Minute Clock Speed Offset [Intel, 8th gen]

[jerrylee22]

I'm not having problems, but I had a technical question which I couldn't articulate with the proper syntax to get any results on the internet. I was wondering if anybody knew exactly why there's inconstant real time clock speeds with processors, and whether this comes down to architecture, voltage, sensors, or software. I've seen a lot of entertainment and professional coverage of processors and overclocking, and these inconsistencies are sometimes mentioned but were never explained.

For example, my Intel i7 8700k is overclocked to 4.8GHz six-core stable. The real time clock speed is often 4800 MHz, but also varies between 4797.6-4802.4 with an average of ~4799.6 MHz. It's a stable fluctuation across the six cores evenly. I'm noticing a similar trend with Uncore clocking, and different sensor programs show the same results. Windows 10 Task Manager is not even close with the clock speed, but I'm not worried about that since literally everything else (like HWiNFO64/CPU-Z/CoreTemp/BIOS) is displaying it correctly.

This doesn't actually matter overall, and the only impact I've seen (and why I noticed it in the first place) is with CoreTemp, since it doesn't use a dynamic clock speed reading, so when you start the program it just reads your static clock speed from when the instance was initialized. This means if it was ~4801.15 MHz at launch, the program will read out 4801.15 MHz an hour later even if it's actively running at 4800mhz.

It's just a little weird, and I was wondering if anybody knew the reason why this happens with processors.

 

[shmoochie]

For one, the sensor are fairly precise and the bus speed is never exactly 100MHz. So a 99.86MHz bus speed running with a 48x multiplier would show as 4793.28. There are probably other factors, but I think that is the main thing you are seeing.

 

[jerrylee22]

For one, the sensor are fairly precise and the bus speed is never exactly 100MHz. So a 99.86MHz bus speed running with a 48x multiplier would show as 4793.28. There are probably other factors, but I think that is the main thing you are seeing.

Your comment and utilizing "bus speed" to articulate my query confirmed to me the simple answer that "it's normal".

Overall, it seems to "simply" be transmission/communication inconsistencies. Seems like first grade now, but it does just have to do with the bus speed in particular, although this still might be perpetuated by a number of factors. (Like your example, it needs to be measured in 100ths or 1000ths to understand its effect). It's still a little confusing how my bus speed will be 100.00 MHz for a decent amount of the time, though, so it's not always (but can be) working perfectly.

(I suppose?) the reason for the inconsistency has to do with the amount of components required to work together in order for the CPU to function as designed. It has to be powered by the VRM, has to transmit via the bus to the Northbridge and by extension it has to utilize the Southbridge/RAM/PCI-e; all of these seem to be necessary to work together in order to transmit data with a certain frequency.

I (apparently) have slightly unrealistic expectations for technology, but it's still very impressive nonetheless. We're dealing with 14/22 nanometer lithography and (undoubtedly) billions of transistors. I figured with such detailed lithography, the inconsistencies would be ironed out (but now I realize it could make them even more likely). Even if the CPU is working perfectly, every other component would also need to be working perfectly in order to produce and transmit a perfectly stable frequency from it.

I guess it just boils down to the cliché "nothing's perfect". I'm obviously reading much further into it than necessary; I just like having a more in-depth comprehension about what my components are doing. Thanks for nudging me in the proper direction!