4790K stress tests give big differences

[wiebevb]

Hi there,

 

I was running some stress test on my CPU. I noticed that prime95 in "Small FFTs" was giving me much higher temps than tests such as AIDA64 or Intel XTU.
In prime95, a 5 minute stress test resulted 83 degress celcius on the highest core. In XTU around 60 degress celcius on the highest core and in AIDA64 around 69 degress celcius.

 

Are those difference normal, and would you ever reached the prime95 given temperature? I think it's a little bit too high, even in heavy PC usage.

 

By the way, i am using the following components:

 

CPU: i7 4790K (clocked at 4.4 GHz, 1.2 V).

Cooler: Noctua NH-D14 (air)

 

Do someone have a clue?

[SWAGGA]

Hi there,

 

I was running some stress test on my CPU. I noticed that prime95 in "Small FFTs" was giving me much higher temps than tests such as AIDA64 or Intel XTU.

In prime95, a 5 minute stress test resulted 83 degress celcius on the highest core. In XTU around 60 degress celcius on the highest core and in AIDA64 around 69 degress celcius.

 

Are those difference normal, and would you ever reached the prime95 given temperature? I think it's a little bit too high, even in heavy PC usage.

 

By the way, i am using the following components:

 

CPU: i7 4790K (clocked at 4.4 GHz, 1.2 V).

Cooler: Noctua NH-D14 (air)

 

Do someone have a clue? 

thats normal with this program, I still think its important to know that under full load your cpu could become that hot, cuz if you are doing rendering or in general something where you need the horsepower then it will also get very hot

[wiebevb]

thats normal with this program, I still think its important to know that under full load your cpu could become that hot, cuz if you are doing rendering or in general something where you need the horsepower then it will also get very hot

 

I think I am not using the cpu very much for rendering, so I think I will be safe with those temps. By the way, i just runned aida 64 for 5 minutes and I only stressed the FPU and the max temp was around 80 degress, instead of FPU and CPU.

 

Another question, are these temperatures reasonable for a NH-D14 aircooler, or to high?

[basestamp]

It's known that Prime95 will increase core voltages on Haswell chips.

[runit3]

The only thing I have ever gotten as high as P95 temp wise have been using IBT and F@H with 100% CPU+GPU utilization. You will never see those temps in any real world scenario. Intel XTU and AIDA64 give a much better representation of absolute worst case scenarios, and even then you're never going to hit those temps doing something like gaming.

[ThomasD]

If you are running Prime95 in Haswell/Devil's Canyon make sure your voltage is set to manual.  If set to adaptive Prime95 is know to cause substantial overvolting.

 

Many people will advise you to simply avoid Prime95 on the entire Haswell line.

[wiebevb]

It's known that Prime95 will increase core voltages on Haswell chips.

 

Yes, I know, but i am looking directly in monitoring program and I can see that the voltage is steady around 1.200 volts.

 

 

The only thing I have ever gotten as high as P95 temp wise have been using IBT and F@H with 100% CPU+GPU utilization. You will never see those temps in any real world scenario. Intel XTU and AIDA64 give a much better representation of absolute worst case scenarios, and even then you're never going to hit those temps doing something like gaming.

 

That confirmed my assumption.

 

 

If you are running Prime95 in Haswell/Devil's Canyon make sure your voltage is set to manual.  If set to adaptive Prime95 is know to cause substantial overvolting.

 

Many people will advise you to simply avoid Prime95 on the entire Haswell line.

 In aida 64 I see that the voltage is steady around 1.200 V, sometimes 1.208 V, but that can be incorrect readings I think. I don't see any voltage spikes in the graphs, so I think it is safe?

[ThomasD]

 In aida 64 I see that the voltage is steady around 1.200 V, sometimes 1.208 V, but that can be incorrect readings I think. I don't see any voltage spikes in the graphs, so I think it is safe?

Yes, AIDA64 is safe, it is not known to kill Haswell chips like Prime95 is.  Those readings are not incorrect - voltage control is never perfect.

[wiebevb]

Yes, AIDA64 is safe, it is not known to kill Haswell chips like Prime95 is.  Those readings are not incorrect - voltage control is never perfect.

 

I checked the voltage in AIDA64 while I was running the Prime95 stress test

[ThomasD]

I checked the voltage in AIDA64 while I was running the Prime95 stress test

Then I assume you are either running manual voltage, or your UEFI/OC software allows you to set a max voltage limit on adaptive.

 

Either way, so long as Prime95, or whatever stress test you are using, does not drive your voltage above 1.3v you should be safe.

[Queek]

It's known that Prime95 will increase core voltages on Haswell chips.

Not quite.  Any stress testing program will result in increases in voltage in any mode other than manual voltage.  This is why you should use manual voltage while validating overclocks.  See http://linustechtips.com/main/topic/179353-prime95-should-be-just-fine-with-devils-canyon-dmm-testing/

 

 

Hi there,

 

I was running some stress test on my CPU. I noticed that prime95 in "Small FFTs" was giving me much higher temps than tests such as AIDA64 or Intel XTU.

In prime95, a 5 minute stress test resulted 83 degress celcius on the highest core. In XTU around 60 degress celcius on the highest core and in AIDA64 around 69 degress celcius.

 

Are those difference normal, and would you ever reached the prime95 given temperature? I think it's a little bit too high, even in heavy PC usage.

 

By the way, i am using the following components:

 

CPU: i7 4790K (clocked at 4.4 GHz, 1.2 V).

Cooler: Noctua NH-D14 (air)

 

Do someone have a clue?

It is perfectly normal for different stress testing programs to have vastly different temperatures.  Prime95 is particularly hot compared to others (as you have seen).  I saw the same thing when I was testing my 4690k.  

Prime95 is not recommended for testing haswell or DC chips because it doesn't really validate your cpu's ability to run instructions (I still don't fully understand why).  Generally AIDA64 or IXTU are preferred for Haswell or DC overclock validation.  

[wiebevb]

Not quite.  Any stress testing program will result in increases in voltage in any mode other than manual voltage.  This is why you should use manual voltage while validating overclocks.  See http://linustechtips.com/main/topic/179353-prime95-should-be-just-fine-with-devils-canyon-dmm-testing/

 

 

It is perfectly normal for different stress testing programs to have vastly different temperatures.  Prime95 is particularly hot compared to others (as you have seen).  I saw the same thing when I was testing my 4690k.  

Prime95 is not recommended for testing haswell or DC chips because it doesn't really validate your cpu's ability to run instructions (I still don't fully understand why).  Generally AIDA64 or IXTU are preferred for Haswell or DC overclock validation.  

What options are recommended to test in AIDA64, FPU and CPU or only CPU/FPU (in 2 different stress tests). I noticed when I only stressed the FPU, the temps reach much higher (high 70) than when I stressed FPU en CPU together. Which options do I have to select for a best indication?

[Queek]

What options are recommended to test in AIDA64, FPU and CPU or only CPU/FPU (in 2 different stress tests). I noticed when I only stressed the FPU, the temps reach much higher (high 70) than when I stressed FPU en CPU together. Which options do I have to select for a best indication?

I used AIDA64 to validate my 4690k overclock, and I just used the CPU option.  It seems to have worked well, as I haven't had any crashes on manual voltage.  Getting an adaptive overclock with appropriate offset took a few tries.  

[MatadorZA]

As far as i know Prime 95 v26.6 and newer take advantage of the AVX2 instruction set introduced in Haswell CPU's which causes ridiculous temperatures. If you want to P95, use an earlier version than 26.6. I however, do not use P95 anymore. I use IBT as a heat test and then XTU for a stability test.

[wiebevb]

Ok, thanks for all your answers guys