Delta temperature between CPU and H100i

[Brian Stirling]

I just completed a new PC build for video editing using an Asus x99-PRO/USB3.1 motherboard, i7-5820K CPU, Corsair H100i GTX CPU water cooler, EVGA GeForce GTX 980 TI Hybrid Gamer water cooled GPU and 32GB of G.Skill Ripjaws V series RAM in a 4x8GB DDR4 3200 configuration with 14-14-14-34 timing.

OK, I haven't done much to up the clock at this point and I just did a test render from Premiere Pro of a 21 minute edit and noticed something I have to ask about. During render the CPU temps got up to about 56C but the H100i GTX CPU cooler only reported about 36C or about 20C less than the CPU. Does that delta sound about right or should they be closer in temperature. If they should be closer in temp would that imply the cooler block is either not seated flush with the CPU or that the thermal paste is maybe not doing what it should?

In summary, what difference in temp should one expect under load between the CPU temps and the water temp? I should mention that all temps are as reported by Corsair Link. Corsair Link reports 5 temps, 2 fans speeds and the VCPU from the motherboard; 6 core and 1 package temp from the CPU; the temps of both WD 6TB HD's; the temp and fan speed of the GTX 980Ti; and lastly the temp and fan and pump speeds of the H100i GTX CPU cooler.

I noted during the render that in addition to the 56C temps of the CPU there were temps on the MB that were also about 56C while the H100i GTX was reporting about 36C or 20C cooler. As an aside, the GPU temp never got more than 30C but then the test render didn't have any effects added so the GPU would not have played much of a role.


Brian

 

[Carfax83]

What sensor or method does Corsair link even use? I doubt it's accurate at all. Good software temp monitors like RealTemp read directly from the CPU's tjmax and then infer the temperature based on distance from what the DTS is reporting.

If you Google Corsair link, you'll see that a lot of people question it's accuracy.

On a side note, are you really running your RAM at DDR4 3200 with such aggressive timings? I've never seen timings that low for DDR4 3200, especially with a 32GB configuration.

 

[Brian Stirling]

What sensor or method does Corsair link even use? I doubt it's accurate at all. Good software temp monitors like RealTemp read directly from the CPU's tjmax and then infer the temperature based on distance from what the DTS is reporting.

If you Google Corsair link, you'll see that a lot of people question it's accuracy.

On a side note, are you really running your RAM at DDR4 3200 with such aggressive timings? I've never seen timings that low for DDR4 3200, especially with a 32GB configuration.

No, the system is largely unchanged from the default settings and the RAM is definitely not running that aggressively -- I believe it's currently running about 2100 versus the 3200 its billed as being able to provide.

http://www.newegg.com/Product/Product.aspx?Item=N82E16820232219

Yeah, I'm not sure how accurate Corsair Link is, but I find it hard to believe that if they have a sensor it's not being read with reasonable accuracy. It's not clear exactly what temp is being monitored but my guess is they are measuring the water temp as that would be the most meaningful thing to measure.


Brian

 

[know of fence]

The lower the temperature in your cooling system, the more effective it is. Try running a sustained stress test, giving the water time to heat up, to see if you eventually reach an equilibrium temperature for your coolant, but 36° C may be as high as it gets.
CPU temperatures are mostly determined by load, voltage and frequency, elaborate cooling can only shave off a couple degrees, which makes AIO-coolers basically a waste of money and electricity, though not necessarily a giant waste. You have to push the diminishing returns of your OC and to the very edge to see any benefit.

 

[guskline]

What was the ambient temperature of your room? I suspect the H100i temp you are referring to is the temp of the coolant in the radiator and NOT the temp of the cpu.

In custom water cooling I measure this way. I use Aida64 (the section under stress test which shows min/max and average and measure temps of all of the cores, mb, vrm etc). I also have a temp sensor in my distilled water cooling loop which shows the true temp in the loop and finally a separate external temperature guage that shows the ambient room temp in Celsius (all other temps use Celsius vs Farenheit).

The Delta you need to measure is the difference between the loop temp and the ambient room temp to see how well your cooler is doing.

The 36C temp you reported for the H100i was probably the stress test temp of the liquid in the radiator at it's hottest point. You need to compare that temp to the ambient temperature both when the computer is near idle and at a point during stress testing to see the true delta in temp.

I also have the heat of the gpu dumping into my loop (both of my rigs below are custom water cooled).

For a stock 5820K, not a bad temp during stress. In my 5960x rig (8core/16 thread) which is OC'd to 4.4Ghz with vcore set @1.35 my highest temps on the cores during stress are in the mid 60s but the delta I refer to above between the idle state and OC'd state is at most 6C ( a little higher if I stress the gpu also).

I'm sure the H100i is working well for you and even Oc'd somewhat will be a good investment.

I bought my ambient room temp device online for @$6. Has a nice big digital readout. Get one of those and then see how the H100i holds up.

 

[Brian Stirling]

As I type this the PC is on (I'm using it to type) and the temps are:

CPU cores average about 26C with CPU package at 35C.

Most motherboard temps are about 27C with a high of 36C.

The 980 Ti is at 23C while the H100i is at 27C.

Both 6TB hard drives are reporting 20C and 21C. but I've not accessed them so they are likely idle.

Room temp is about 20C.

So, under very low loading as it is now the difference in temp between the CPU and H100i is ... well the H100i is about the same as the CPU core temps but about 8C cooler than the CPU package temps.

I agree that the temp reported for the H100i is almost certainly the water temp and likely the temp of the water on exit from the pump mounted on the CPU block. I don't know that for a fact but that would be the most reasonable place to put it.

I can make another test render and the one I did last night ran for about 25 minutes. What else should I be looking at as far as readings are concerned?


Brian