Intel quality control: voltage differences

I have a few T440s units with the i5-4200U. It is interesting how much higher the voltage requirement is in one of them.

800 mhz: .6794 vs .7194
1.5 ghz: .7138 vs .7865
2.3 ghz: ..945 vs .8580

That's a 10% difference in voltage at the same speed. And the difference in power consumption with these voltages is over 20%. It's a pretty astonishing variance between two units of the same model.

 

How was it measured, and were they all with the same load? Running "idle" there can be fluctuations of voltage because of constant "spikes" of activity. Just running at a specific clock speed doesn't inherently mean same voltage. Same clock speed and same load will be the best test, as well as battery life with the same battery. That isn't to say that there isn't some fluctuation, although those do seem a little bit extreme.

 

Sounds interesting, maybe you can get send me one to confirm your results

 

Measured with HWinfo and CPU-Z. Same load. There was fluctuation in both processors at times but it was .005V or less.

I should clarify that the power consumption difference isn't a system level measurement, but calculated for the processor given that it's consumption goes up quadratically with voltage.

The ones I don't keep are going back to Lenovo.

 

Yes this is normal and so long as the TDP is within spec.

 

Were these stock voltages or was this after tweaking and finding the lowest stable voltages at each clock speed?

 

Stock voltage.

 

Try finding the minimum voltages and they should be far more similar.

 

There are small differences in even Intel's fabrication processes. When I used to build desktops, I'd hunt for the chips that ran at lower voltages because they had superior overclocking potential.

I'd only worry about the differences you're seeing if they translate to worse battery life and/or abnormal temperatures; it would be highly surprising if either of those were the case.

 

^ I would agree with this. 0.95 to 0.85 volts may seem like a lot (> 10%) but with process variations, small errors in the voltage read out and in the software itself would be enough to show that amount of variation.

Notice that the voltages don't actually follow a pattern. CPU1 shows lower voltages for two clock speeds than CPU2 and a higher voltage at turbo. If the chip itself had enough defects that it needed larger voltages to maintain a certain clock, then it would require a larger voltage across the board.

 

I wouldn't complain that much about Intel CPU's voltages... With AMD and NVIDIA GPU's, the voltage differences is even worse...