CPU v GPU temperatures

[MagR]

Hi All

I’m new to watercooling and have just built my first loop (with a lot of help from this forum). Now I have my temperature results I wanted to ask a couple of questions. My loop consists of dual gtx 570’s overclocked to 800mhz core clock and 4ghz effective memory clock and a Core i5 2500k overclocked to 4.5ghz at 1.23v. I am using a 4x140 full height radiator from Alphacool.

My GPU temps are amazing me. On air they could get up to 80C and the fans were very noisy. Using EK full cover blocks for graphics card watercooling Unigine Heaven gives GPU max temps of around 40C (varies slightly depending on ambient temps and fan speeds). This is far better than I expected as I was getting 40C at idle on air. I also get the same temperatures using the Metro 2033 benchmark on a continuous loop which I’ve always found very GPU intensive.

My CPU temperatures are great but the difference is not so pronounced. On air with the Be Quiet Dark Rock Advanced aftermarket cooler I was getting a max temp of 63C (some cores slightly lower) using Prime 95 small FFT’s (or blend for that matter). This always struck me as pretty good but I think I am lucky as I can get a perfectly stable 4.5ghz with vcore of only 1.23v which keeps the temperatures down. In my new watercooling loop with an XSPC Raystorm CPU block I’m getting a max temp of 51C using Prime.

Is this pattern normal? It seems watercooling is great for CPU’s but absolutely outstanding for GPU’s. Is this a reflection on how poor or inefficient air coolers on GPU’s compare to aftermarket air coolers on CPU’s? Would I notice a bigger difference between air and watercooling on my CPU if I went for a more adventurous overclock such as 4.8-5ghz? (I’m not going to – just interested). At the moment the CPU is positioned after my graphics cards in the loop. Would the CPU benefit much from being first in the loop before the graphics cards? – my graphics cards are in series and I notice a temperature difference between them of only 1C at most.

I’m very happy with the way things have panned out and am just trying to get a better understanding of how watercooling works.

Thanks in anticipation

Mag

 

[Deeks]

I am also newer to the water cooler community myself. i have built 2 water cooled rigs 1 i use myself and i have noticed the same differences you are talking about. I have noticed that if i push my 17 2600k to 4.8 in compared to say 4.3 my temp differences is 1c depending on ambient temp. I have found the water cooling pays off the most for extreme OC's in gpu and cpu. the higher the oc the bigger differences you will find from air to water delta T's. As you mentioned before i am not to sure why the GPU benefits so much more then the CPU i have often wondered that myself. As for the positioning in the loop i have always put CPU before GPU's but from what i read it wont make much of difference at all. I also am plumber by trade and when we install cooling system within buildings the difference between the first cooling panel and the last one in the loop isn't much different ( a couple degrees within a big loop )

 

[rubix_1011]

Is this pattern normal? It seems watercooling is great for CPU’s but absolutely outstanding for GPU’s. Is this a reflection on how poor or inefficient air coolers on GPU’s compare to aftermarket air coolers on CPU’s? Would I notice a bigger difference between air and watercooling on my CPU if I went for a more adventurous overclock such as 4.8-5ghz? (I’m not going to – just interested). At the moment the CPU is positioned after my graphics cards in the loop. Would the CPU benefit much from being first in the loop before the graphics cards? – my graphics cards are in series and I notice a temperature difference between them of only 1C at most.

You are referring to delta, specific heat and thermal conductivity of a watercooling loop. 40C load temps are very common; I personally see about the same temps, as do many others.

Loop order does not matter when it comes to temps.

 

[MagR]

I am also newer to the water cooler community myself. i have built 2 water cooled rigs 1 i use myself and i have noticed the same differences you are talking about. I have noticed that if i push my 17 2600k to 4.8 in compared to say 4.3 my temp differences is 1c depending on ambient temp. I have found the water cooling pays off the most for extreme OC's in gpu and cpu. the higher the oc the bigger differences you will find from air to water delta T's. As you mentioned before i am not to sure why the GPU benefits so much more then the CPU i have often wondered that myself. As for the positioning in the loop i have always put CPU before GPU's but from what i read it wont make much of difference at all. I also am plumber by trade and when we install cooling system within buildings the difference between the first cooling panel and the last one in the loop isn't much different ( a couple degrees within a big loop )

Thanks for your reply. It seems my results are perfectly normal then. I might experiment and see what happens if I push my CPU up to say 5ghz (it could do this on air but with a lot of voltage). For gaming I'm quite happy with it running at 4.5ghz but it might be interesting to test your hypothesis. I read a technical report on using separate loops for CPU and GPU with the same flow and rad area in total and it made no difference to temps really compared to running everything in a single serial loop. This supports your findings that position of components in loop does not matter much.

Thanks again for your reply - much appreciated

Regards

Mag

 

[MagR]

Is this pattern normal? It seems watercooling is great for CPU’s but absolutely outstanding for GPU’s. Is this a reflection on how poor or inefficient air coolers on GPU’s compare to aftermarket air coolers on CPU’s? Would I notice a bigger difference between air and watercooling on my CPU if I went for a more adventurous overclock such as 4.8-5ghz? (I’m not going to – just interested). At the moment the CPU is positioned after my graphics cards in the loop. Would the CPU benefit much from being first in the loop before the graphics cards? – my graphics cards are in series and I notice a temperature difference between them of only 1C at most.

You are referring to delta, specific heat and thermal conductivity of a watercooling loop. 40C load temps are very common; I personally see about the same temps, as do many others.

Loop order does not matter when it comes to temps.

Thanks for your reply. I'm starting to feel more comfortable with the technical side of watercooling. Thanks for helping me learn more.

Regards

Mag

 

[boiler1990]

I think the difference in temps has to do with a difference in packaging (the parts of the CPU/GPU that aren't circuitry). There's a decent amount of packaging material in CPUs (a lot of copper; you actually have to wear away a significant amount off the top to reach the die), and I am betting the temperature sensor in the CPU is below the copper somewhere in the die.

Therefore, you'll see a significant conduction effect across that copper layer, and conduction heat transfer is dependent on the temperature difference. In air cooling, the temperature difference starts large but gets smaller because the heatsink eventually begins to heat up above room temp. With watercooling, you get a more constant difference because the coolant temp remains relatively constant, so your temps will be better than air, but not humongous.

GPUs have significantly less packaging material on the die, so you're getting more of a convection effect (conduction still exists through the block), which will dissipate much more heat since it's dependent on the coolant temp, speed, etc.

TL;DR: Conduction is less effective than convection, and CPUs by design utilize more conduction effects, and therefore the temperature effects are not as drastic.

 

[dingo07]

I learn something new every day

Thanks for the lesson all!