i7-7700K runs hot (100°C) under Blender render load despite H100i v2 cooler

Status
Not open for further replies.

NeverConvex

Commendable
Mar 27, 2017
9
0
1,520
I recently completed my first PC build and have been happily working/gaming/3D modeling/yadda yadda, but I've noticed that when I cycles render an image in Blender my CPU's normal operating temperature leaps near-instantly from 30°C (when browsing/playing League of Legends/playing Netflix or Youtube videos/) to around 96°C. This spike is detected in both the Asus Suite and Speccy, and has me rather worried; I'm avoiding rendering anything in Blender except briefly for use as a load test while I try to figure out how to fix this.

The fans seem to respond (although not as dramatically as I'd expect..) when the temperature spikes: CPU cooler fan jumps from 0 RPM to >1500 RPM, other fans installed near the front of the chassis spike from 1200 RPM to 1500 RPM according to Speccy/Corsair Link; Asus Suite thinks this fan is running at 600 RPM and 900 RPM respectively for some reason, but in any event does detect a spike in RPMs. Corsair Link's readout for the pump RPMs spikes too, from about 1300 RPMs to about 2000 RPMs as soon as I start rendering in Blender. Strangely, Corsair Link's temperature sensor for the H100i V2 never budges, always staying around 27.7°C; this leads me to believe heat isn't transferring properly to the H100i V2.

Current guesses in the order that I plan to pursue them as possible problems/fixes:

- the H100i V2 was a pain in the butt to install and I would not be surprised if it is seated improperly. I may either have put the backplate on in the wrong orientation or may need to mess with the thumb screws that secure it from the front to the top of the CPU (I've read that it may be best to gradually screw the thumb screws in, not tightening at one of them completely before the others?). This is the easiest 'fix' to attempt so I will try this first, probably as soon as I get off of work this Thursday

- I think I smudged some of the thermal compound that came preapplied on my H100i V2 CPU cooler when I was clumsily smashing my way through trying to figure out how to build a computer. I should have returned to this before now but I had some other issues (bent mobo pins) that took a bunch of headaches and time to fix and completely forgot about this since temperatures seem fine for most of what I do. I've ordered a tube of Dow Corning 340 Silicone Heat Sink Compound Lubricant Grease and some rubbing alcohol and am planning to remove the thermal compound remaining on the H100i V2 and replace it. From what I've read online it doesn't seem like this is likely to be sufficient to explain the dramatic overheating I'm seeing under load but I figure it's worth trying

- ??? other things. I guess I'm hoping through this thread to solicit feedback on whether I'm approaching this correctly, whether I've forgotten/overlooked/am unaware of other possible causes, if any of my guesswork here seems completely wrongheaded, etc.

Things I have not yet checked:
- no idea if the H100i V2 radiator ever gets hot. I guess I could let Blender render for a minute or two and see if I feel any heat reaching the radiator, though I'm hesitant to let the rendering continue for more than a handful of seconds given how quickly it seems to increase the CPU temp

- haven't actually felt any of the components with my hands to see if I can feel the temperature or RPM changes; this could be used to rule out faulty sensors or something I guess but given the behavior is consistent in Speccy/Asus Suite/Corsair Link and that it reacts systematically, reliably, and understandably to Blender rendering I'm not feeling very skeptical of the sensors

Random things I've checked:
- I thought maybe rendering to a 4K resolution monitor was a key part of the problem, so I tried rendering on my second (27") monitor, but got exactly the same temperature spikes, although the render finished much more quickly & seamlessly

Things that confuse me:
- switching Blender's rendering to GPU compute doesn't alter the rapid spike in CPU temperature when rendering at all. I'm not familiar with the details of Blender's rendering algorithms, but it seems very strange that GPU compute wouldn't relieve the CPU of substantial load?

System components of primary interest:
Intel Core i7-7700K Kaby Lake Processor 4.2GHz 8.0GT/s 8MB LGA 1151 CPU w/o Fan
Corsair Hydro Series H100i V2 Extreme Performance Water / Liquid CPU Cooler
ASUS Republic of Gamers Maximus IX Hero Z270 LGA 1151 ATX Motherboard
Corsair Carbide Clear 400C Mid-Tower Case (White)
Hisense 50H8C 50-Inch 4K Ultra HD Smart LED TV (2016 Model) (primary monitor)
Samsung SE330 Series 27-Inch FHD Monitor (S27E330) (secondary monitor)

Other components (don't think any of these are relevant, but I'm a newbie at system builds so including these for completeness' sake):
Toshiba DT01ACA300 SATA III Desktop 3TB Hard Drive
Mushkin ECO3 480GB Solid State Drive MKNSSDE3480GB
GIGABYTE GeForce GTX 1080 XTREME Gaming GV-N1080XTREME-8GD Video Card
SeaSonic Platinum SS-860XP2 860W ATX12V / EPS12V
Kingston ValueRAM 16GB 2133MHz DDR4 Non-ECC CL15 DIMM 2Rx8 Memory

Update: still haven't had time to dig into the PC to try re-seating the cooler, but I was at least able to figure out that there are two separate places in Blender where I need to tell it to use the GPU to render. Now when I GPU render in Blender CPU and GPU temps both spike modestly (increase of ~15°C each) and remain at tolerable max temperatures (~50°C). This solves my immediate problem but there are still plenty of situations where I might need to run the CPU under heavy load, so I've got more problem-solving work to do...
 
Solution


And I thought I had a problem, because I reached 70c in 30% usage. Your case is worse.. So If I were you, I would reinstall the cpu liquid cooler, with your own paste, get a quality one.. If your problem persist, you might have a problem in your cpu.
 
Solution


Haha yeah, I'm rocketing right past 70C, although my CPU load's typically at a fully maxed-out 100% (according to both Task Manager and Corsair Link - didn't look at Asus Suite or Speccy but I imagine they agree) when Blender relies on the CPU alone to render.

Do you have any opinions on good thermal paste to use? I have some Arctic Silver 5 here already but I'd read online that Dow Corning has been performing better recently and was supposedly what was on the H100i V2 originally before I fudged it up, so I thought I'd wait on that to come in and use it.

It's very startling to me that the CPU hits 100C so quickly - it only takes maybe 5 seconds or so after I click render in Blender for both Asus Suite and Speccy to record a temperature jump from 30C to 90C. I guess if the liquid cooler's heat-transfer block isn't touching the top of the CPU then the heat has nowhere to go and this is not unexpected, though?
 
Update: still waiting on my new thermal paste to come in (after reading some more I changed to ordering Shin Etsu, which seems to be the sort that was preapplied anyway), but I tried reseating the CPU cooler. Made sure the thumb screws on the front are tight, and that the backplate doesn't float about after the thumb screws are tightened. No noticeable change in temperature behavior (on standard low-load operations it runs maybe 5C lower, but I also have fewer programs open and both side panels off) - still spikes rapidly to 90C or higher if I try to render in Blender on the CPU.

When the Shin Etsu comes in Friday I'll try that out. A little unsure whether it would be better to apply it directly to the top of the CPU or to the CPU cooler's heat sink (which is where the preapplied stuff was), or if it matters very much.
 

Actually, the one you have is one of the best ones.... What type of case do you have? Does it have good airflow? Are the fan spinning? You are getting extreme high temperature, that is not normal. What about dust? Does it have dust all over it? What about cable management? Check all that.. If no results, then is your cpu the problem.
 
This is my case: Corsair Carbide Clear 400C Mid-Tower Case (White). Here's a stock image of it (no phone/camera to take pictures with right now):

carbide_400c_wht_02.png


My cable management is a bit of a mess and could certainly use some work. Some of the cables sit in the airflow path between the front-mounted radiator fans and the back-panel fan.

All my fans spin, yeah. In particular the fans on the front-mounted radiator and the single fan at the back of the case are always on. The down-facing PSU fan and the GPU fans are not always on, but spin in response to load (well, the GPU fans spin in response to load; I'm just guessing that that's when the PSU fan spins).

This PC's only about 1-1.5 moths old and sits on top of a large table in a carpeted room, so it hasn't built up any noticeable dust yet.

I'm at work until 6 PM EST tonight, but I now have Dow Corning and Arctic Silver thermal paste tubes. The Shin Etsu doesn't arrive until tomorrow but I'm doubtful the specific brand of thermal paste will matter as much as properly applying it, so tonight I'm going to use some rubbing alcohol and a lint-free cloth to strip the old (partially smudged off) thermal paste from my CPU cooler/top of my CPU and reapply either the DC or the AS to the CPU cooler/CPU point of intersection. I'm hoping that fixes the problem.

I may also try max'ing out the fans in BIOS (in PMs you mentioned you had tried this and it had some effect, I think?) if replacing the thermal paste doesn't work.

If all of that fails I'll probably start feeling desperate and call Intel, heh. :pfff:
 
I also have a much larger, full case which would provide better airflow and that I am not currently using - this one: https://smile.amazon.com/gp/product/B00P7TPYDM/ .

I am hesitant to take my PC entirely apart and transition it to a new case, but I guess that's yet another option if nothing else seems to work.
 


xD Yeah, it had some good effects. Doesn't go around 70C anymore while 35/45% usage. While playing animation games, the max C is around 63C. If you can, work in the cable management.. Is one of the most important thing, make sure you have a good airflow, if the problem persist.. Cal Intel, like you said.
 
I'll try to improve my cable management, thanks. I don't think there's much I can do about the CPU cooler's cables (they're very inflexible), but I can probably use ties to reroute most of the rest of my cables so that they avoid interfering with air flow.
 


Let me know, how it went after you apply the thermal paste once again, and doing good cable management. Remember.. Don't put too much, neither too little. Watch a youtube tutorial, if possible.
 
See mate ,today i spent a whole day researching into this.

1)I7 7th gen chips have a bad TIM, dilidding may reduce your max temp in range of 5-25c
2) users noticed that in a matter of fee seconds the cpu reaches idle temp of 50 to 90-100c when 100% loaded.even opening a browser resulted in cpu temp spikes.
3) doesnt matter what cooler you have...it isnt about the cooler...its about the cpu made by intel...its cheap quality..stop investigating the cooler.
4) now the solution to these-
-Undervolt the vcore voltage,as i did to my i7 7700(1.030v),its about trial and error..start from anywhere below but above the minimum value as specified in intel datasheet.
- Disable these: thermal monitor,svid support ,intel speed up technology,turbo boost(you can turn it on after getting a stable underclocked voltage).
-Stop any overclock.
5) there was a mention of smbus modifications due to latest intel chipset driver..find these in device manager and roll back to the oldest driver....this is a temporary fix..


Follow all the above points and tell us again
 
I appreciate the advice, montj. I might try some of that in the future.

However, as it turns out the CPU cooler's layer of thermal compound was, in fact, the problem! My smudging of it when I first opened it, or my constant fumbling with parts while figuring out how to build this PC, or .. something .. clearly ruined it, because right after work today I took apart my PC setup to make it easy to get to, jumped in there, threw another stick of new RAM in for good luck, popped off the CPU cooler, cleaned off it and the CPU of thermal compound using 99% isopropyl alcohol, and applied a 'pea method' or so's worth of Arctic Silver 5 (I know I kept saying I'd use DW or SE but I read that AS5 is easier to apply and.. seeing how many steps I've screwed up so far, that seemed like a good thing, haha).

Plopped the CPU cooler back on, reattached monitors that had been in the way, started her up, and voila! Low 4% load idle temperatures are now 10C or so cooler (26C versus the 30-45C I was getting before), and when I turn on Blender and kick off a 100% CPU load CPU-only render, temperatures now asymptote at 60-65C!

Problem seemingly solved! Woo.
 
Status
Not open for further replies.