock? perhaps too much/bad thermal paste?
also, should i look into lapping the cpu for optimal heat conductivity?
No its a failure to understand how components work and how heat is distributed.
Because you have the the same water temp going though your loop, does not mean the gpu and cpu will heat up the same accordingly.
If your cpu is overclocked and is running hotter then your gpu, then you will have a spread that large.
Why do you see so many temperatura variations? Well, thats because the water temperature is more or less constant and the temperature of your heatsink is not, touch it while idle and touch it again while you've been stressing the cpu for long: the HS gets hotter so gets worse to disipate the heat. Thats why you see less variation with a HS and a lot more with water.
Because nothing but FULL LOAD will dish out a constant heat load.
This is why WE NEVER LOOK AT IDLE PAST MOUNT!!
Bad idle = BAD MOUNT
FULL LOAD = PEAK OUTPUT PERFORMANCE
Your thinking is flawed.
Also wtf do you mean less variation?
My CPU loop on Nadeshiko has VERY VERY VERY VERY LITTLE variation.
Why? because im running an overpowered cooling loop on the cpu only.
Your looking at 2 different heat sources, holding the same ambient.
The only way you can get a spread of less then 5C between GPU and CPU is if you can mirror the heat load off each to be within 5C.
The only way you can mirror is by dropping your CPU overclock so its matches on the fly.
The OP's thinking is just WRONG.
He's assuming his gpu's are within 5C, by soley looking at TDP values.
Well YOU THROW TDP OUT THE DOOR WHEN YOU OC.
you can talk about tdp all day long....but as quoted below..the surface area will change how much heat is dissipated/transferred...
That's why i reduced the hypothesis to the same surface area of transfer, if you read my post closely ;) Of course that's a futile excercise since a GPU and CPU have a very different thermal profile, surface area, and even waterblocks that go on them are completely different.