High temps with stress test's !!

O

Omionous11

Weaksauce
Joined
Mar 30, 2014
Messages
78
-I5 4670k
-Corsair H75 54.0 CFM Liquid CPU Cooler

Hey guy's fairly new to overclocking but I got my system OC to 4.6ghz with a voltage of 1.5 to be stable.

I ran prime95 at first to check stability and I was checking out perfect with an average temp of 40c. I then ran Intel(R) Extreme Tuning Utility and I did a benchmarking with temps as high as 89c and I was stable. However when I ran the stress test under IETU after about 5minutes I hit temps over 100c !!! Is my cooling not good enough ?? Why are two different stress tests giving me such huge gap's in temp's ?? Did I do something wrong ??

I seem to run normal at these setting for my overclocking and my average temp is about 37c
 
1.5V? Is that a typo? You are dumping way too much voltage into the chip to achieve that 4.6 overclock. If you need to use that much voltage for 4.6, you shouldn't be running at 4.6.

Also if you are only hitting 40C in Prime95, you aren't running it correctly or for long enough.
 
Isnt there an issue with running the newest verison of prime95 on haswells? I think its v30 that over stresses haswell and you need v27? Not 100% sure on the verisons. I see that when you ran prime95 your temps were ok, but that might be why there are such huge gaps. I've just heard to be careful with haswell and prime95.
 
Don't use Prime95. Intel has a program out to use for their cpus. Gaming, encoding and normal use will also tell you if it is stable.

P.S. I hope the 1.5 volts was typo also. Way too high for that cpu.

Intel Extreme Tuning App
 
Thanks for the reply's ! I had a great overclock at 4.4 with 1.225 voltage that was sitting at max temps of 75c but I got greedy and wanted more. Every time I benched-marked at 4.6 with IETU and under 1.5v the system would freeze. I guess I'll have to settle for 4.4 :)
 
Omionous11 said:
Thanks for the reply's ! I had a great overclock at 4.4 with 1.225 voltage that was sitting at max temps of 75c but I got greedy and wanted more. Every time I benched-marked at 4.6 with IETU and under 1.5v the system would freeze. I guess I'll have to settle for 4.4 :)
Click to expand...

For what it's worth, 1.5V is way too high of a voltage. Your CPU was almost certainly throttling the clock speed back down to keep the temperature in check, so you weren't really running at 4.6GHz most of the time anyway.
 
Ohhh ok I didn't know it worked like that but thanks !! I'm still learning about OC ! I just like to game :)
 
Newer versions of Prime95 implement AVX2 (FMA).

Haswell applies a voltage offset (increase) when using AVX2, this is why newer versions of Prime95 heavily stress and generate very high temperatures with Haswell CPUs.
 
limitedaccess said:
Newer versions of Prime95 implement AVX2 (FMA).

Haswell applies a voltage offset (increase) when using AVX2, this is why newer versions of Prime95 heavily stress and generate very high temperatures with Haswell CPUs.
Click to expand...

Note that if you go full manual on the voltage control, there is no AVX2-triggered voltage increase to the best of my knowledge. This could be motherboard-dependent, so tread lightly and check carefully rather than taking my word for it. The only downside is that you lose some efficiency at the lower multipliers.

Prime95 will still generate high temperatures at a fixed voltage because it's giving the processor more of a workout. That's why it's an ideal worst-case stress test. The fact that it's an ideal worst-case stress test is why some people choose to not use it. You will encounter AVX2 instructions in other applications, but you probably won't find them used as heavily as Prime95 for a while.
 
Omionous11 said:
This is good to know, I'll update my prime95 and give it a go !
Click to expand...

Just to clarify my statement was merely to give an explanation on why older and newer versions of Prime95 behave differently with Haswell and not an endorsement (or the opposite) of using a newer version of Prime95. It's just an explanation of why you will hear suggestions of using an older version.

Like mentioned above you need to consider if the test is really applicable to your actual workload. But in general (eg. for gaming) you won't be leveraging AVX2 anywhere that heavily, as such is it really important to test with the newer versions of Prime95?
 
You must log in or register to reply here.
Back
Top