H100i GTX hitting 105c package temps

Even with the pump on high and the fans at 100%, I'm getting thermal throttling almost instantly.

 

The system components are in my sig.

 

Cpu is running 4.2ghz @ 1.3v, when I cut voltage I lose stability quickly.

 

This doesn't seem like what I expected from such a well reviewed water cooler.

the corsair link reports the pump running at 3000rpms is that accurate or could it be misreporting?

What is the H100iGTX coolant temperature?

 

In general it would be easer/possible to help you if you attached a screen shot of the CL4 [Home] tab to your post, so I advise you to do this.

With the system idling

 

cpu 5820k @ 4.2ghz, 1.3v

fans @ max, 2700rpm per link

pump @ max, 3090rpm per link

 

ambient temp: 25c

coolant temp: 37c

cpu cores: 48-56c

package temp: 57c

 

When I had the fans at 1800rpm, water temps went up to 41c.

here's the screen shot with the fans turned down to 2000rpm.

Interesting. You seem to have two separate issues developing. 1) Your water temperature (H100i GTX Temp) is quite a bit above your room temperature. This can be from a weakened pump or other flow restriction, but it can also be from environmental or case heat management issues. 2) Even at the elevated water temperature, your idle CPU temps are way above that water temp. At that menial 3% load, they should bounce around near that 39C water mark. Instead, they are all up in mid 50's. Now perhaps a false peak was created when took the snap shot, so correct me if this is not true.

 

Neither of these issues are related to the fans, so you can set them to something that won't damage your hearing while we sort this out. Make sure your Windows power plan isn't set to Performance or something that will skew the expected core temp values.

 

1) How is the H100i GTX set up in the 550D? I'll assume in the top. Are the fans set to exhaust or intake?

 

2) The very high idle CPU temps above the water temperature are usually caused by a contact issue between the CPU and cold plate. This is the number 1 problem on 115x sockets by far, but it is rare on 2011/3 since you don't need the extra backplate. Make sure you used the right standoffs for the mount. The 2011/3 screws are shorter one end. After that, you only need to screw it down and hand tightening is enough. That leaves the TIM. Did you have any trouble getting it mounted or something else that might have prematurely smeared the thermal paste? When you have a TIM problem, both your idle and load CPU core temps are very jumpy, almost random. Is that like your current situation? Or are they steady, but too high?

Even with the pump on high and the fans at 100%, I'm getting thermal throttling almost instantly.

 

Instant throttling usually suggests a contact issue of some sort. Most likely an extremely small gap between the CPU and cold plate. Even if you unplug the pump, it will take it a few minutes to heat up all the water and reach the 90C+ throttle level. I have to think the block didn't go down straight or on right. When you remove it, take a look at the TIM. If it did not spread, this was the issue.

I just went through and remounted the cooler, which I just had done on Friday troubleshooting the situation. I had noticed rising temps maybe 3 months ago and reseated the cooler with good results, but last week it started popping up again.

 

I've attached the pictures to show how the TIM had spread from the last install and my install process for this current time (Cleaning and tim application) as well as the screw I'm using. As with the install on Friday I seem to be immediately getting better temperatures after the re-install with the case on it's side.

 

The TIM is noctua NT-H1.

Also to directly answer the questions

 

1) radiator is top mounted directly to the case, fans are below the radiator set to intake into the case. Normally the case is vertical and I had the side panel removed during testing.

2) cpu temps were not a one off jump, they idle there but I do have my power profile set to high perfomance. I just tried setting it to balanced and it makes a small difference 2-4c range.

The pictures look like a fairly normal distribution of TIM and there is nothing wrong with NT-H1. Obviously, the wrong threads didn't put themselves on last week, but if it had been operating normally, you remount, and then it runs hot, the most likely reason is related to that moment. On the other hand, if you are saying this was already the situation before the remount and/or this comes and goes, I am a bit perplexed.

 

The other side of the coin is voltage and power settings and we might has well look at that before taking another look at the contact issue. Obviously, performance mode gives you 100% CPU up time, minus EIST and Speedstep fighting to drop it back down again. It is easier for me to interpret the idle temps when the power settings will allow the voltage to drop, although on a 5820K I can compensate. If you are not running EIST and speedstep, plus using a fixed voltage at 1.30v, let me know. That might be just enough to explain the higher than water CPU idle -- although I am reaching a little. I think you can lower vcore a bit for 4.2, probably at least to 1.25v, but I don't want to change those things until this is resolved. Regardless, it should not responsible for your idle temps. I ran 1.315 on my 5820K for quite some time. It idles at the water temp, 22-30C, seasonally dependent. Again, that is why this is unusual. If you unplug your pump, the water temp will slowly start to climb. 40-45-50 etc. The CPU idle temps will go right along with it. They don't hover 10-15C above it unless there is another issue or the PC is constantly running something in the background. The 3% load suggests it is not.

 

Ultimately, I think you will want to flip the fans around to exhaust out the top and it may be responsible for your higher than expected water and case temperature. However, that is also something better left until later so we can maintain a consistent set of variables.

 

Go ahead and start a support ticket here describing the issue. You'll need to create an account (it's not the same as the forums). In the event the cooler needs to be replaced, this is a required step for the RMA. I don't know if that is necessary or not yet, but in the interest of a quick resolution, it would be better to get this started during the work week so Corsair can respond.

 

My normal trouble shooting methods rely on the problem being one or the other, and not both. You can check if there is a flow problem in 5 minutes by running a simple stress test and watching the water temp rise and (hopefully) come back down. However, it sounds like even the slightest loads send your core temps into the stratosphere, making that a non-starter. Also, your motherboard temp sensor indicates it also is close to the water temp and thus most of the interior air is also in the upper 30's. This again suggests the heart of the issue is a contact problem of some type. The only other thing that comes to mind is the thread(s) are slipping out of the factory backplate. If you gently wiggle the block when mounted, does it feel tight?

So it's deffinetly not the pump then, water temps do go up and down.

 

The water block feels very secure on the cpu, doesn't wiggle an iota.

 

I tried playing around with some settings, I had to tweak my overclock a little to get it more stable. I was able to drop the voltage to 1.25, but temps still max out. It's not instant, but maybe 20 seconds before it starts throttling.

 

Here's a snapshot of what a stress test looks like.

 

I also opened Ticket Number 6812067.

 

I really appreciate your help and support. I'll absolutely remember the next time I go shopping.

Yikes, and that is back at the stock settings? Just for comparison, on a 5820K and H100i you should get rolling wave CPU core temps in the 55-62C range at 4.2GHz and probably in the low to mid-40's at stock. Everything about this screams contact problem, but I can't see it. The "slip" theory is the only thing I can come up with that would explain the even TIM spread and the shifting performance.

Perhaps it's just a really crappy 5820K

 

With everything pure stock, I do idle right around 35c, with water temps at 31c.

 

Loading the proc up at stock I get load temp of about 52c, with water temps around 32-33c, pump perfomance, fans max.

 

Simply adding voltage to 1.25v (only change), temps jump right up to 70c under load, add clockspeed to 4.0ghz and temps jump to 80c. The full overclock has a nudge to ram voltage (1.36v) and runs ram at 2600mhz, a nudge to cache voltage(+0.15v) and runs cache at 4.0ghz, and a nudge to the system agent voltage (+0.1v).

 

If you'd expect stock to hit low forties and I'm hitting mid fifties, it's either a really ****ty proc or poor heat transfer from the proc to the cooling system.

 

For reference, my first PC was a P5 MMX 200 I oc'd to 250mhz, so I've been at this a while.

My 5820K was no prize winner either, but it should be consistently bad - not sporadic. If it's not contact, then it has to be voltage or loading. The part that sticks out in your last post is the stark difference between stock idle temp and overclock idle. Your idle temp should be closely tied to ambient temperatures, not peak voltage levels. This mid-50's idle is strange. Rather than an over-inflated voltage (all of yours look within range to me), I wonder if we are looking for a nonvoltage BIOS setting relating to processor or load behavior. I don't know the ASRock X99 BIOS in the way I know the Asus, but that may be something to look at after a good night's sleep (at least for me). XMP often trips some variables, although usually related to BCLK and multipliers. Are EIST and Speedstep enabled? Fixed voltage? That might explain the idle, but it does not explain why you rocket to 90C+ when loaded at 1.30v. The last XTU shot is better and if I equalize the water temps to mine it would be close to expected.

I believe speedstep is always enabled.

When overclocking it disables the Cstates, but when reset it has them enabled, with C3 disabled, but C6 enabled.

 

So everything stock, cstates enabled, speed step enabled, only change voltage to 1.25v, we get idle cpu = water temp +7c (37/30, it's getting later so everything is 2-3c cooler now).

 

In my mind it is consistently bad, it's a little hot at idle and way hot under load, it doesn't really bounce, it just shoots up.

Having C-states enabled probably saves you a few degrees on idle, but not what we are looking for. My C-state limit is set at C0/C1 with the deeper phases disabled. That is enough to drop the clocks and voltage to ~0.72v when idle and water temp level idle.

 

Take a look in the BIOS at OC Tweaker Menu ---> scroll way down to CPU Integrated VR Faults and VR Efficiency. Both are probably on AUTO. VR fault management should probably be disabled when overclocking. I imagine the AUTO feature takes care of this and I would expect instability rather than extreme CPU temps. However, the VR Efficiency does control the current and the BIOS may default to some sort of 100% on state when it detects overclocking. In relation to this, what kind of VRM temperatures do you get at the desktop when overclocked? That may give us a clue as what the kind of load the FIVR and CPU are swamping back and forth.

 

What happens if you use one the "factory preset" overclocks? It looks like there are a couple fast settings from 4.0 up, although I am not looking at the BIOS for your exact X99 board.

I know that I had both VR Faults and VR efficiency disabled, currently on auto for stock.

 

How do I monitor VRM temps? I don't see a temp sensor for that in any of the monitoring software I have loaded.

 

 

I'll try the 4.0 factory o/c and report back.

Hmm.. good question. I don't see VRM in the BIOS screen shots. It is a prominent feature on the Asus monitoring and I use AIDA64 which has it as well. I found a review comparing VRM temps on the Extreme 6 and Asus Deluxe, bu they did not specify how the value was obtained.

 

I don't know if it's that critical, but I set my Vcore to 1.40v this morning and Input Voltage to 2.0 to see how it affected idle. Maybe 1-2C difference. This drastic jump from stock to mild overclock is very strange.

 

Perhaps more relevant is in Full Manual mode with all C-states and EIST off, the idle temps were steady 32-34 at a constant 4.5GHz (1.275v) 4.0 cache (1.17v) at 26C water temp. Dropping back to my normal adaptive voltage and C1/EIST put them right back to 26-28C at 26C water temp.

The factory 4.0 overclock uses XMP settings for the ram, my ram wants a 125 bclck for xmp.

 

So I get

4ghz cpu, 3ghz cache, 125 bclk, 3000 DDR4

 

Voltages are all default except cpu is at 1.18v, the two VR features are disabled, every other settings looks to be at auto.

 

With that setup I get similar idle temps water +6c (37/31), pump on perf, fans on custom - 1300rpm.

 

Under load I get water temp +40c (72c/32c) right off the bat pump on perf, fans on max. If I let it run for 15 minutes temps climb a bit still roughly +40c over the water (76/36).

The forced 125 strap likely changes it to a fixed voltage and that might bump idle temps a little.

 

Was this Intel XTU for load?

Yes I'm using XTU to generate the load, I do have AIDA64, OCCT, and prime95 so I could load the proc with a different kind of load.

No, I think we'd better stick will XTU. Aida64 is mild too, but it's about 8-10C higher than XTU. OCCT and prime are in a different category. You may need a substantial bump in Vcore to even run the test for longer than 60 seconds and the heat would be way too much in these circumstances.

 

I don't want to tell you this 5820K is the worst piece of silicon ever pressed, but I am running out of ideas. Has it always been this bad? If it is the CPU, I can see where it would scale vertically with voltage and past a certain point, you would get tremendous jumps at all levels. I was hoping to momentarily create that situation with the 1.40v this morning, but it did not bring back the same kind core temp jump at idle. My current 5930K is a very nice CPU and that is uncharted territory up there, so perhaps if I pushed further, but I am reluctant to do.

 

Do you have a spare air tower or anything else that could be temporarily substituted to see if this carries over to another cooling device? This has been happening for 3 months?

It's always been hot.

 

I think on my initial setup last year (8/15) I was seeing peak load temps right around 90c with 1.35v and I believe idle temps of around 45c, but I was able to get 4.5ghz stable. It seems that the system is either less efficient removing heat, or has started producing more heat as time has gone on.

 

I had assumed pushin it to 1.35v and 4.5ghz 90c load and 45-50c idle were normalish.

 

I don't have tower cooler handy right now to test that.

I agree. 1.35v on a H100i might get you to 90C on some stress tests with the normal room temperature variability. I might even concede the 45C idle at the very end of voltage band. But now it idles in the 50's and goes even higher. With all the information we have, it seems like it has to be on the voltage & CPU/TIM/cold plate side of the cooler. We've looked at voltage. I am rather doubtful the cold plate went through some sort of alchemical change. We have looked at the BIOS settings and nothing stands out, except the results get scale worse with voltage and the idle is always bad.

 

Is it possible the NT-H1 tube went bad? It would explain both poor idle and load temps. It certainly appears to be making full contact, but yet you have the characteristics of poor contact.

I doubt the NT-H1 went bad it's in a syringe always stored with the cap on.

 

I just ordered a scythe mugen 4 and some artic MX-4 paste off amazon to do testing with.

 

Obviously if the scythe outperforms the H100i, then we know hte H100i is having a problem. If it doesn't out perform the water unit, then all signs point to having a particularly ****ty CPU.