Ideal stress test temperatures

CommandoGeek

Honorable
Nov 26, 2013
27
0
10,530
Like the title states, what would be the most ideal temperatures for a CPU when using Prime 95 to stress test? Currently, I'm using the 4770K with the Noctua NH-D14 cooler, and I'm getting these temperatures:
Core 0: 55 C
Core 1: 60 C
Core 2: 62 C
Core 3: 61 C
Most of the time, the temperatures float around ~60C, but the maximum can stretch to 80C, though that doesn't happen often.

The CPU isn't overclocked, but yes, it's turboboosted to 3.8GHz.
 
Solution
Seeing is believing! Big difference between 28.5 and 26.6! More than 20C improvement! :D

Since Standard Ambient is 22C, you can add 3C to those core temperatures. Now you have a thermal benchmark at a steady-state 100% workload. You might want to hang on to those numbers for future reference.

If you want to overclock, you certainly have plenty of headroom to do so. Regardless, your core temperatures are excellent. Core temperatures in the mid-70's are fine for everyday real-world workloads.

As far as your observations are concerned at the beginning of this thread, you were probably running P95 v28.5 "Blend", which is the default test, and is a fluctuating workload for testing memory stability that yields fluctuating temperatures...


I'm curious what part of the stress test that's already running, (quote > "the temperatures float around ~60C"), yields a 20c jump to 80c, for a CPU that's not overclocked that seems a little strange to me?

You say it's Turbo Boosting to 3.8ghz and is Hyper Threading Enabled as well?

 

CommandoGeek

Honorable
Nov 26, 2013
27
0
10,530

I'll read that when I have time ;) Thanks for that


I honestly don't know about that sudden jump. Maybe it was just a random spike, since I was running multiple programs and downloading things at once? And yes, I have 4 cores running 8 threads in total, so hyper-threading is enabled.
 

CommandoGeek

Honorable
Nov 26, 2013
27
0
10,530
So, I just ran Prime95 again, but without any unecessary programs in the background. I got temperatures 60-70, but sometimes it rose to 80 and then cooled down again. Is this a problem, and should I clean out my PC and reapply some thermal paste? Since I'm using a pretty high-end heatsink, I don't think that temperatures should be this high...
 

CompuTronix

Intel Master
Moderator
CommandoGeek,

You probably haven't yet read the Temperature Guide. If you had, then you would've run P95 version 26.6 Small FFT's which is a steady-state 100% workload with steady-state temperatures which do not fluctuate. This is why we use it to benchmark Core temperatures.

What you just described does not match this scenario.

Please read the Guide then try again. Use Real Temp to measure Core temperatures.

Also, we need to know your ambient temperature.

Thanks,

CT :sol:
 

CommandoGeek

Honorable
Nov 26, 2013
27
0
10,530
Okay, running 8 workers in a Small FFT test for approximately 10 minutes and getting these results (all measurements are in Celsius):
Core 0 - 78 | Max 83
Core 1 - 79 | Max 85
Core 2 - 78 | Max 82
Core 3 - 72 | Max 77
Power - 125.8W

As for ambient temperature, it's about 19C here in Hong Kong. But should we take into consideration the fact that the computer is located in the corner of the room?
 


Well I just did, Thought I'd download P95 26.6 and run the 10 minutes for the heck of it.

Ambient 24c (Irrelevant with chilled water cooling but there it is.)

i7-3770K (Not Delidded) overclocked to 5ghz at 1.376v

Max core temperatures reached:
Core 0 = 59c
Core 1 = 64c
Core 2 = 62c
Core 3 = 59c

 


It cannot be compared to straight ambient cooling it is in a league all it's own, it was run with a 12c water temperature and that's 12c below ambient.

I was just curious, not trying to derail the thread. Ry

 

CommandoGeek

Honorable
Nov 26, 2013
27
0
10,530
Version 26.6? :O Then I did it wrong.

I just redid the test and got these results:
Core 0- 56|Max 58
Core 1- 57|Max 58
Core 2- 56|Max 58
Core 3- 51|Max 52
Ambient Temperature- 19

Does this seem about right to you guys? Sorry for my incompetence xD
 

CompuTronix

Intel Master
Moderator
Seeing is believing! Big difference between 28.5 and 26.6! More than 20C improvement! :D

Since Standard Ambient is 22C, you can add 3C to those core temperatures. Now you have a thermal benchmark at a steady-state 100% workload. You might want to hang on to those numbers for future reference.

If you want to overclock, you certainly have plenty of headroom to do so. Regardless, your core temperatures are excellent. Core temperatures in the mid-70's are fine for everyday real-world workloads.

As far as your observations are concerned at the beginning of this thread, you were probably running P95 v28.5 "Blend", which is the default test, and is a fluctuating workload for testing memory stability that yields fluctuating temperatures.

Blend also shifts gears at about 16 minutes into the test, when it becomes less memory intensive, and more processor intensive. This causes the core temperature "peaks" to increase by several degrees.

This is what it looks like:

Shown above from left to right: Small FFT's, Blend, Linpack and Intel Burn Test. The Chart spans 13 minutes.

Notice the steady-state thermal signatures of Small FFT's, which allows accurate measurements of Core temperatures. This is why we use Small FFT's only for thermal testing.

Does that make sense to you?

When you get a chance to do so, please sit down and give the Temp Guide a good read.

Thanks,

CT :sol:
 
Solution