Single core benchmark/stress test using more than 1 core?

G

Guest

Guest
In prime 95, I am trying to simulate my overclock that boosts vcore and MHz by stressing a single core. Instead of %100 on one core and the others doing whatever, my cores are at %50-%70, not a single one close to being maxed out. I also tried this with the single core benchmark in cinebench, same problem. How can I fix this???
 
Solution

What you are seeing is windows spreading out the single thread workload among all the available threads.
This behavior commonly misleads some to think that their cpu limited single threaded game is actually using multiple cores.

You could lock a core to the task, or
in the windows msconfig boot advanced options option specify one core.
You will need to reboot to activate that.


And, I suggest using OCCT...

What you are seeing is windows spreading out the single thread workload among all the available threads.
This behavior commonly misleads some to think that their cpu limited single threaded game is actually using multiple cores.

You could lock a core to the task, or
in the windows msconfig boot advanced options option specify one core.
You will need to reboot to activate that.


And, I suggest using OCCT for testing which uses more common and less stressful instructions than prime95 or IBT
 
Solution
G

Guest

Guest
Also, I am having problems with turbo boost. I know that my first core can do 48 GHz at 1.35 v (i5 4690k). But one a all core load, I top out at 4.5 GHz. So I set these settings in the BIOS.
1-core ratio limit 48
2-core ratio limit 46
3-core ratio limit 45
4-core ratio limit 45
(Voltage settings set)
I know that all my cores can do 45 at even 1.32 v, but as soon as I boot into the OS, I get BSOD from low voltage.
What I'm assuming is that it uses the 48 multiplier on my bad cores (3 and 4), then leading to a crash. Will doing what you said previously also solve this or is the natural chararistics of turbo boost?
 
First of all, how high you can go safely is determined by your luck in getting a good chip.
I think your chip is at least average.

I suggest you reset all settings to default. That includes ram.
Then gradually increase all the multipliers on concert and see how well you can do.
Monitor your vcore. Past 1.30v is no good, and it is the voltage that drives up your temperatures.
Stress test with OCCT. It uses more normal instructions than prime95 and IBT.
It will shut down the test at 85c.
Monitor your vcore.
When you reach your limit, perhaps back off a notch.
How high do you really need to go?

I leave ram at stock voltage. Faster ram takes more voltage, and I found that it impacted the ability to oc a bit.
And faster ram is not all that necessary if you use a discrete graphics card.
Read this on haswell ram scaling:
http://www.anandtech.com/show/7364/memory-scaling-on-haswell

When all is done, implement speedstep and adaptive voltage.
That will let your multiplier and vcore drop when the cpu has little to do.
 
G

Guest

Guest
I have the older version of prime 95 (without AVX instruction), my ram is stock at 1866 MHz, temperature don't pass 70.