Help Me Understand CPU vs. Coolant Temp

[dcx4610]

Curious if anyone could answer these questions about the difference between CPU temp and coolant temp on AIO coolers.

I have a Corsair H115 AIO and an Intel 8700k CPU. On a fresh boot and while idle, my CPU temp hovers around 30C while the coolant temp on the AIO is around 26 or 27C.

I understand that coolant temp should always be lower than CPU as is just the nature of how it works. What I'd like to understand is how far apart they should be?

I was playing a 2D game (Guacamelee 2) and noticed that my CPU only got up to 36-39C while playing the game but the AIO was around 34C. Is it normal for them to stay failure close in temperature?

I haven't gotten around to anything intense yet like a FPS game and I know that some of those will probably make my CPU hit the 50s, 60s and maybe 70s. Considering the warning temp for the coolant is at 45C and based of my previous numbers, I can't imagine that would be good.

Hope that makes sense.

[DevBiker]

Check out the Liquid Cooler FAQ at http://forum.corsair.com/v3/showthread.php?t=174442. Details in there when discussing fan curves.

[c-attack]

Have a quick look at A-D in post 2 in the following thread. This person has some environmental issues to address, but it is the same CPU and cooler for some perspective.

http://forum.corsair.com/v3/showthread.php?t=184643

Originally Posted by dcx4610

I haven't gotten around to anything intense yet like a FPS game and I know that some of those will probably make my CPU hit the 50s, 60s and maybe 70s. Considering the warning temp for the coolant is at 45C and based of my previous numbers, I can't imagine that would be good.

Don't worry about this. Your CPU will go to 50-60-70C and it's fine. The coolant will not. Some of this is addressed in the thread, but the supposed max coolant temp of 40C is not a fail point. However, if you were starting a typical room temperature (for PC labs) of 20-23C, then that would represent a +17C coolant gain. That's a lot of watts. On the other hand, if your case is 39C like the guy in the thread, then it means nothing. He can't make it colder and no his system won't fail. However, that extra 15C+ of coolant adds 15C to his CPU temperature as well. If you were 60C, he would be 75C. You take a penalty for having a high coolant temp, but it does not mean you are in danger.

[dcx4610]

Originally Posted by c-attack

Have a quick look at A-D in post 2 in the following thread. This person has some environmental issues to address, but it is the same CPU and cooler for some perspective.

http://forum.corsair.com/v3/showthread.php?t=184643




Don't worry about this. Your CPU will go to 50-60-70C and it's fine. The coolant will not. Some of this is addressed in the thread, but the supposed max coolant temp of 40C is not a fail point. However, if you were starting a typical room temperature (for PC labs) of 20-23C, then that would represent a +17C coolant gain. That's a lot of watts. On the other hand, if your case is 39C like the guy in the thread, then it means nothing. He can't make it colder and no his system won't fail. However, that extra 15C+ of coolant adds 15C to his CPU temperature as well. If you were 60C, he would be 75C. You take a penalty for having a high coolant temp, but it does not mean you are in danger.

Thanks for the post. Definitely seems like it's the GPU waste heat affecting the coolant temp and not the CPU usage as I had thought.

[GuruSR]

Originally Posted by dcx4610

Thanks for the post. Definitely seems like it's the GPU waste heat affecting the coolant temp and not the CPU usage as I had thought.

I'm noticing that same issue with my 6700K and the H80i V2, the rad hoses go right past the GPU at the bottom of the case and over top of the GPU (1050Ti-OC), the heat down at the bottom of the case is all the GPU and pump fan output as the temp creating chips are above the GPU and aren't being affected by any internal fans aside from the one up top pulling only chassis air out. The H80i V2 is up front with dual fans shooting air along the bottom of the chassis which is blocked by the GPU (~80% wall), so most of the air flows along the bottom of the chassis and out, but sadly, the hoses are right in it's path, so it too is also affecting performance that way (shame I couldn't mount that upside down, but it wasn't possible as the GPU and the hoses hit 100% and there was no logical method of putting the hosing in without some serious bends and the tension on the mount would have been violent).

Suggestion would be to run a torture test (I know, everyone says "NO!" about such "Silly to do a fake test"), but Prime 95 on small FFTs will cause some pretty nasty heat, if you can get the temps to stay "sane" (less than 80C is best, the cooler it is, the longer it'll last) and not be super noisy, then you'll soon have a quiet and cool PC, but remember that any good test takes an hour at a minimum, especially with liquid cooling because the temperature of the liquid takes longer to heat up than the CPU does and longer to cool down as well. I have RealTemp running with a "Sleep at 58C" safety so if my cores get to the point where they're going to hit that Tcase value of 64C, I want the machine to stop before they get there, don't need another SWR for a replacement CPU. So for that 8700K, I'd avoid getting it past 80C, just because it *can* hit 100C, doesn't mean you should. Hotter = shorter lifespan for CPU, cooler = longer lifespan for CPU (well, for anything really, even the pump). The nice thing about RealTemp is you can set it to 100ms for monitoring the temps so it'll catch spikes, plus you can have the pump software open (iCUE, Link) to monitor the pump temp. Just have to "try" curves, would recommend being harsh on the fans and working your way down, rather than starting low and trying to recover, less possible damage that way.

GuruSR.

[DevBiker]

Prime95 only stresses the CPU. And it stresses it in ways that aren't indicative of real-world usage.
Something like ROG RealBench is a better option. It stress both CPU and GPU and gives you a better idea of what real-world maximum heat is going to be like.

[GuruSR]

Originally Posted by DevBiker

Prime95 only stresses the CPU. And it stresses it in ways that aren't indicative of real-world usage.
Something like ROG RealBench is a better option. It stress both CPU and GPU and gives you a better idea of what real-world maximum heat is going to be like.

I'll agree to disagree partly on this, Prime95 is like every other application within the Windows environment and is subject to the same time slicing that creates the illusion of multitasking, which is why whenever I want to heat the entire system up, I use (if a GPU card is present) a bench like 3dMark along with Prime95 (at the same time), actually I've run up to 3 benches at once on some systems (because there were multiple video cards and the benches didn't work with both, but recent ones do, though not 100% sold on them). The original OP was asking about temperature of the CPU and coolant, where Prime95 would give the "unearthly" max temperature and would tell them the maximum the CPU would push it to. If I were the OP, and was worried about the temperature of the GPU pushing the thermal of the coolant above the wanted 40C max, would be to buy a new case where the rad and hoses will get nowhere near the card. Using Prime95, I was able to determine that my coolant hits roughly 37C after 16 hours of torture, then with 3dMark running along side it, the temperature stayed there and the fans just kicked up a bit more (on the GPU mostly and a bit on the pump), though I rarely push that card above 40C anyways, so for my system, I didn't need to worry about the bench for the GPU. If the OP is having issues still with the GPU, perhaps a side photo of the inside of the case, could help assist in maybe relocating the rad and hoses to avoid GPU heat, or if the case permits fans on the bottom, putting a fan on the bottom and using it to force air into the video card, though you may need to customize air fins to avoid backwash into the rad.

GuruSR.

[DevBiker]

I stand by my statement. Unless you are running AVX all the time (and hardly anyone is), Prime95 stresses your CPU is ways that are not indicative of a real-world situation. Running it with the likes of a GPU bench tool will give you an idea of the max heat capacity of your entire system but the OP wasn't talking about that. They were testing with Prime95 only. And it's still nothing close to real world - it'll heat the CPU far more than any normal usage will.
And with multicore systems (and what modern system isn't multicore?), multitasking isn't an illusion. It's an illusion on single core systems. Sure, Windows handles the scheduling just like any other app (though Prime95 tends to run on lower priority threads) but that's not my point.
RealBench does a decent job of stressing both pretty well and simultaneously. It's a bit closer to a realistic scenario.

[c-attack]

This is something of a moot point. We originally were discussing why coolant does not rise to same degree as the CPU itself and also why coolant may go up in relation to environmental factors. Those factors are typically measured in single digit values.

Testing is always specific in its application. If you run Prime 95 for 16 hours, you have demonstrated your computer can run Prime 95 for 16 hours - nothing else. It does not mean it can run it again for 16 hours the next day or your fan settings are perfect for video encoding or your system is indestructible. All of these stress tests are predictors. "I ran 10,000,000 calculations without error, so I can probably run 10,000,000 more without error." And of course the answer to that is always maybe to probably. If you are trying to set-up your fans and curves, do what you normally do. Nobody accidentally initiates a multi-instance Prime 95/3D mark run and needs their system to instinctively react because they were out having lunch. Design your system for the task at hand. Don't put roll bars and 5 point safety belts in your station wagon.