CPU temp spiked within <1 sec of starting Prime95

freakasis85

Honorable
Apr 3, 2014
27
0
10,540
i5-4690k
Corsair H100
ASRock Fatal1ty Z97 Killer

New motherboard/CPU...everything else was working fine in other rig.

I have reapplied thermal paste, despite it looking perfect when I pulled the cooler off the first and second times. I thought it was a bit loose on there so i added some thin washers between the back of the board and the bracket, no change. H100 kept the A10-7700k nice and cool with a high overclock. Pump is working according to the RPM lead and I can feel the vibrations when I touch it. Any Ideas?


Capture_1.png

 

ShadyHamster

Distinguished
That's what normally what happens when you load up prime95, my 4690k shoots up to 60C instantly with a custom loop.
What version of p95 are you using? The recommended version is 26.6 for intel cpus.
I prefer to use AIDA64 stability test, i suggest giving that a go.
 
Make sure you're using prime95 v26.6 with small fft's and try using realtemp or hwinfo64 to see what the temps are. Is that stock or overclocked? If it's the latest version or newer than 26.6 of prime95 it will push temps even higher under unrealistic loads. Do both hoses feel warm (aka to make sure the fluid is in fact circulating). Those temps look more like what I'd expect from an oc'd 4690k. Also make sure in the bios that there's no fan speed control in effect dropping the rpm to the cpu fan header controlling the pump (make sure the pump isn't just running, but running 100%). Hopefully one of those will help bring your temps down.


As ShadyHamster said, the temp changes from idle to load will happen almost instantly, that part is normal. It doesn't 'slowly' build up to temp. When the load is removed and whatever stress test is disabled, the temps should fall right back down again.
 

freakasis85

Honorable
Apr 3, 2014
27
0
10,540
It was Small FFTs and version 28.4 @ 4.4 Ghz. I have never seem temps spike up 50C in less than 1 second before. I confirmed same temps with Speedfan, hw_monitor and the F-tuning software that ASRock provides, they were all the same. I updated the bios and the problem is gone. It was version 1.50 for any Googlers out there and 2.10 is working fine. Max temp after a few min of Prime are 60 C

Also, the H100 gets pump power from molex and only outputs on the RPM wire for monitoring purposes, to allow you to shut down if the pump does fail.

Thanks guys.
 
There's no real value to using Prime 95 on HW / DC CPUs.

1. It puts an unrealistic, yet limited, load on the CPU
2. You can experience unhealthy voltages when overclocking and raising vCore if using adaptive voltage control.
3. You can pass P95 for 72 hours and still fail under RoG Bench.

Temp will always spike because the system is designed to respond to temperature. There is usually a lag built into the response curve so that the system doesn't chase its tail so to speak causing the fans and pump to whir up and down casing temps. Kinda look..."ooh look at that, temps spiked, if it hangs around for more than [insert user's / designer's inputted time interval] , I'll speed things up".
 

freakasis85

Honorable
Apr 3, 2014
27
0
10,540


So, you're saying that it always spikes up that fast, it's just not normally reported that fast? I also do not typically run Prime. I was seeing a spike in temps and wanted a simulation for a quick test, so I used it. Should I have used something else instead? RoG RealBench?

Also, I forgot to re-overclock after the update. There is still a jump, but it is not nearly as significant after the BIOS update.
 
Prime95 isn't bad for thermals but versions after 26.6 aren't designed for the way haswell/d.c. handles it. Other tests can be used for stability like ibt which will often show flaws right away as well.

This sticky at the top of the cpu forum here helps explain it better and why it's used for thermal testing rather than stress testing and why v26.6 is preferred for haswell/d.c.

http://www.tomshardware.com/forum/id-1800828/intel-temperature-guide.html

The reason for the small fft's is it's a consistent load rather than a fluctuating one which is more desired for thermal monitoring.
 

freakasis85

Honorable
Apr 3, 2014
27
0
10,540


I understood the small fft's but I wasn't considering that the makers of Prime95 would have an older version that works better for a new CPU. 26.6 downloaded. TY
 

TRENDING THREADS