H100i GTX Water Temp v. Core Temps

N

NExUS1g

Gawd
Joined
Aug 15, 2004
Messages
554
I did some searching, but I couldn't find this anywhere, and I'm a bit concerned about the performance I'm getting from my H100i GTX and wondering if it may be a heat transfer issue (if there even is an issue).


CPU:
i7-5820K @ 4.5 GHz (100x45) @ 1.328v core
Ambient: ~22ºC

At 10% CPU usage (basically idle), I'm reading about 30º C across all cores.
Under stress, the temperatures climb to a (what I consider at least) questionable 75-80º C. Specifically Cores 0, 2 and 4 are hitting 80. The others tend around low-to-mid 70's.

In checking my Corsair Link, it shows the water temp at 32º C. That seems low to me given the core temps.

My concern is that the heat isn't transferring properly from the IHS to the block. Any thoughts on this would be appreciated, even if they are comments that I shouldn't be concerned. Worth noting is that I reseat the water block a few times with no difference in temps (pea dot method).
 
Last edited:
That does seem awfully hot.

Have you verified the pump RPM's are within their appropriate range?
 
There was another guy complaining about a very similar thing with an H100i GTX earlier this week. Do you have a regular heatsink you can use to test with, to see if the problem is the CPU itself or the cooler?

I wonder if maybe there could have been a bad batch of coolers that have the mating surface misshapen, such that it doesn't make very good contact.
 
NExUS1g said:
My concern is that the heat isn't transferring properly from the IHS to the block.
Click to expand...

What is the program you are using to monitor the CPU temps?

I think either you aren't getting proper heat transfer to the block, or one of he temp monitoring programs isn't working correctly (the one for the cpu or the corsair one.) I would lean toward the former since your closed loop cooler should be performing better than that, i would expect it to keep all cores under 70C.
 
Zarathustra[H] said:
That does seem awfully hot.

Have you verified the pump RPM's are within their appropriate range?
Click to expand...

Pump's showing 3K RPM in Speed Fan and Corsair Link. Radiator fans are showing 1.5K RPM.

---

RazorWind: I'll have to see what I have around. Get back to you on that.

---

snyder: To monitor the CPU temps, I'm using Speed Fan and Core Temp. Using CPUz to stress the CPU. There is a minor discrepancy of about 2º C between the two tools (e.g. Core Temp reading 80º C while Speed Fan showing a max 78º C on the chart).
 
You might also check the flatness of the water jacket with a straight edge. Also check the IHS on the CPU itself. It's possible one of them is warped.

Also check to make sure that nothing is preventing the water jacket from actually contacting the IHS. I had a similar problem last week, when I put a heatsink on one of my graphics cards, where it would idle OK, but as soon I tried running a 3D application, the temp would spike to like 100 degrees, and the system would shut down. Turned out, one of the ram heatsinks was getting in the way of the GPU heatsink, and causing it to not quite touch the GPU die all the way on one side. I removed the offending RAM heatsink, and it worked perfectly. The telltale, in my case, was that the TIM goop wasn't even spread all the way across the die.

If you remove the water jacket/pump assembly from your CPU, do you see a pattern of the thermal goop that indicates it's not making contact in one or more places?
 
RazorWind said:
You might also check the flatness of the water jacket with a straight edge. Also check the IHS on the CPU itself. It's possible one of them is warped.

Also check to make sure that nothing is preventing the water jacket from actually contacting the IHS. I had a similar problem last week, when I put a heatsink on one of my graphics cards, where it would idle OK, but as soon I tried running a 3D application, the temp would spike to like 100 degrees, and the system would shut down. Turned out, one of the ram heatsinks was getting in the way of the GPU heatsink, and causing it to not quite touch the GPU die all the way on one side. I removed the offending RAM heatsink, and it worked perfectly. The telltale, in my case, was that the TIM goop wasn't even spread all the way across the die.

If you remove the water jacket/pump assembly from your CPU, do you see a pattern of the thermal goop that indicates it's not making contact in one or more places?
Click to expand...

I will check the flatness of the two surfaces this weekend. Perhaps a good lapping is in order if that's an issue.

The TIM does spread, but certainly not as far out as I've had others spread. This one is about half a cm from the edges all around where others tended to be more like just to the edge or 1mm from it.
 
If you're really using a pea sized drop of the TIM (which is a lot), it should be running out the edges of the mating surface. I usually have to wipe a little off the sides of my IHSes, even if only use a grain of rice's worth.

Something else to check is whether or not you've got the cooler's mounting hardware installed correctly. My recollection, from installing my own H100i, is that it comes with four studs that thread into the motherboard, with a shoulder on them to keep you from over-tightening the nuts. You might check to make sure that these are screwed into the motherboard all the way. If they're not, that would leave a small gap between the mating surfaces, leading to your issue.
 
Alright, so I checked the IHS and the block's mating surface. The water block is pristine. The IHS on the other hand has a rather prominent "hill" in the middle. It's not insane, but it's about 1/4-1/2 of a mm higher than the edges, and it's very uneven all around. I have a feeling that's why the TIM is not spreading all the way to the edges.

Something else to check is whether or not you've got the cooler's mounting hardware installed correctly
Click to expand...

Funnily enough before I even saw your post here, I noticed the offsets were loose when I was checking the flatness. My temps do seem to be lower, but ambient here has changed dramatically since I last tested. It's more like 17º C ambient right now instead of 22. With that being said, tightening down the offsets with a driver, my temps are idling low-to-mid 20's now.

Cores 0, 2 and 4 are still running hotter than the others. I wonder if they're on the same physical side of the die and how that's related to the contact being made due to the hump on the IHS. I'll have to look for a diagram of where they have the die under the IHS (I think I've seen them off to the side, rather than directly in the middle) and see where the least contact would be made.

Under stress, the CPU is getting up to high 60's on the cores 0, 2 and 4. The other three are high 50's, to about 61/62. So I still think there's an issue with efficiency of heat transfer between the die and the water block.

Any other ideas? Would lapping the IHS be recommended at this point?

/edit

Can't seem to find a schematic online that shows the core numbers and where they are in relation to the orientation of the IHS. Anyone know if something like this exists? But what I have found, and what I see from the placement of the cores on the die, I would put money that 0, 2, and 4 are the outermost cores of the die.
 
Last edited:
Shoot, I keep meaning to test my own H100i-equipped machine under load and post the results; I'm sorry about that. When you say it's hitting 50s to 60s under load, are we talking Prime95 synthetic full load, or an actual application of sort, like a game?

For what it's worth, 50s to 60s is the range in which I would probably quit worrying about this. That's not even close to the range where you need to worry about damaging anything.

That out of the way, if your IHS isn't flat, your CPU sounds like it might be a good candidate for delidding, or at least a thorough lapping.
 
Yeah, I mean, in a long term Prime95 my custom loop - while it stays in the high 50's C most of the time will occasionally peak as high as 69-70.

That's with a hexacore Sandy-Bridge-E though, which produces more heat, and it's overclocked to 4.8Ghz with the voltage up at 1.448. Thing warms up a room.

Even with my keeping my water temps below 30C, the block (EK Supremacy EVO, one of the best) just can't remove the heat to the water fast enough to do better.

When I had a H110i GTX with 4x140mm Noctua 2000rpm jet engine fans in push-pull on this system, it would still peak in the 80's.

You only have two thin 120mm radiator slots on the H100i. Don't expect miracles out of it :p
 
RazorWind said:
Shoot, I keep meaning to test my own H100i-equipped machine under load and post the results; I'm sorry about that. When you say it's hitting 50s to 60s under load, are we talking Prime95 synthetic full load, or an actual application of sort, like a game?

For what it's worth, 50s to 60s is the range in which I would probably quit worrying about this. That's not even close to the range where you need to worry about damaging anything.

That out of the way, if your IHS isn't flat, your CPU sounds like it might be a good candidate for delidding, or at least a thorough lapping.
Click to expand...

Civ 5 is hitting mid-to-high 60's. Other games don't go over 59 or so. I thought that 72 was the temp where damage may start occurring. I've seen that number in a couple of locations around, though I can't for my life recall where.

I don't think delidding is something I really want to do as it seems a bit ... violent, but lapping I feel secure doing.


Zarathustra[H] said:
Yeah, I mean, in a long term Prime95 my custom loop - while it stays in the high 50's C most of the time will occasionally peak as high as 69-70.

That's with a hexacore Sandy-Bridge-E though, which produces more heat, and it's overclocked to 4.8Ghz with the voltage up at 1.448. Thing warms up a room.

Even with my keeping my water temps below 30C, the block (EK Supremacy EVO, one of the best) just can't remove the heat to the water fast enough to do better.

When I had a H110i GTX with 4x140mm Noctua 2000rpm jet engine fans in push-pull on this system, it would still peak in the 80's.

You only have two thin 120mm radiator slots on the H100i. Don't expect miracles out of it :p
Click to expand...

Fair enough. So your summation is that it's performing within expected ranges then?
 
NExUS1g said:
Fair enough. So your summation is that it's performing with in expected ranges then?
Click to expand...

Well, I'm not sure, since I've never had either a 5820k (lower power than mine) or a H100 dual 120mm cooler. They don't necessarily stand out as crazy, but they might be a little high.

High 60's does seem a bit high for Civ 5 gameplay though. What's ambient again? Is it at the 22C you mentioned above? Is your GPU maybe sending its heat through the CPU radiator?

If you just do a prime 95 blended test and let it go for a while (maybe 30 min) what max temps do you see?

With Civ 5 there are too many variables to be comparable to others setups. A prime95 run is more repeatable across systems and might tell us more.
 
Alright, so I'm back again, unfortunately. After the last mounting, I was doing really well. 65-67 C tops. Now the temps, doing the same things with the computer that I always have done before, have climbed back up to up to 79 C peaks, and the mins are even up by about 8 C. I'm not sure what's happening here, but it sure is frustrating.
 
Hi, I've got an H100i GTX and a [email protected] 1.280v
At idle (1.50GHz) Average temps are (around) Liquid=30C CPU=50C Individual cores=35-40C (I keep the fans low for desktop around 300-500 push/pull)
I'll run Prime95 later on today to give you an idea of what I get.
off the top of my mind the cores stay around 70-75, gaming I think mid-60s


P.S. DO NOT DELID A 5820K (or 5930K/5960X)
 
ChoGGi said:
P.S. DO NOT DELID A 5820K (or 5930K/5960X)
Click to expand...


Agreed. While I have read that it is possible to delid a Intel -E part, it is considerably more difficult, will probably result in destroying the chip, and even if successful, the benefit - as opposed to non -E parts - is tiny.
 
My 5820K is at 4.5GHz with 1.310v, also with an H100i, and encoding Big Buck Bunny 4k60 to 'High Profile' preset in Handbrake it gets high 60s, with the third core getting around 71-72C. Fans are a pair of Scythe GT AP-45 at full speed, so they're quite a bit stronger than the stock fans. I have a small case, but with a pair of Noctua 80mm fans and a GT AP-14 as case fans pushing in, and the radiator fans pull out through the radiator. GPU radiator is separated by the side wall of my Corsair 240 case, so it shouldn't contribute a lot to heat in the 'CPU side' of the case.
 
ChoGGi said:
Hi, I've got an H100i GTX and a [email protected] 1.280v
At idle (1.50GHz) Average temps are (around) Liquid=30C CPU=50C Individual cores=35-40C (I keep the fans low for desktop around 300-500 push/pull)
I'll run Prime95 later on today to give you an idea of what I get.
off the top of my mind the cores stay around 70-75, gaming I think mid-60s


P.S. DO NOT DELID A 5820K (or 5930K/5960X)
Click to expand...

The really messed up part that I forgot to mention is that this isn't full CPU usage. It's just playing CSGO for about 45 minutes that I'm getting to 80+.

Zarathustra[H] said:
Agreed. While I have read that it is possible to delid a Intel -E part, it is considerably more difficult, will probably result in destroying the chip, and even if successful, the benefit - as opposed to non -E parts - is tiny.
Click to expand...

Yeah, it's my understanding that these chips are using a solder-like material between the IHS and the lid.
 
NExUS1g said:
The really messed up part that I forgot to mention is that this isn't full CPU usage. It's just playing CSGO for about 45 minutes that I'm getting to 80+.



Yeah, it's my understanding that these chips are using a solder-like material between the IHS and the lid.
Click to expand...


Yeah, there is something there that clearly isn't right.

Do you have any other CPU coolers handy you can compare it to? This way you can determine if it is a cooler issue or a system issue.

Could be a motherboard putting more voltage than expected through the CPU, or a rapidly degrading CPU from excessive past overclocking at high temps.

Since you verified the pump speed, I don't think it is the cooler internals. My money is still on that the mount is somehow not clean.

What does the paste contact pattern look like when you unmounted it? Maybe take a pic and show us?
 
NExUS1g said:
Yeah, it's my understanding that these chips are using a solder-like material between the IHS and the lid.
Click to expand...
It's not solder-like, it is a very expensive indium solder with a layer of gold on top of the nickel-plated die. ;)
 
Zarathustra[H] said:
Yeah, there is something there that clearly isn't right.

Do you have any other CPU coolers handy you can compare it to? This way you can determine if it is a cooler issue or a system issue.

Could be a motherboard putting more voltage than expected through the CPU, or a rapidly degrading CPU from excessive past overclocking at high temps.

Since you verified the pump speed, I don't think it is the cooler internals. My money is still on that the mount is somehow not clean.

What does the paste contact pattern look like when you unmounted it? Maybe take a pic and show us?
Click to expand...

I'm going to remount again today, and I'll grab some pics of the spread.
 
My 6800k (stock, no OC yet) hits 66C with Prime95's default test, and pretty much the same in all cores. My H100i GTX is set to Performance in Corsair's software, however for me this means about 2450RPM on the pump and 1600 on the fans.. I tested this remotely while my house is set to 78, so ambient temps were right around 27C. Water temp sustained 37C.

Even in a cooler house and after long gaming sessions I don't see anything over 45C on the CPU. I would lean towards an RMA.
 
Last edited:
That's a 14nm at stock vs a 22nm at 4.5

Turns out I don't have a solid oc in prime95, good thing I never use it :)
converting a 1920x1080 x264 to x265 at the placebo setting in handbrake gives me an average of 72 on the cores vs 75 on the CPU (left it about 45m or so).

In the image that CPU Syn is the sum of the cores, but as you can see I keep the RPMs lowered on the fans for noise (1600RPM on your fans eep!).
That PSU CL and HDD CL are temp sensors from the Corsair commander mini.
They work good as ambient case temps for me, the HDD one is near input air which is why it's lower. The H100iGTX below is the liquid temp.
I don't use Corsair link software though (too buggy and sucks up CPU even while apparently idling).


Edit: I doubt it really matters, but what goop do you use?
Edit2: did an hour or two Witcher3 last night, CPU hovered around 55-60
 
Last edited:
Have you considered getting your hands on a cheap heatsink and see if that fixes the issue. The Cooler Master 212 can be had for like $20, and works fine for the 5820K. You could probably find a used one in the for sale section, or Ebay, for even cheaper.

If this fixes the problem, I'd probably just leave the 212 on there. If it doesn't, you know the problem is internal to the CPU or motherboard.
 
You must log in or register to reply here.
Back
Top