Confusing thermals

[Short_Circut]

Ok, so I finally decided to switch things up from air/aio cooling to full water cooling. Almost all my parts are hydro x (XD5 pump, XC7 cpu, xr5 360, 6 QL fans) apart from the gpu which is from phanteks.

 

After finally building my updated rig I decided to check temperatures. So I was dissappointed to see that the temperatures were not only higher than on air, but also unecessarily higher. While before I was getting ~60C on CPU (i7-9700k) and ~55 C on GPU (strix 3090) (both were undervolted) with the liquid setup I was seeing both CPU and GPU hit ~70.

 

So I ended up taking off the panels and stress testing again, and lo an behold the temperatures were what I was expecting it to be. CPU hit 50's ish and GPU settled around ~52-53 which is not bad. However, once I put the side panel back on expecting the temperatures to rise up to prove my potential theory, I noticed they did not really change all that much apart from maybe a degree or 2 (but nowhere near the 70 mark I was hitting), so now I'm confused as to what is causing the super rise in temps. If I game with the panels on, then temps soar high. If I remove the panels, stress and then put the panels back on, temps seem to be better

 

Idle temps arent too bad, GPU sits around ~30C and CPU around ~40, but what surprises me is the coolant temps. They seem to idle around ~37-38C and under full load they can get up to as close as 50 C

 

My current setup (o11 dynamic xl) has Pump --> GPU --> CPU --> rad (mounted on top) --> pump. I currently have 2 QL 140's as bottom intake, 1 ql 120 as rear intake, and 3 ql 120's for the rad. I tried experimenting with switching between intake and exhaust, and found that intake gave better temps so stuck with that

 

I'll try and give benefit of the doubt and wait at least a couple of days for things like thermal paste to get up to speed and whatnot, but I am really confused as to what is going on (particularly with coolant temps). I feared it could've been a cpu/gpu mounting issue, but those times where the temps were perfectly fine seem to indicate otherwise?

[c-attack]

Before you had two separate cooling mechanisms for each device. Now you've combined them into one. A 9700K can output between 125-200W at 100% depending on instruction type and your voltage/power settings. It doesn't seem like you have set a high overclock, so that is going to be on the 125W end for stress tests and more like 60-80W during gaming. That is a fairly nominal amount of heat for a 360mm radiator to get rid off and it can handle that without much effort or coolant temp change. However, you have added a 300W GPU into the mix that can use over 400W depending on model and power settings. In effect you quadrupled the wattage that needs to be dissipated without making any other changes. Your gaming mixed load coolant (and thus device temps) are going to be much worse with any AAA game at 60W+300W compared to the 125W max running a fixed load CPU test. You need a 2nd 360mm radiator.

 

The other thing I noticed is your ambient case temp and idle coolant temp seem a bit high, unless your room is 30C+. Fully warmed up and running all day long, my XL build is going have a coolant temp of about +5C over the ambient room temperature measures in the area of the case. That has been a consistent value through all of the dozen or so configurations I have put together during the past year. I think part of this could be your 3090 is like many of the others that idles in a moderate power state rather than completely stepped down. However, we probably need to look at case layout for you anyway since you are going to need to add a second 360mm.

 

There are three basic layouts most people use for a O11D/XL. The first is dual 360mm exhaust. This is going to be top+side radiators as exhaust and the bottom as intake. I would leave the back slot open as passive intake for the top radiator. I am running this now. It has the advantage of dumping waste heat directly out of the case and thus keeping other components cooler. The trade-off is the intake side of the radiator is going to see air that is 4C or so warmer in that nexus where the two rads come together. That is all of your MB, RAM components heat being exhausted. Ultimately, the amount of wattage a radiator can dissipate is affected by the temperature of the intake air into that radiator. However, this is most notable if you are pushing your radiators to the limit and getting -1C more cooling from the radiator if you then have add +2C to the intake air temp may not always be a good trade.

 

You can run both radiators as intake. The logical placement is bottom + side, then use the top and rear as exhaust. This keeps the intake air temp into the radiator as cool as possible (coming from outside the case), but then you are dumping that waste heat into the box. For some users this has no meaning and they do not care the heat passes through the case. The items most sensitive to this are m.2 drives and RAM. If you have one of those "on a stick" m.2 holders that projects it naked out from the motherboard, I strongly advise not using an intake set-up. It could cost you 10-20C on the drive temps. I started this way when I first built my XL, but it costs me about +8C on RAM temps during gaming/load. That's not a lot, but it takes me over 50C during Summer. Well within spec, but different things matter to different people. I run my RAM pretty high and do not want to add another obstacle.

 

Funny enough the most common layouts you will see if you peruse the internet is the top/bottom radiator, often with a flat wall reservoir or the like in the side slot. Aesthetically this is pleasing but the thing that kills me with most of these is they have a bottom intake radiator, a top exhaust radiator, and then the side wall is blocked up. This means they are exhausting the waste heat from one radiator into the other. Don't do this. The O11 is too versatile to be forced into that kind of set-up. If you want to do this, either do top/bottom as exhaust or top/bottom as intake. If you do the exhaust, you can get away with blocking the side wall and even letting the rear be passive intake. That is a quiet set-up. If you do intake, you are going to need active exhaust probably on both sides to evacuate the heat. I ran both of these about a month or two ago as test builds. The top/bottom intake led to the same results as the bottom/side intake with noticeable warmer internal temps and those were passed onto the coolant/radiators if I did not keep the exhaust flow up. The top/bottom exhaust was an interesting set-up, however be aware there is a definite limit to how much radiator you can stick up top (~45mm thick) and if you are exhausting from top and bottom, you want to make sure you don't put the case under a desk, in a tight corner or other space with no room airflow. It is susceptible to recycling it's waste heat if there isn't room air movement.

[Short_Circut]

Before you had two separate cooling mechanisms for each device. Now you've combined them into one. A 9700K can output between 125-200W at 100% depending on instruction type and your voltage/power settings. It doesn't seem like you have set a high overclock, so that is going to be on the 125W end for stress tests and more like 60-80W during gaming. That is a fairly nominal amount of heat for a 360mm radiator to get rid off and it can handle that without much effort or coolant temp change. However, you have added a 300W GPU into the mix that can use over 400W depending on model and power settings. In effect you quadrupled the wattage that needs to be dissipated without making any other changes. Your gaming mixed load coolant (and thus device temps) are going to be much worse with any AAA game at 60W+300W compared to the 125W max running a fixed load CPU test. You need a 2nd 360mm radiator.

 

The other thing I noticed is your ambient case temp and idle coolant temp seem a bit high, unless your room is 30C+. Fully warmed up and running all day long, my XL build is going have a coolant temp of about +5C over the ambient room temperature measures in the area of the case. That has been a consistent value through all of the dozen or so configurations I have put together during the past year. I think part of this could be your 3090 is like many of the others that idles in a moderate power state rather than completely stepped down. However, we probably need to look at case layout for you anyway since you are going to need to add a second 360mm.

 

There are three basic layouts most people use for a O11D/XL. The first is dual 360mm exhaust. This is going to be top+side radiators as exhaust and the bottom as intake. I would leave the back slot open as passive intake for the top radiator. I am running this now. It has the advantage of dumping waste heat directly out of the case and thus keeping other components cooler. The trade-off is the intake side of the radiator is going to see air that is 4C or so warmer in that nexus where the two rads come together. That is all of your MB, RAM components heat being exhausted. Ultimately, the amount of wattage a radiator can dissipate is affected by the temperature of the intake air into that radiator. However, this is most notable if you are pushing your radiators to the limit and getting -1C more cooling from the radiator if you then have add +2C to the intake air temp may not always be a good trade.

 

You can run both radiators as intake. The logical placement is bottom + side, then use the top and rear as exhaust. This keeps the intake air temp into the radiator as cool as possible (coming from outside the case), but then you are dumping that waste heat into the box. For some users this has no meaning and they do not care the heat passes through the case. The items most sensitive to this are m.2 drives and RAM. If you have one of those "on a stick" m.2 holders that projects it naked out from the motherboard, I strongly advise not using an intake set-up. It could cost you 10-20C on the drive temps. I started this way when I first built my XL, but it costs me about +8C on RAM temps during gaming/load. That's not a lot, but it takes me over 50C during Summer. Well within spec, but different things matter to different people. I run my RAM pretty high and do not want to add another obstacle.

 

Funny enough the most common layouts you will see if you peruse the internet is the top/bottom radiator, often with a flat wall reservoir or the like in the side slot. Aesthetically this is pleasing but the thing that kills me with most of these is they have a bottom intake radiator, a top exhaust radiator, and then the side wall is blocked up. This means they are exhausting the waste heat from one radiator into the other. Don't do this. The O11 is too versatile to be forced into that kind of set-up. If you want to do this, either do top/bottom as exhaust or top/bottom as intake. If you do the exhaust, you can get away with blocking the side wall and even letting the rear be passive intake. That is a quiet set-up. If you do intake, you are going to need active exhaust probably on both sides to evacuate the heat. I ran both of these about a month or two ago as test builds. The top/bottom intake led to the same results as the bottom/side intake with noticeable warmer internal temps and those were passed onto the coolant/radiators if I did not keep the exhaust flow up. The top/bottom exhaust was an interesting set-up, however be aware there is a definite limit to how much radiator you can stick up top (~45mm thick) and if you are exhausting from top and bottom, you want to make sure you don't put the case under a desk, in a tight corner or other space with no room airflow. It is susceptible to recycling it's waste heat if there isn't room air movement.

hmm those are some nice points thanks. Do you think a 280mm rad would work? Would like to at least use my 140's without putting them to waste.

 

In light of this I decided to use the Cpro to monitor temps at different parts of the case and run heaven. I had set probes at right in front of the CPU block (so in between the rear exhaust and one of the rad intakes), one around the pump (so below another rad intake) and one outside to measure ambient. Average CPU power was ~40W and GPU was ~290W.

 

 

Temps didn't start off too bad on a completely cold boot. Looking at the results though.......yeaaaa almost a +20C delta coolant temp over ambient with just a GPU load is not looking like the best of things. I'm gonna try and run a gaming scenario with both and load and see how catastrophic things really do get rip

[Short_Circut]

ok quick update. Decided to remove the top cover and let the fans freely flow with no restrictions, ran heaven again. This time however, coolant temps stayed pretty within the ~10C window of ambient, eventually peaking at around 20C above. CPU and GPU temps however were great though.

 

So now I guess this leads to the whole aesthetic vs. performance scenario. I won't lie having the top panel removed does make things look pretty unpleasing, but I guess if it makes temps more in line with what it should be then that's fine. I did end up purchasing a 280 which is set to come in 4 days so we'll see what happens with that

[c-attack]

I am surprised there is so much resistance from the top, unless that includes a dust filter. You don't need that for top exhaust and that will have a clear impact.

 

The O11 is designed for 120mm x3 sets of fans, but you can probably center a 280mm on the side and not throw the proportions off too much. The top has clear 140mm channels as well. A 280mm can theoretically keep up with a 360mm radiator, but it requires pretty high fan speed to do so. The difference between at like speeds (noise) is usually around 2C, so adding a 280mm is absolutely preferable to not adding anything.

 

The slow, continuously building coolant temp is usually a sign you can't expel the heat from the system, one way or another. Physical restriction is a possibility, but I don't think we have talked about fan or pump speed either. You don't have to max these fans out, but 120mm radiators fans will have a dramatically higher coolant temp delta at 750 rpm vs something like 1300 rpm. This would be magnified with a combined GPU/CPU load. Pump speed is trickier to assess because each loop is unique and the level of flow restriction difficult to measure. 1 rad + 2 blocks is not a lot of resistance, but if you are keeping the pump at 1500 rpm or lower you likely will see measured temperature increases.

[LeDoyen]

To put it bluntly, QL fans suck on radiators... you can use them but you better have a lot of radiator surface to make up for what they lack in static pressure. That's basically why removing the top panel has so much effect on coolant temp. they can't overcome both the rad and the panel restrictions very well, unless you really crank up the speed.

When gaming if it's a demanding one i imagine it could be a case of bad thermal runaway, or close to. Remember the water should never reach or exceed 60°C. That's the maximum rating of most pumps.

 

With a 3090, i'd slap as many rads as i can put in the case to have good temps without too much fan noise.

Two rads would be a minimum to start seeing benefits from watercooling the GPU. More than 2, you make the PC more silent for the same temps, or cooler with the same fan speeds.

Just as an example, my water temp caps at 37 - 38°C when gaming with ambient around 24°C, and GPU hovering around 46°C (TUF 3090, the overclock doesn't change anything since it hits power limit. same temps as factory settings). That is with two 360 and one 240mm rads.

 

And regarding the 3090 idle power, make sure you set the power mamagement mode in the Nvidia control panel to "normal" instead of max performance.

That should drop idle power usage from almost 120W to under 40, and drop your idle coolant temp quite dramatically. But it won't change the troubles you have under load, of course ^^'.

[yzonker]

If I understand correctly, you have a single thin (30mm) 360mm rad trying to cool both a 3090 and CPU?

 

If so, then yea that's not nearly enough radiator area. Maybe if you had the best 120mm fans (Noctua A12's) you could make it acceptable. The A12's are so quiet that you can almost run them at full speed (2000 rpm) without them being overly loud (audible though of course).

 

I'll give a little history of my build effort very similar to yours. Through all of this my ambient temp is in the 20-23C range.

 

I started with a single EK 360mm thin radiator to only cool my 3090 running at 390w. I had a CPU AIO, so wanted to keep it. Well even though this setup did work, I was disappointed with the temps I got. Water would hit 40C or more with full load on the 3090. That was with the Noctua F12's running full speed (yes I bought the wrong fans the first time, should have gotten A12's).

 

So then I added a thin 280mm rad to the front of my case (Corsair) along with adding the cpu (5800x) to the loop. This helped quite a bit despite adding more load from the cpu. But I still hadn't done my fan research and used Noctua A14's for the 280mm rad. But water temps at this point would at least stay below 40C.

 

Finally I started researching fans. Discovered how much better some of the newer style fans are (Noctua A12 and Arctic P12/P14 to be specific). These fans can run at higher rpm while delivering the same noise level as older style fans at much lower rpm. So bit the bullet and swapped out all 5 of my rad fans and finally got down to the mid 30's for water temps. Giving an air to water delta of around 11-12C. 3090 core temp is around 47-48C @390w. (actually 385w right now running the Zotac OC bios). Ambient 22-23C. Corsair block.

 

All my fans are exhaust. Rad performance would be the best with them as intake, but 3090 VRAM temps can go very high even with case temps near ambient. The last thing I want is to pump 5-10C warmer air in to the case from the rads.

 

It's just a lot of heat to dump out with a cpu+3090.

[c-attack]

I do have a case full of QL fans. While not an ideal radiator fan, when you get into large amounts of radiator surface area fan choice makes only a small difference. You might see 1 or maybe 2C difference between the best and worst. That is not the 10C we are looking for and you can never get that with fan choice. Most fans at like speeds are going to be similar. That is different than fan B has a different tone so I don’t mind running it at 2000 rpm vs 1300 rpm for another. I squeak just under the coolant delta from other two examples above with 2x360 and QLs at 450W, so there are other ways to steal back that 1-1.5C. These represent fine tuning options. Right now you need a major change (another radiator) and that process is underway. Edited April 25, 2021 by c-attack