Fx-8350 overheating with a corsair H115i

tbeilbywhite

Commendable
Sep 4, 2016
15
0
1,510
A while ago i had a 6350 with a h55 cooler on it. i started to get the overheating problem where the computer just shuts off. so i bought a 8350 and a H115i for a good upgrade. i ran it at 4.2ghz for half a year and just recently i have started to have the overheating problem again. i have cleaned all the dust from the fans and reapplied thermal paste twice and still nothing

Specs:

CPU: FX-8350
Mobo: CROSSHAIR V FORMULA-Z
Graphics card: 2 R9 380 4gb in crossfire
Ram: corsair vengeance two 2x4g kits
Psu: Rm1000i (yes i know this is stupidly over kill)
 
Solution
Ok starting to get the picture. Your rail voltages are within spec but temps on cache test are higher than normal. What about CPU and FPU.? Just to confirm. Probably will be out of range too.
Please monitor constantly as sustained temps above 80C can damage the CPU.
Go into your Bios and set VCORE to 1.236V and keep frequency as is. All other setting on auto for now.
If the system boots OK then do the test on both CPU and FPU. I want to see your temps at those settings. If the system is unstable then raise VCORE to 1.275V and test again. Just small voltage jumps can make the difference.
BTW what is your ambient room temperature at the moment.
The H115i is more than capable of keeping an FX-8350 at 4.2GHz cool and should be no more than 65C-70C max under load.
Describe how you applied your TIM (thermal paste) and have you determined that the Cooler pump is actually working properly.?
A series of Stress tests will show any deficiencies and indicate temperatures when under load. Are you sure the system shuts down due to heat related issues.? How did you measure this.?
Have you made any changes in Bios.? If so what changes did you make.
 

tbeilbywhite

Commendable
Sep 4, 2016
15
0
1,510


For applying my TIM i used the pea method directly in the center of the cpus heat spreader i then placed the cooler over it and tightened the screws in a cross pattern (top left, then bottom right, then bottom left, then top right).
according to corsair link the pump is running at ~3100rpm which i assume is fine.
I used a program called heavy load to max out my cpu at 100% and while it was running i used HWmonitor to see temps. all other temps were fine apart from the cpu which was reaching 83 degrees c. this was after running the test for 50 seconds at which point i turned off the test because my pc would shut down if i did not.
I had made no changes in the bios when the problem occurred but right now i used the jumper to activate LN2 mode and the switch to turn on slow mode this lowers my clock speeds to 1405mhz and my cpu runs at 40 degrees c under 100% load but the performance hit is just to great for it to be a long term solution.

 


A good quality TIM is important. I use AS5 and still use the pea method. I have an FX 9590 OCed at 5.2GHz with same MB and a H110 and my temps are around 70C under max load. My pump and cooling fans are connected via y connection to CPU_FAN header.
Check the cooling pipes to the radiator. Is the pipe from the CPU to rad hot.? Is the pipe from the radiator back cool.? It should be under load. Are the RAD fans spinning up to around 2k.? reason for checking is that there may be air inside the loop and getting caught in the pump. I've heard of cases like this during benchtesting where they shifted the radiator below the pump and it started to work efficiently when the air bubble was moved. The pump was returned RMA.
If all this checks out OK then only thing I can think off is the CPU overvolting.
Yes switching LN2 switch is no solution, I agree.

Please report back.
 

tbeilbywhite

Commendable
Sep 4, 2016
15
0
1,510


The TIM i use is e2 ic essential (silver)
under load the pipe from the cpu to the rad is hot. And the pipe from the rad to the cpu is cooler than the one from the cpu, but still a bit warm.
The fans that i am using are not the fans from the watercooler they are AF140 Quiet editions they only run up to 1,080rpm the reason i am using those fans is one of the fans that came with the cooler broke and i have a bit of ocd so i replaced both fans even tho only one of them was broken.
i will go into the bios now and tell you the current voltage
 

tbeilbywhite

Commendable
Sep 4, 2016
15
0
1,510
Here are the voltages i am getting with LN2 and slow mode off

IMG_1357_1.jpg
 
I can't comment on e2 ic essential for TIM as I don't know it.
Your voltages are OK in auto mode and I don't think that would change with LN2 switch.
This is getting curious. Hmm.

Download AIDA64 and HWMonitor and put them on your desktop side by side. Run separate tests on the CPU, FPU and cache for 10mins. Stop the test if temps reach 80C. Take screen shots at the 10min mark and post the results. I want to see 100% utilization on all cores and all rail voltages in HWMonitor to check your PSU.
You can use IMIGUR as file host. Obtain the BB url to link here.
 

tbeilbywhite

Commendable
Sep 4, 2016
15
0
1,510


Here is the cpu bench it got to hot very quick
q2pEj5c.png

 

tbeilbywhite

Commendable
Sep 4, 2016
15
0
1,510


Here is the FPU Bench once again very hot very quick

rzNisEV.png

 

tbeilbywhite

Commendable
Sep 4, 2016
15
0
1,510


Here is the cache result. the first time i did this test i did it for to long and my pc shutdown so i lost the results after waiting for my pc to cool down i did them again and the cpu temp started at 40c which was 20c lower then in all of the other benchmarks so the other tests may be inaccurate i can do them again if you need.

HUIamuN.png

 
Ok starting to get the picture. Your rail voltages are within spec but temps on cache test are higher than normal. What about CPU and FPU.? Just to confirm. Probably will be out of range too.
Please monitor constantly as sustained temps above 80C can damage the CPU.
Go into your Bios and set VCORE to 1.236V and keep frequency as is. All other setting on auto for now.
If the system boots OK then do the test on both CPU and FPU. I want to see your temps at those settings. If the system is unstable then raise VCORE to 1.275V and test again. Just small voltage jumps can make the difference.
BTW what is your ambient room temperature at the moment.
 
Solution

tbeilbywhite

Commendable
Sep 4, 2016
15
0
1,510


I am in the uk so ambient temps wont be a problem. here is what i got for the cpu test with 1.275v going to the cpu. i think you could have fixed the problem
i will do the other benches so we can find out then i will play some games. i hope you have
6UgdrNX.png
 

tbeilbywhite

Commendable
Sep 4, 2016
15
0
1,510


i can only do fpu stress tests for very short amounts of time even after waiting for the pc to cool back to 39c within a few secs it reaches 80+

bizo6EF.png

 

tbeilbywhite

Commendable
Sep 4, 2016
15
0
1,510


For consistency i did the cache test again and the results where perfectly stable
so it must be something with the fpu what do you think
EseBpFs.png
 
Yes your CPU test indicates a much better thermal reading at using 1.275V with max 70C for 12mins.
The AIDA64 FPU test is harsh and many report rapid rise in temps when stressing the FPU. I expected this.

I think you may be stable enough at current settings. Now to test this.
There is a Realworld stress tester which uses Realworld applications and not as harsh as the synthetic algorithms.
Download Realbench which is available from ASUS ROG. Just use HWMonitor when you run the test to monitor CORE temperatures. If you can pass that test at current settings and complete the benchmark then you are stable for most games and apps.

If you crash during the test then Go into your Bios and adjust the LLC (Load line calibration) to high and current capability to 120%. This will assist the CPU during heavy load and limit voltage droop.

And again there are further Bios Tweaks you can do depending on results.
BTW: Do not use any software within windows that make Bios changes. ie: AISuite


 

tbeilbywhite

Commendable
Sep 4, 2016
15
0
1,510

thats good news i will download realbench now and report back and no i do not use anything like AISuite
 

tbeilbywhite

Commendable
Sep 4, 2016
15
0
1,510
i just did the real bench stress test the first time i did it my pc overheated in a miniute. Ithen went into the bios and applied the settings you suggested this time it reached 80c in ~3mins which is much better than last time but still too hot i will do the benchmark now which i hope will be successful
any more ideas would be great
 

tbeilbywhite

Commendable
Sep 4, 2016
15
0
1,510
just finished the Benchmark test during the H.264 encoding and the heavy multitasking tests the cpu reached up to 83c however it did not crash do you think these temps are ok
(i scored 84123 dont think you need this but meh)
 
No that's pretty good tbeilbywhite and I get the same temps during the Realbench test.
I did a small mod to gain about another 4-5C drop in temps. I installed a small 80mm fan behind the backplace and glued it on with Aquarium Silicone. Wasn't prepared to drill holes.

UsJW2Ft.jpg


When every little bit counts.
 

tbeilbywhite

Commendable
Sep 4, 2016
15
0
1,510


ah thats great thanks so much for all the help your a genius. i might do that mod to i have a few 90mm fans lying around
is that fan pulling air away or pushing air against the mobo also did you make an air vent in your case just wondering
 

tbeilbywhite

Commendable
Sep 4, 2016
15
0
1,510


i was going to update my bios but i am using my only usb stick for my unraid server so i need to get another
and once again thanks just did a benchmark map in csgo which is the game i play the most and my cpu only reached 59c so thats perfect
 


You can update your bios using an empty USB stick without your system powered up.
Download the Bios file to USB stick and formatted to fat32. Use your rear IO panel and use the Bios flash button.
With the USB stick in the correct Bios flash port, press the flash button till it starts to blink and when it stops your done. Its the best and safest method imo.
 

1991ATServerTower

Distinguished
May 6, 2013
141
4
18,715
It really does sound like a plumbing issue. Air in a water pump will prevent the pump from sucking water. In your case, no moving water, no heat dissipation.

See if your cooler has a manual with trouble shooting for clogs or gas bubbles.

My 8320 doesn't go over 45C at 4.2GHz using a Silverstone air tower when compiling with all 8 cores for 6+ minutes straight. In prime 95 it didn't break 60C after 5 minutes, so you really shouldn't have issues with that super duper cooler you are using.