Question Hwmonitor Clock speed different than HWinfo effective clock

[jayleonis]

HWmonitor shows under full load (all 64 cores SMT disabled) im running at 3800 ghz on all cores. But effective clock shows 3500. What is the difference? is one more accurate? When I go to activate PBO/overclock which should i use to see my speed on all cores?

CPU 3990x Threadripper
MOBO - Asus Prime trx40 pro s
256 gb ram 3600mhz

 

[jojesa]

Since HWM, shows each core speed, you should use that one to see each core clock.
If you want to see all cores average speed, use HWiNFO

You are seeing two types of data.
HWM shows each clock current (Value), Min and Max speeds.
CPUs do not run at a fix speed and HWM is showing the current speed (Value) around 3817Mhz, but the cores are reaching up to 4341MHzs, with a minimum clock around 2196MHz.

HWiNFO shows the Current, Min and Max clock speeds and an average of all core speeds.

 

[drea.drechsler]

HWmonitor shows under full load (all 64 cores SMT disabled) im running at 3800 ghz on all cores. But effective clock shows 3500. What is the difference? is one more accurate? When I go to activate PBO/overclock which should i use to see my speed on all cores?

CPU 3990x Threadripper
MOBO - Asus Prime trx40 pro s
256 gb ram 3600mhz

Hover over a sensor reading in HWInfo and it will explain the reading to you.

HWInfo's Core Effective Clocks is the average clock calculated across the polling period with respect to sleeping states. As such, it can't be compared directly to HWMonitors Core Clock speed: you should use HWInfo's Core Clock readout instead. That's the current clock calculated from the current core multiplier and bus clock, most likely much like HWMonitor's.

Drag out the core clock sensors in HWInfo to a graph of each core's clock over time and see how it varies so dynamically.

Modern CPU's are extremely dynamic and constantly changing clock speed (core mulitiplier, more precisely) in response to varying processing demands and make frequent drops into deep sleep states to operate with maximum efficiency. Zen processors are particularly so, changing clocks up to 100 times per second. It's impossible for any monitoring utility to capture that activity and not adversely affect performance of the system it's monitoring. Two different utilities running simultaneously are unlikely to sample the clock speed of a CPU at the exact same moment and therefore will almost always report different readings.

The core effective clock reading in HWInfo is the more accurate indicator of how the CPU is actually being used but neither (with one exception) are particularly useful when evaluating CPU performance for establishing PBO settings. The best way is to run benchmarks, probably the best being Cinebench. The exception is when undervolting with Curve Optimizer. If you achieve optimum setting for a core you might find that effective clocks will match core clocks extremely closely when running a certain Prime95 FFT processing load on that core.