Turbo boost and overclock in wrf/ems-uems

[dominic]

Hello everyone, I was wondering two things:

1) WRF / EMS / UEMS uses the intel turbo boost? a presser to 3 GHz in turbo (of course if the certificate) must be such as to 3.4 in wrf-ems works?

2) An overclock the cpu, portandaola top positive influence in wrf?

Of course I speak of i5 / i7 processors

[meteoadriatic]

Turbo boost works only if CPU is not fully loaded and it controlled by hardware:

http://www.intel.com/content/www/us/en/ ... ology.html

So, when you use WRF you probably want to use all available power, which means that turbo boost does not work in such condiitions.

But you can overclock it if you want to accept the risks that come with such action.

[dominic]

Hello Ivan, let's take a concrete example. A quad core i5 3750k, base clock with Turbo Boost 3.4 to 3.8 ghz.

If on 4 core I will use only 3 wrf with turbo boost should work right?

In ubuntu using the utility that measures the cpu frequency, I always see a stable 3.6GHz, I wonder whether it is right speech.

As for overclocking believe that if done with sound principles, and using the safe dissipation of heat by liquid cooling and with good ventilation, a esacore i7 can beat a xeon core 10, thanks to higher clock, what do you think?

[meteoadriatic]

Hello Ivan, let's take a concrete example. A quad core i5 3750k, base clock with Turbo Boost 3.4 to 3.8 ghz.

If on 4 core I will use only 3 wrf with turbo boost should work right?

In ubuntu using the utility that measures the cpu frequency, I always see a stable 3.6GHz, I wonder whether it is right speech.

Use i7z software to find out what is going on. Check for multiplier, you will see then when it (and which core) enters boost and for how long.
https://github.com/ajaiantilal/i7z

As for overclocking believe that if done with sound principles, and using the safe dissipation of heat by liquid cooling and with good ventilation, a esacore i7 can beat a xeon core 10, thanks to higher clock, what do you think?

No idea. I believe you can pretty much linearly extrapolate expected performance for frequency percent increase.

[dominic]

Personally tested, made overclocking a processor i7 6800k, dispelled liquid with corsair hydro h115i, up to 4.2 ghz (stable). The performances are there and the increase on the clock of 3.4 GHz on a simulation with two nesting to 24 hours is about 5.6 minutes less, than the classic clock of 3.4GHz