[SOLVED] Why is my CPU clock speed dropping during stress testing?

[George9t8]

Some background, I'm not overclocking, and don't want to do anything risky, I'm just curious about CPU speeds, throttling and limits, and want to understand why my CPU is behaving the way it is.

I have an Asus Z390P motherboard with 32GB RAM clocked at 3200, and just upgraded to an i7-9700KF, which has a base speed of 3.6 and turbo boost up to 4.9. I have a Cooler Master Hyper 212 Evo v2.

I had experimented with some minor adjustments in Intel's XTU and OC Tuner in my BIOS because my clock speed was stuck at 3.9 (I think because that was the max speed of my old CPU), but after restoring BIOS defaults the problem was resolved and I then uninstalled XTU. The only thing I changed in the BIOS is to enable XMP and set fan speeds.

What I'm struggling to understand, and can't find an answer to anywhere, is why when I stress test the CPU the clock speeds drop within a few seconds at temperatures under the CPUs safe limit:

Intel's Extreme Tuning Utility
XTU reports Power Limit Throttling after just a few seconds after package temperature peaks at around 75 degrees, with power reading 125W. Then clock speed drops to 4.2 and power reading reduces to 95W for the duration of the stress test.

Prime95
The clock speeds begin at the safe turbo frequency of 4.6 and all core temperatures reach mid 70s. The speeds very quickly begin to drop, first to 4.2, 3.9, then settle at around 3.7 for a while then back up to 4.1 where it settles, with all core temperature at around 65 after the initial peak. No errors or warnings are reported after 45mins.

Cinibench multicore test
Clock speeds start at 4.7 and temperature quickly reaches 83 for the hottest core, then speeds drop to around 4.25 and temperatures settle at mid 60s for the duration of the test.

GPU stress tests (Heaven, 3DMark Timespy)

Clock speeds stable at 4.6, even during the CPU Test in Timespy, but this stresses the CPU for just a few seconds.

Strangely, when I stop the stress tests and the CPU cores idle, the clock speed increases to 4.78, but that's another issue.

So my question is what's causing the clock speeds to drop even when temperatures at 74 degrees? It doesn't seem like a cooling issue to me, as this isn't that hot. I don't want to mess about with the temperature limits, voltage or power limits as I don't think I should have to.

Can anyone help me understand what's happening and is there anything I can do? Or should I just forget about it and move on with my life?

Thanks in advance.

 

[Eximo]

It isn't throttling for temperature, it is likely throttling for power. You need to go into the BIOS and start changing power limit durations and the power limits themselves.

I believe hardware monitor or hwinfo64 can tell you what is causing the CPU to throttle as well.

Those test are basically power viruses, so you are likely exceeding the TDP of the chip instantly, and then it throttles back to keep that limit at the TDP. Intel specifically excludes XTU as a valid test for power on their CPUs as I recall.

 

[mdd1963]

XTU has settings to easily override the boost power max and short boost power max...

(set both to 'unlimited'...)

It is normal for mainboards/CPUs to throttle down in clock speeds after 30-90 seconds to settle on lower clock speeds to keep overall power draw under control...

 

[TommyTwoTone66]

Intel's Extreme Tuning Utility
XTU reports Power Limit Throttling

Seems like you're power limit throttling. Try increasing your power limits.

I don't want to mess about with the temperature limits, voltage or power limits as I don't think I should have to.

Then you aren't going to achieve high clock speeds under constant load.

The chip is throttling back so it can stay within your motherboard's default, factory-set, conservative power limits. If you want to take full advantage of its built-in overclocking functionality you will need to increase them. If you're happy with "only" 3.7Ghz under extreme loads, then don't bother. I doubt you would ever approach the default power limits running anything other than a synthetic benchmark anyway.

That said, 83C is pretty hot for the Cinebench test for a CPU running at stock speed. To keep under 80C and avoid thermal throttling, even at stock, you're going to need a bigger cooler it looks like. Is it very warm in the room the PC is in?

 

[George9t8]

Thanks all for your responses. I'm getting the impression that this is normal behaviour because of conservative limits imposed, and that throttling would happen with most setups with default BIOS settings. I just wanted to rule out a hardware issue with my VRM or something.

I've fallen into the benchmarking and stress test trap. When doing real world stuff, I don't have an issue. The most CPU intensive application I use is MS Flight Simulator, clock speeds are consistent at 4.6 and temps average 65-75. So I don't feel need to change anything just yet, I'm fine with 4.6GHz speeds, I need a water cooler for the maximum 4.9 anyway.

The only option to remove limits in XTU I could see were for Turbo Boost Short Power Max, disabling that resulted in the speed dropping to 4.4 instead of 4.2.

That said, 83C is pretty hot for the Cinebench test for a CPU running at stock speed

Sorry this was 83c was max clock speed, once it throttled it went back down to mid 60s.

 

[George9t8]

Had a response from Asus which I thought I'd share, as I found something out that I didn't see anywhere online.

After my initial tests above I undid all overclocking settings in the BIOS and just set XMP (technically this too is an overclock, for RAM) and just enabled Turbo Boost and set Power Saving & Performance Mode to ‘Performance Mode’. This kept the clock speed to 4.6GHz throughout a 10min stress test in XTU, and in Prime95 it remained at 4.6GHZ for about 25 minutes before power throttling the CPU to to 4.2GHz, much longer than when it was set to Auto.

I was told that this throttling happens when then board cannot cope with the higher power consumption of the processor, in order to keep things in reasonable limits based on temperature and consumption. With the power setting set to Performance mode, the throttling still happens in order to protect the board components and CPU, as it did before, but is more generous with how much power it allows the CPU and board to use.

The throttling is to keep the components within reasonable temperatures, not the CPU's temperature, but the VRM temperature. This gives me some reassurance there's no issue with my CPU cooler.

When Performance Mode is set to Auto it is following Intel's rules about power, whereas Performance Mode defies those rules and allows more power so the maximum clock speed can be sustained for longer.

No one really mentions this Power Saving & Performance Mode on forums. But I think it's important. If you're using a laptop then you'd probably use Power Saving mode or Auto, otherwise you might as well set it to Performance Mode.