Liquid Temp > CPU Package temp

Looking at CorsairLink I'm seeing that my liquid temps are around 4 degrees Celsius higher than the hottest core/CPU Package temp.

I've seen in many forums that this is not possible. If it we true then the H115i I have would be warming my CPU and not cooling it.

Speedfan doesn't work on my CPU which is a de-lidded I9-7900X, but I checked my temps with HWMonitor and it agrees with the temps listed for each core in Corsairlink.

The current and previous versions of CorsairLink both report the higher liquid temp.

So is this a faulty H115i, a buggy Corsairlink or something else?

OS is Windows 10 1709. Case is NZXT Phantom 820. There are Front, Rear, bottom, side and Internal fans in the case. The H115i is set as push, and I've recently vacuumed the dust from the radiator. Motherboard is MSI X299 Xpower Gaming AC.

Looking at CorsairLink I'm seeing that my liquid temps are around 4 degrees Celsius higher than the hottest core/CPU Package temp.

 

I've seen in many forums that this is not possible. If it we true then the H115i I have would be warming my CPU and not cooling it.

 

Speedfan doesn't work on my CPU which is a de-lidded I9-7900X, but I checked my temps with HWMonitor and it agrees with the temps listed for each core in Corsairlink.

 

The current and previous versions of CorsairLink both report the higher liquid temp.

 

So is this a faulty H115i, a buggy Corsairlink or something else?

No. It's not unusual. Temperature sensors, and especially the CPU sensors, aren't exact but have a accuracy range. It's not an issue.

The entire environment is far too dynamic to apply simply rules like coolant temp is 25C, thus cores are also 25C. You are right in theory, but it gets complicated by a few things. First, power saving features like EIST, Speedstep, Speedshift, C-states, etc. The cores may be dropping voltage constantly and as each one of your 10 cores pops on or off the micro local temp area changes. Nano second changes with poling tools that only gather data every 1-2 seconds.

Generally, you can think of package temperature as socket temp, rather than the total cumulative heat from all cores. On some CPUs it is a specific temp sensor and the location may vary, but this is cpu specific and not likely to run in parallel with the coolant temp. My HW-E CPUs were wacky with this, even with their soldered IHS. +10-13C over coolant at idle. This was completely different from my "CPU temp", a specific CPU sensor and maybe la closer representation of the lid/IHS temp. I don't know enough about the 7900X be specific about its various sensors, but across multiple CPUs Link does not always have the same CPU package value as AIDA, HWinFO, etc.

Also keep in mind the package temp sensor, cores, IHS, and coolant temp sensor are on different planes of heat transfer within that area. While it would be nice if heat transfer was 100% efficient, you should expect to see differences between the layers.

I agree it isn't an issue, but I understand the curiosity. Your delid likely alters the true package temp reading and you may get some differing results. If you are interested, I would also use HWiNFO and AIDA (without Link running) to see what they come back with. Unfortunately, you can't use Link with any other monitoring program right now and the data is not going to be reliable.

Thanks for your detailed explanation @c-attack.

My De-Lidded CPU is 6 months old. It's only recently that the behavior changed. A couple of weeks ago the coolant temp was lower than the CPU temps. What changed - I upgraded from W10 1703 to W10 1709. I'm going to restore my 1703 image and see what it shows.

I've also tried a couple of other programs, and they agree with CL's core temps.

I'm also running MSI Afterburner and have been for a long time. I haven't upgraded that program so nothing's changed there.

I turned my system on 15 minutes ago. The coolant has gone from 27.3 to 31.4 in that time. I've only been web browsing.

The only reason I'm concerned is because my fan curve, which up until a week or so ago was silent. Now doing nothing on the PC the coolant temp climbs until the curve hits a spot where it increases fan speeds, and within an hour or so it'll be sounding like a jet plane.

If I play doom, it takes several hours for the coolant temp to drop from 41 degrees back to 35 degrees. It will never drop back to where it started before starting the game. Previously it used to drop back within minutes. There was an issue with nVidia drivers > 382.53, but I'm using 382.53 with doom which doesn't exhibit the temp problems.

At 35 degrees the fans are very loud. Sure I can change the fan curve, and when I do the cooler stops cooling as efficiently and the CPU temps jump a few degrees.

Ambient temp is 19 degrees. Up until a week or so ago my coolant temps were around 5-8 degrees above ambient. My CPU cores are still reporting 5-8 above ambient, but coolant is now 10 - 16 degrees above ambient.

I'll report back when I've tested my 1703 image.

OK, well then you are right to be concerned. It is typical for the coolant to go +3-4C doing light work after booting or waking from sleep. That would match my 5930K clocked fairly high. My lower watt 8700K is more like +2C on the coolant after full warm up.

However, I am concerned about the difference in full idle coolant temp to ambient and especially at the very slow reduction post gaming. Those are warning signs the cooler has a flow issue. There are some other possibilities you need to eliminate.

1) Excessive dust build up or some other fan restriction (like a dust filter). I don't mean a little dust. I mean enough to block the flow of air. It would be carpet-like. I'll assume this is a no.

2) Anything else that can trap the heat and prevent the radiator from shedding its heat. Dust filter over the radiator. One of those dome tops that keeps heat in. Putting the entire case in a cabinet, etc.

3) Some other general airflow issue that keeps the case at an elevated temperature. Obviously is the entire case interior is 39C, then your coolant can't go below that as well.

Since this is a recent and developing problem, I expect this is a no to all of above. Every once in a while someone throws a curve ball and makes a physical change not mentioned.

I don't think you need to load your C drive image. I do think you should contact Corsair through the ticket support system at the top of the page and start a RMA request. These things usually get worse, not better. Anything abnormal with the pump speed? It should be ~2000 ish for Quiet and just under 3000 on Performance. Which are you using now? Do you see a drop in coolant temp when switching to Performance mode while on the desktop in light load?

Ok, I've done two things.

1. Restored my 1703 image

2. Reinstalled my old Kraken X62.

First, apologies for the length of this post.

1. The results seem confusing. The same Corsairlink version. Fan #1 is reported at 1427 RPM on 1703 and 976 on 1709 at around the same ambient temp (which is now 24 degrees Celsius). The Samsung 960 Pro boot drive is reporting 34c on 1703 and 39c on 1709. Coolant on 1703 started at 34.7 after boot and dropped to 33.5 after 10 minutes. CPU cores are around 27 - 29 at idle. On 1709 the coolant temp remains fairly constant at 34.5c-35.3c at idle. After 20 minutes of doing nothing it's sitting at 35.2c

Clearly I was mistaken about the coolant being lower on 1703. It is however, closer to the CPU temps, and reacts more quickly after quitting doom, but still slower than a few weeks ago.

I checked my fan curves to make sure they were identical.

2. After installing the X62 with 1703 the coolant was 1c lower than highest CPU core temp. I've never trusted the CAM software (It's riddled with bugs and telemetry), which is the sole reason I switched to the H115i. The fan #'s don't match up with CorsairLink but there were no fans reported over 900 RPM. CPU cores are sitting around 27c-28c at idle, about the same as what CorsairLink reports, but coolant is @ 26c with radiator fans at 840RPM.

I then restored 1709 with the X62 and the results were the same as 1703.

Now, to respond to your suggestions.

1. The case, all fans and radiator were cleaned two weeks ago. There was very little dust (I clean my system every few months).

2. The case is an NZXT Phantom 820. Yes it has a "dome" of sorts. I removed the filters when I removed the stock 200mm fans.

3. I'll describe the airflow configuration. Let's keep in mind this fan configuration has been the same for many many months. The only physical change inside the system was an upgrade of my 960 EVO 500GB to a Samsung 960 Pro 1TB (These are both NVMe M.2 drives).

Corsair links reports temps 1 - 5 as 41c, 46,5c, 46c, 41c, and 29c respectively. I have no idea how to correlate those temps to physical locations within the PC.

The radiator is top mounted in a push configuration with pump on performance mode (2910 RPM), with fans @ ~1100 RPM

At the Back I have a Noctua NF-A14 PWM exhaust running at ~500RPM

On the front is a stock NZXT 200mm case fan intake running at 450 RPM.

On the bottom is a Noctua NF-A14 PWM intake running at 430 RPM. There is 2"/5cm clearance under the case to allow air intake. This fan blows upwards across the PCH.

On the side is a stock 200mm NZXT case fan intake running at ~400RPM.

There is also an internal fan; A noctua NF-A14 PWM at an angle of 45 degrees approx that takes airflow from the bottom fan and distributes it around the CPU/VRM area. Running at ~575 RPM

The PSU fan exhausts downwards out of the case, not into the case.

I've now increased the rear fan to 650RPM to remove some of the positive air pressure that might be swirling around inside of the case. There is no demonstrable difference in temps though.

I'm quite happy to change the case (Price is no issue). But I am quite partial to white cases. I picked the Phantom 820 because of it's size (I had an E-ATX board at the time), the number of drive bays and fan mounts. I also have 3 optical drives installed. There are now only two SATA drives as I now have 3 M.2 drives on the motherboard, two of which used to be SATA.

The motherboard is an MSI X299 XPower Gaming AC, which I bought because it has 10 fan headers and 3 M.2 slots with pre-installed heat sinks. The graphics card is an ASUS Strix GTX 1080 which is managed by MSI Afterburner.

I'm also considering the purchase of the Corsair Commander Pro so I can manage fans from CorsairLink instead of having to boot into UEFI to do it.

Don't worry about the case configuration. Small differences in fan speed do not lead to doubling your coolant delta. Every once in a while someone forgets to mention they have 980 Ti x 2 in SLI and their CPU temps are strangely warm after gaming in 4K. They also don't notice the entire case is 42C and this the coolant must be as well. Since these were all new developments in the same case under the same conditions, it is very unlikely any of this is part of the problem.

I can't think of any reason why it should be different one from one instance of Windows to the next. The most likely reason is software error. It's just s simple temp probe inside the pump housing so this really should not happen. However, Link often assigns different pieces of hardware different names from one boot to the next on my Asus Z370 Code. Fan #3 may not be the same thing each time, so I am not sure how much weight you want to give it. In fact, since you no longer have the H115i installed, I would exit Link on each boot or uninstall it. I suppose it still could be used to check on the PSU, but we know Link conflicts with multiple monitoring programs, including CAM which you now must be running. Link is not the greatest total system monitoring tool ever.

To that end, with the X62 installed are things back to normal? You cannot make straight coolant temp comparisons between the CAM software and Link. I would like to say to this is because the temp probe is a different place and thus readings are different, but that fact is NZXT has taken to manipulating the liquid temp reading for their own unstated purpose. Which version of CAM are you running? The older series of the software preferred by some (3.3xx and before) will give you a higher liquid temp reading compared to Link. However, coolant deltas and end CPU temps will be more or less the same. Why it reads higher is unknown and I asked many times for an explanation, never getting one from them. It was then quite magical when NZXT released the 3.5 series for CAM with the redesign. They apparently also redesigned their temperature scale. The new CAM reads -4 to -5C on the old software and you can swap versions back and forth and see. Coolant deltas still the same. I can only assume they have done this to "make their coolers seem better" by posting lower temps, but the problem is they doctored the value to make it go below their controllable scale. I had coolant temps in the 18-19C range at idle, but the controls have a 25C minimum threshold. My normal max coolant delta is +6C, so I had to get to a full load before the fan controls became usable. This is a very long way of explaining you can't take the X62 liquid temp at face value. However, they are linear and the increase under load should be the same. It should also drops it heat after the load session. If it does not, that may be an indicator there is a case heat management issue.

A Commander Pro would give you desktop software control of fans. However, in order to regulate by measured dynamic variables like CPU temp or GPU temp, the Link software must be active. On the other hand, it can regulate from the 4 temperature probes without Link running. I find this to be very useful and it is why the C-Pro stays in my system even when I don't have any other Link hardware installed.

The X62 is not installed. I reinstalled the H115i. I only installed the X62 as a comparison test. I absolutely will not stay with the NZXT product and CAM.

Oh, also, with all this changing of coolers... It might be relevant to let you know that in all cases (pun intended), I put the side cover back on with the side fan running. I didn't do the tests with the side off.

According to UEFI there are no fans spinning over 900RPM. CorsairLink on 1709 insists that one of them is spinning at 1440 RPM. If that were true I'd hear it. I've just turned the system on and ambient is 17c and everything is quiet. So it appears it's not reporting the correct speed. Coolant temp is 30c @ 960RPM with pump @ 2910RPM. System is idling.

Also, HWMonitor shows the fastest fan spinning at 893RPM.

I only have the one graphics card.

The reason I'm concerned is that with no hardware changes the behavior changed recently.

CAM - I installed 3.1.71 and 3.3.50. They are the only versions that reliably detected the X62 and work at least semi-reliably. Understand your opinions on CAM - They are similar to mine, and why I will never voluntarily run an NZXT product again (I would like to get rid of my case too as a statement of disgust :)

According to the CAM software, my coolant temp drops back down within a minute or so. It's very quick.

It wasn't only the temps that were different - fans speeds were also being reported differently. I guess my point is if the same version CorsairLink reports two different fans speeds and temps between two Windows updates then how can one trust CorsairLink?

Last night I did some tests with Handbrake, which runs x264. This always warms my system. I had the coolant @ 33.6c when I started. Coolant reached 41.8c during the encoding, and after it finished it dropped to 34.5c but took 60 minutes to do so. Again, Pump on performance mode. CPU temps maxed at 70c, which I'm rather happy with on my I9-7900X. De-lidding works :)

I was planning to run CorsairLink with the Commander Pro.

How would I measure and test case heat management?

As a side note.... Reinstalling different versions of Windows and software is a trivial exercise which takes less than 10 minutes per version as I have imaged each version of Windows, and versions with CAM and versions with CorsairLink (But never together at the same time).... So I can do these tests very quickly.

Anyway, I took your advice and have opened a ticket with technical support. I'll see what they say.

Where is link reporting 1440 RPM? Can you post a screen shot?

By the way ... when you connect the pump to the CPU Fan port, it reports a fan tach signal to the header. This is about 1/2 the pump speed (not exactly but close). That 1440 sound suspiciously like that. This should also be reported as the CPU Fan speed in UEFI/BIOS. Now ... the default setting for the pump is Quiet, which runs the pump at ~2000 RPM. IF you changed that in Link to performance mode (as you stated) AND you did not write the settings to the device, then you'll see the higher value when Link is running because you set the pump to performance mode (~3000 RPM).

For non-Link hardware (CPU, motherboard, etc) ... Link actually uses the CPUID SDK. HWMonitor is built on the same. If you are looking for an "alternate opinion", use HWInfo or AIDA.

You mentioned that the pump is on performance mode. What are the fans set to? That makes more of a difference with the coolant temperature than the pump speed does.

What we don't have are your internal case temperatures. Sure ... ambient may be 17c but that's not going to be the internal case temperature. If you are running the pump as exhaust, your internal case temp matters far, far more than your external ambient temperature. That, plus your fan speeds (not your pump speed) can help us understand if what you are seeing is expected or if it's out of the ordinary.

The Commander Pro's temp sensors can help you get a good picture of what temperatures are actually doing inside the case. I also like to measure the radiator intake and exhaust temps as well as two different spots inside the case - 1 of these being just above the GPU.

CAM - I installed 3.1.71 and 3.3.50.

 

According to the CAM software, my coolant temp drops back down within a minute or so. It's very quick.

OK, that is really what I was digging at. This is further evidence the H115i is not quite performing as it should. It should have dropped 5-6C off the peak within 5-7 minutes unless it was a really heavy GPU+CPU mixed render that elevated the entire case temperature. I don't think that is what we are talking about. The change is performance also coincides with NO hardware or physical changes, further implicating the cooler. I do not think case heat management is the issue. I do think you have a developing blockage somewhere in the system.

Yes, CAM 3.3.50 is my preferred version as well. Just keep in mind you can't make 1 to 1 comparisons to Corsair coolers on the liquid/coolant temp. The CAM software will read higher, but coolant delta and end CPU temps will be the same.

I can't help you with the extra fan and temp readings in Link. It is going to be motherboard/chipset specific. My advice is to shape the Configuration panel picture with readings and temps you know to be valid. You can use that as the monitoring center. The front Main page of my Link has 21 entries for the motherboard alone. I obviously don't have 12 temp sensors on the board the fans are duplicated as well. Some temps perpetually read 0.0. Others are nonsensical. There is nothing on my board that would have a 78C value.

Here you go...

Ambient at the time of the screenshot is 25.7c

My fan curve; 1(20C=25%), 2(30c=30%), 3(35c=40%), 4(40c=55%), 5(45c=-75%), 6(50c=100%)

Playing doom will put me in points 4-5. At point 4 my fans are spinning at 1600RPM.

Well, I got a response from technical support. It was nothing more than a fob off telling me that the unusual behavior I'm seeing is completely normal.

I'd already paid for a replacement unit (as sending back a faulty unit takes 3 - 12 weeks as I'm in Australia), so I need a cooler in the interim. After the response I got, had I not already paid for a new unit, I'd be changing brands. Pretty disappointing.

It is unfortunate someone did not take the time to troubleshoot this thoroughly. You are running +11C idle coolant over room ambient. Normal ranges are 4-7C. You might run toward the higher end with a 10 core, but only if you have all c-states/EIST, etc. turned off and perhaps the power plan on performance with 100% uptime. The really long cool down is a red flag as well, as long as the entire case isn't that same temperature.

At the desktop in normal idle, what happens to the coolant temp if you downshift the pump into quiet mode (~2000 rpm)? On a normal working cooler, this should have zero impact. On a cooler with a partial blockage, it may have a measured effect.

Dropping either the fans speeds "marginally" or changing the pump to quiet mode results in the coolant temp increasing by 1c - 2c; On idle over a period of 10 or so minutes

I guess the 1440 RPM fan is the pump, as you mentioned it's half the pump speed.

I'll put in the new cooler tomorrow. If that solves the problem I've then got to find a way for Corsair to accept that mine is faulty.

I was quite disappointed they didn't ask me any questions.. No troubleshooting at all. They simply stated that Ambient temperatures are always subjective and the escalating temp and non-cooling of the coolant is a symptom of the heat expelled by the cooler, in effect, raising the ambient temp in a self perpetuating cycle.

Running the pump in performance with fans @ 1500RPM I cannot get the coolant below 34c @ 24c ambient. Of course, 1500RPM is intolerable noise wise.

I'll know more tomorrow.

EDIT: My power plan is "balanced". I am using all power saving features pretty much at default. Only rarely do I run on a performance power plan. So typically my CPU is idling at 1.19ghz.

Dropping either the fans speeds "marginally" or changing the pump to quiet mode results in the coolant temp increasing by 1c - 2c; On idle over a period of 10 or so minutes

Yeah, that is about right and not conclusive either way. Sometimes on a flow restriction, the lower pump speed will reveal a large change in temperature, suggesting flow rate is an issue. 1-2C does not indicate that.

For comparison, my 5930K@4.5/1.275 and with a cache OC would chew up 205-225W at full load. With a 280mm radiator (X62 or H110), my coolant delta is +6C with the fans only at 1200-1300 on a full bore stress test. In a similar 25C Summer environment, my peak stress test coolant temps are the same as your idle. Granted 10 cores is more load, but not at idle.

Putting the X62 on and temps drop quickly was enough to nail it down for me. I am not sure what the opposing view point would be.

My 7900X was delidded by Silicon Lottery in the US... Done by professionals and not by me :) I'm pretty happy with the CPU temps. My old 6900K used to hit 72c under load. The 7900X with the liquid metal is still only hitting 70c under load. Both CPUs were set to 4.2ghz

That 1 - 2c I mentioned in my previous post was after 10 minutes. Had I waited for 30 minutes I would have reported a 4c increase.

Anyway, I put in the new cooler. Guess what. temps drop much quicker after load drops off. The coolant is -2c on idle after boot from the previous cooler, and +4 on ambient.

After loading the system up, it's cooling back to where it started, and doing it in minutes. Previously it was taking hours to drop and would always end up around 34c - 36c regardless of what the ambient temp near the PC was.

Under load CPU is still hitting 70c, but coolant now is not exceeding 41c, whereas before at times it would approach 45c under certain loads (Eg 2nd pass x264 encodes).

I used to have a 5930 a few years ago and my experiences were similar to yours.

My CPU is @4.2ghz (Not a massive OC), and vcore is 1.24v. On those numbers I would not have expected to see my coolant temps. And so it goes...

Well, Corsair look like they might be considering an RMA. The new cooler is performing as expected and I've now got the commander... I'll connect that up in the next few days and see how that goes.

Thank you for sticking with me through this ordeal :)

Tanya

Anyway, I put in the new cooler. Guess what. temps drop much quicker after load drops off. The coolant is -2c on idle after boot from the previous cooler, and +4 on ambient.

Good, that is more like the expected range.

After loading the system up, it's cooling back to where it started, and doing it in minutes. Previously it was taking hours to drop and would always end up around 34c - 36c regardless of what the ambient temp near the PC was.

That is the smoking gun for me, although the X62 was enough to demonstrate this. Hopefully someone else will see this part clearly. Fortunately, this was still in the early stages and you were not critically affected. These types of issues tend to get progressively worse.