Hydro X i9 9900k Temps Advice

[Jules MK]

This is my first custom loop so I'n not sure if the temps are to be expected.

I have a an i9 9900k in a Corsair Obsidian 500 RGB SE case (biggest case I can fit with some airflow around) with a custom loop, I have "messed" about over the last 2 months and the temps I get while gaming for 2 hours this evening were:

 

CPU i9 9900k - Peaked very briefly at 78 deg but sat high 60’s - Is this high or ok?

GPU RTX 3070 - Max 56 deg and held around 54

Coolant XL5 - Max temp was 36 deg and it held that

 

The CPU is on an XMP1 setting, if I stop OC it peak at 68 but sits 64 ish

The GPU is Graphics 2295 MHz - Memory 8201 MHz 104% Power OC using Afterburner and is stable

 

The cooling system is:

280MM rad in the top 2 x ML140 Fans exhaust

360MM rad in the front 6 fans push (3 x ML120) pull (3 x LL120) intake.

120mm LL120 fan exhaust on the rear of the case, plugged in to the CPU fan header

XD5 pump

XC7 CPU water block

Alphacool GPU Eisblock (Ice Block) Aurora Acryl GPX-N RTX

 

Custom fan/pump setting all based on the coolant temp:

Thanks to LeDoyen for setting me down this path

 

120mm fans

20 - 600

25 - 775

30 - 1200

33 - 1500

37 - 1690

40 - 1951

 

140mm fans

20 - 600

25 - 700

29 - 1000

33 - 1300

37 - 1400

38 - 1500

 

Pump

23 - 2424

26 - 3030

29 - 3484

33 - 3817

36 - 4241

40 - 5000

 

Does the difference between the coolant temp and the CPU temp seem too large or as the i9 apparently runs hot this nothing to worry about please?

 

Are the GPU temps ok considering the OC please?

 

Also is there anything I can do differently with fan curves that won't increase the noise but help cooling.

 

Pic for ref (I have topped the fluid up now):

 

 

 

Thank you for reading!

[c-attack]

Does the difference between the coolant temp and the CPU temp seem too large or as the i9 apparently runs hot this nothing to worry about please?

 

No, not at all that 36C to 60-70C gaming temp is submaximal. Good way to test this is with the stress test in CPU-Z. It's a fixed load so the CPU temp you see 1 second in will clearly show your 100% coolant to CPU temp differential. That also tells you your max coolant temp in regards to your CPU temp safety zone. (max coolant to CPU delta is 45C. I don't want to be over 85C at the CPU, thus max allowable coolant temp is 40C, etc.)

 

 

Are the GPU temps ok considering the OC please?

 

This one is harder to assess. I would normally say +20C for 104% power (320W?) is slightly high. However, this is both a unique property to each GPU chip and I don't have a lot of 3070 data from other users. So definitely do not strip the block off tomorrow. The Ampere cards as a whole run with a slightly higher differential than Turing. One thing you can check out when gaming or running any kind of GPU test is to set the pump up to 80% PWM or so, then check coolant to GPU temp differential again at 100%. If you get a marked drop of 3-4C, you know the fluid is lingering in the GPU block a bit longer than is optimal. If it remains unchanged, then you know you don't need to run those pump speeds, but you can check overall performance as well. I would not run higher pump speeds for CPU benefits when gaming because the load is fairly varied compared to running a maxed out.

 

 

Also is there anything I can do differently with fan curves that won't increase the noise but help cooling.

 

You should make your fan curves to your noise tolerance level. Water cooling is not so sensitive that +-100 rpm makes any difference at all. For example, a 300W load into a 360mm radiator with 3 fans at 1300 rpm will typically have a coolant rise of around 8.5 to 10C depending on the radiator characteristics. Ramping the fans up to 1800 brings that down to 6.5-7C or about a 2C improvement. For most people the noise increase from 1300 to 1800 far outweighs the fairly nominal 2C improvement to liquid and component temps. The only place you may see large swings is when hanging around at very low fan levels. The difference between 400 rpm and off is very large and the difference between 750 and 400 rpm is also significant. Once you get up to about 1000 rpm, the curve and returns flatten out quite a bit.

 

 

If you only get up to 36C with this set-up I would say you are doing quite well, especially since the front radiator is dumping its waste heat into the top. However, you have set it up in a workable way with the post GPU load going to the top first. That puts the warmest coolant up top and the exhaust air coming off the front array will be just slightly cooler so you still get some cooling from the top. Theoretically you might get the best results from running the top and front as exhaust while the turning the rear fan to intake. The radiator fans will pull in the extra needed air through the back slots and mesh. I am not going to tell you to do that and a reverse flow is not for everyone. It also puts a fair amount of dependence on having cool air in the back of the case. That might be true in Summer if you have an AC vent back there or it could be jus the opposite in Winter with heat pumped into the case.

[LeDoyen]

the 10th gen intel have thinned die which will account for a lot of the temp difference you see compared to 9th gen.

[Jules MK]

No, not at all that 36C to 60-70C gaming temp is submaximal. Good way to test this is with the stress test in CPU-Z. It's a fixed load so the CPU temp you see 1 second in will clearly show your 100% coolant to CPU temp differential. That also tells you your max coolant temp in regards to your CPU temp safety zone. (max coolant to CPU delta is 45C. I don't want to be over 85C at the CPU, thus max allowable coolant temp is 40C, etc.)

 

 

 

 

This one is harder to assess. I would normally say +20C for 104% power (320W?) is slightly high. However, this is both a unique property to each GPU chip and I don't have a lot of 3070 data from other users. So definitely do not strip the block off tomorrow. The Ampere cards as a whole run with a slightly higher differential than Turing. One thing you can check out when gaming or running any kind of GPU test is to set the pump up to 80% PWM or so, then check coolant to GPU temp differential again at 100%. If you get a marked drop of 3-4C, you know the fluid is lingering in the GPU block a bit longer than is optimal. If it remains unchanged, then you know you don't need to run those pump speeds, but you can check overall performance as well. I would not run higher pump speeds for CPU benefits when gaming because the load is fairly varied compared to running a maxed out.

 

 

 

 

You should make your fan curves to your noise tolerance level. Water cooling is not so sensitive that +-100 rpm makes any difference at all. For example, a 300W load into a 360mm radiator with 3 fans at 1300 rpm will typically have a coolant rise of around 8.5 to 10C depending on the radiator characteristics. Ramping the fans up to 1800 brings that down to 6.5-7C or about a 2C improvement. For most people the noise increase from 1300 to 1800 far outweighs the fairly nominal 2C improvement to liquid and component temps. The only place you may see large swings is when hanging around at very low fan levels. The difference between 400 rpm and off is very large and the difference between 750 and 400 rpm is also significant. Once you get up to about 1000 rpm, the curve and returns flatten out quite a bit.

 

 

If you only get up to 36C with this set-up I would say you are doing quite well, especially since the front radiator is dumping its waste heat into the top. However, you have set it up in a workable way with the post GPU load going to the top first. That puts the warmest coolant up top and the exhaust air coming off the front array will be just slightly cooler so you still get some cooling from the top. Theoretically you might get the best results from running the top and front as exhaust while the turning the rear fan to intake. The radiator fans will pull in the extra needed air through the back slots and mesh. I am not going to tell you to do that and a reverse flow is not for everyone. It also puts a fair amount of dependence on having cool air in the back of the case. That might be true in Summer if you have an AC vent back there or it could be jus the opposite in Winter with heat pumped into the case.

 

 

Thank you for the advice, I have slowed the fans down to a level I am happy with the noise, 3 hours this evening and the CPU was actually slightly cooler (ambient might be cooler I guess).

 

I did try all the fans except the rear on exhaust but it made no difference (or maybe slightly worse), I guess this is a combination of the little cubby the case in and also the Obsidian 500 SE case has quite a small area for air flow in the front.

 

I will test the GPU and let you know the results.

 

Thanks again.

[Jules MK]

 

This one is harder to assess. I would normally say +20C for 104% power (320W?) is slightly high. However, this is both a unique property to each GPU chip and I don't have a lot of 3070 data from other users. So definitely do not strip the block off tomorrow. The Ampere cards as a whole run with a slightly higher differential than Turing. One thing you can check out when gaming or running any kind of GPU test is to set the pump up to 80% PWM or so, then check coolant to GPU temp differential again at 100%. If you get a marked drop of 3-4C, you know the fluid is lingering in the GPU block a bit longer than is optimal. If it remains unchanged, then you know you don't need to run those pump speeds, but you can check overall performance as well. I would not run higher pump speeds for CPU benefits when gaming because the load is fairly varied compared to running a maxed out.

 

 

I tested the GPU this evening and the temps didn't change when running the pump at 100% comparted running the pump at 80% while gaming. I assume this is a good thing!

 

I think the location of the PC is quite a contributory factor as the air drawn in from the top right front of the case is really warm, so that's not going to help but I can't change that so it will have to do.

 

The coolant temp remains at 36 degrees playing for 3 to 4 hours so I guess that's the best I can do until I can change the case and or at least the location of the case.

 

Thanks again.

[LeDoyen]

That temp difference on the GPU often varies from water block manufacturer to manufacturer. With those large dies, the cooling engine (fin types, flow type), the thickness of copper and type of thermal paste plays a bigger role than with CPUs.

 

If you have 20°C dT between water and GPU that's directly tied to the block and paste efficiency. Could also be the mounting itself, but usually mounting issues lead to higher temps than that.

 

I just did a test out of curiosity on the 3090 at 300 - 320W in game with water at 35°C, the temp difference was ~11-12°C at a quiet pump setting (72%) and ~9°C with the pump at 100%

It was close to 20°C with the pump at idle settings at 50% PWM.

 

 

Moving the computer will mechanically drop the GPU temp by the same amount the water temp drops, but this 20°C dT should remain i believe.

Maybe you could try switching OC / no OC while gaming to see if that figure changes at all. In the end the GPU hits a hard power limit so i'm not sure overclock will change anything.

[Vegan]

When i first installed my water cooler I removed the OEM thermal pad and used a small dab MX-4. I have an H115i RGB 280mm cooler and my CPU barely flinches even with my most sadistic software.

 

I have tools 100x more harmful than CPUID can do.