Estimating coolant temp with GPU idle changes

Grim5 · 27 Apr 2021 at 23:05

So I'm trying to get a rough estimate of my coolant temp without a sensor and want to know if this idea I came up with gets in the ball park.

So here is my idea

* I've started the PC up, made sure the room temp is stable (lets say its 20c).
* Then I've let the GPU idle for 10 minutes, temp sits around 22c.
* Then I'll run a stress test for 10 minutes with the GPU going all out at 480w. Temps flatten at 54c
* Then I shut off the stress test and within 1 second, the GPU temp drops to 30c.

So what I'm thinking is the first temp recording after the stress is completed (30c) minus the GPU idle temp (22c) plus the room temp (20c) = rough coolant temp of 28c?

SirConfused · 28 Apr 2021 at 14:00

To my understanding, all you can say with that is that your water delta is 8 degrees. What you can't tell is the water temp on idle because you've raised it from the 20 degree room temp but you can't guarantee that it's exactly what your gpu temp is.

It's probably not far off, assuming you have good rad capacity and your fans are pushing at a decent rate, but there's still an unknown that will throw off the values by at least a small amount.

MikeTimbers · 28 Apr 2021 at 15:54

cooking thermometer in the reservoir? body temperature thermometer in the reservoir?

Drumroll · 28 Apr 2021 at 15:57

Close enough. Idling 3900x + RTX3090, between 1C and 1.5C over ambient. Under load, and fans at 1000isg RPM, delta stays between 5C and 6C under load.
Considering your GPU is dumping much more heat than mine, 54C is quite good.

LuckyBenski · 14 May 2021 at 16:40

Grim5 said:
So I'm trying to get a rough estimate of my coolant temp without a sensor and want to know if this idea I came up with gets in the ball park.

So here is my idea

* I've started the PC up, made sure the room temp is stable (lets say its 20c).
* Then I've let the GPU idle for 10 minutes, temp sits around 22c.
* Then I'll run a stress test for 10 minutes with the GPU going all out at 480w. Temps flatten at 54c
* Then I shut off the stress test and within 1 second, the GPU temp drops to 30c.

So what I'm thinking is the first temp recording after the stress is completed (30c) minus the GPU idle temp (22c) plus the room temp (20c) = rough coolant temp of 28c?

Good logic so far. Maybe obvious to you but a few facts we can use to make assertions:

- Your silicon will never really be cooler than the liquid temp as it's the source of heat.

- The delta between silicon and block gets bigger with voltage and load, and this changes dynamically (no inertia/warmup time).

- The coolant temp is negligibly the same throughout the loop at any given time, so any variance between CPU and GPU is down to load.

So for example if your GPU maxes out at 55 long term load, you know your liquid has never been over 55 (and probably way less due to delta).

But also you can more or less estimate coolant temp by dropping the load after reaching max component temp. The GPU should only be a couple of degrees above this, as you've estimated.

It also highlights how much hotter overclocking makes the silicon delta, especially for CPUs. Going from small to moderate overclock bumped me 8-10 degrees on the CPU for a given liquid temp.

Honestly though I've spent 3 years guessing at this stuff and finally added a liquid temp sensor this weekend. It's so much better, I can set everything up around that instead of volatile component temperatures. Highly recommend it.

Brizzles · 14 May 2021 at 23:54

Drumroll said:
Close enough. Idling 3900x + RTX3090, between 1C and 1.5C over ambient.

I really wish people wouldn't post such rubbish as gospel.

There is no way you are running at 1 degree above ambient with any standard cooling including water, you are kidding yourself if you thing you are.

Drumroll · 15 May 2021 at 22:34

Brizzles said:
I really wish people wouldn't post such rubbish as gospel.

There is no way you are running at 1 degree above ambient with any standard cooling including water, you are kidding yourself if you thing you are.

Temperature of coolant, between 1C and 1.5C. Idling.
Over load, between 5C and 6C.
If helps to solve your bitterness, I can upload few photos.
Not a skinny rad, 2 PE 360, one push/pull and one XE 360 push/pull.
All sensors properly checked with icy water and boiling water, as I wanted to make sure all sensors were calibrated.
Idling won't be much load for my loop.
What scares me is when people say they have coolant temperature of 40-45+. If the coolant is that hot, what temperature they're getting on the actual CPU and GPU?
Normally for GPU I notice around 20C ish over coolant.
Sorry to disappoint you, but it's factual.
Sensor for coolant inside reservoir, before the XE which is the last rad before the cycle starts again.
Gaming for over 5 hours yesterday, Assassin's Creed, room temperature 24.5C, coolant 29.7C.
Fans at 35% idling, 45% under load.