Customize sensor value based on another sensor's readout

ssateneth

Well-Known Member
So I encountered an interesting bug today, not HWiNFO specific. I use the CPU Package and Total DRAM power sensors to estimate how much of a load I'm putting on my equipment and to better determine what load I'm putting on my cooling system.

Well turns out that with, at least my Intel CPU, a haswell-E part, that the power figures are affected in a linear fashion by the base clock that the CPU is running at. They normally run at a base clock of 100MHz. I didn't notice it when I ran 125MHz base clock as I thought it was a normal power increase, but just today I was running a 168MHz base clock and got some rather, well, unreal power figures. After some fiddling and with a kill-a-watt device to measure power at the wall, I concluded that the power sensor readouts are affected by how far off the BCLK is from the default 100MHz, so a 168MHz BCLk will report numbers 68% higher than normal, a 125MHz BCLK will report 25% high numbers, and so on.

So I wanted to know. You already allow customizing the value by adding/subtracting/multiplying the value with arbitrary numbers. But what if I want to do that based on the number returned from another sensor? Would probably also have to allow expressions with the field too, now that I think about it more. It becomes more difficult to think about on how to refer to other sensors or the sensor you are modifying.

Let me get your thoughts on this.
 
I have already heard about this happening, but haven't dealt with it yet.
I think best would be if HWiNFO would automatically do the proper scaling so it reports correct power values.
Can you please attach a Report File+Debug File when running at different BCLK values and then I'll try to implement this into HWiNFO...
 
Alright, sorry about the delay. Here's the goodies.

https://dl.dropboxusercontent.com/u/9768004/BCLK.zip

A note: I could not get the 100MHz BCLK stable with my current RAM settings. They were wet to 2666MHz with CAS13-13-13-30 as opposed to 3000MHz CAS15-15-15-35 with 125MHz and 166.7MHz BCLK. All other speeds and voltages remained the same.
Idle power and Kill-a-watt readings were added in after the fact. Heavy CPU load was Prime95 with 8 threads, 448K FFT, FFT's done in place. Reports were compiled during Prime95 loading. HWInfo was started before Prime95 loading and exited after Prime95 was stopped and exited out.
 
Thank you for the detailed report. I'll fix HWiNFO so that it will automatically adjust the power values based on the actual BCLK strap.
 
Back
Top