Overclock.net banner

1 - 7 of 7 Posts

·
Registered
Joined
·
4 Posts
Discussion Starter · #1 · (Edited)
Ok, so I've been overclocking my 9900k past few days to 5ghz, its alright BUT I noticed that when doing a stress test the effective clock shown in HWINFO stays at 5000mhz for around 10 minutes and then drops and fluctuates from 4,2ghz to 4,95 ghz, never reaching that fixed 5G after the 10 minute mark. I think I tried everything to try and stop that, the only thing that "works" is ThrottleStop app, obviously I don't want to use it all the time I need to set it down somewhere in bios.

I have z390 Aorus master on newest bios with modded faster microcodes F11o from a thread somewhere around this site. My Power Duration 1 and 2 are set to max 4090, time value is also maxed out at 127. I tried raising all the values there with WATT at the end to max, and it didnt help. The only thing I noticed is that at max values it "throttles" after 10 minutes, on default it throttles after around >5 min, so I assume increasing Power durations does something but not enough. Also in XTU or Throttlestop I notice that values for Time don't match the bios. In bios I've set 127 where in both apps it shows at i think 112.

I never noticed a performance downgrade in any game. Is the effective clock actually showing throttling or it is supposed to fluctuate? Also no flags in HWINFO or anyother app like XTU show that im throttling in any way, just the effective clock. How do I stop this?

Temps are in check in check I think. OCCT Large AVX2-70C, Small SSE-85c. VRMs don't go above 50C. I dont even have to use a heavy stress test. Its enough that I play some games for more than 10 minutes and when I run a stress test to check if it goes to fixed 5000mhz it doesn't. Temps in games dont go above 60.

Also I use adaptive mode OC, setting manual voltage doesn't help. I have MCE, VT-d, Voltage optimization disabled and pretty much everything else on left on auto. Tried with all settings on auto doesn't help. All stock also "throttles". Is it actually throttling? Anyone had a similar thing happen to him?

Could it be a bug in HWinfo or something? It behaves very strangely on my pc, some times effective clock drops to 0 which is impossible and even random sensors appear out of nowhere which don't read any values and stay at 0 dissapearing after PC restart.
 

·
Registered
Joined
·
999 Posts
The throttling problem you have discovered is real. Users with both 9th and 10th Gen Intel CPUs have reported this throttling problem. About 10 minutes after booting up seems to be typical but it can happen at anytime.

If you are still not sure if this throttling is real or not, run something simple and consistent like Cinebench for a baseline soon after you start up. After you notice this throttling problem, run Cinebench again. You will see a significant drop in your score. This helps confirm that the CPU is throttling internally. None of the usual throttling flags are triggered within the CPU when this happens so HWiNFO does not report anything.

Here is a thread on the LTT forum about this issue.
9600k hwinfo64 effective clock readings inconsistent between OC's

I wrote a little tool to try to troubleshoot this issue. When a CPU is fully loaded, the cores should be spending 100.0% of their time in the C0 state. When throttling starts, that can drop to 90%. This example shows 89.50% in the C0 state but none of the throttling flags within the CPU are being triggered. Running ThrottleStop 9.3 instantly fixes this problem but I have not yet been able to figure out why.

2487584
 

·
Registered
Joined
·
4 Posts
Discussion Starter · #3 · (Edited)
Thank your, I thought I'm alone with this. Tried multiple settings and nothing really solves it other than ThrottleStop. It has to be pretty common right? Even when I overclocked my old 9700k I remember the effective clock fluctuating. Im still going to try a few settings to see if anything stops that. I also thought about doing the Cinebench test to check the performance hit and it does drop, 13000 points in R23 before the throttle and 12500 after, did multiple tries
 

·
Registered
Joined
·
96 Posts
Thank your, I thought I'm alone with this. Tried multiple settings and nothing really solves it other than ThrottleStop. It has to be pretty common right? Even when I overclocked my old 9700k I remember the effective clock fluctuating. Im still going to try a few settings to see if anything stops that. I also thought about doing the Cinebench test to check the performance hit and it does drop, 13000 points in R23 before the throttle and 12500 after, did multiple tries
ThrottleStop is a great solution to this, and a variety of other throttling problems - and it can be set to load with Windows. Just be sure that XTU is not also installed, since it does not work well to have two programs modifying the same registers.
 

·
Registered
Joined
·
4 Posts
Discussion Starter · #5 ·
ThrottleStop is a great solution to this, and a variety of other throttling problems - and it can be set to load with Windows. Just be sure that XTU is not also installed, since it does not work well to have two programs modifying the same registers.
Yeah, I've been using it for past few days, but you know I tried to make it work in BIOS. I think I tested everything, it will just not work I guess. I have XTU installed but I've never used it anyway.
 

·
Registered
Joined
·
999 Posts
I think I tested everything
Other users have done the same. No BIOS adjustment has been able to solve the mystery. I am starting to think that it is a bug at the CPU level. If both 9th and 10th Gen CPUs have this bug then it is likely a more wide spread problem than many people realize. Keeping a close eye on the HWiNFO Effective Clock data is an easy way to spot this problem.
 

·
Registered
Joined
·
4 Posts
Discussion Starter · #7 ·
Other users have done the same. No BIOS adjustment has been able to solve the mystery. I am starting to think that it is a bug at the CPU level. If both 9th and 10th Gen CPUs have this bug then it is likely a more wide spread problem than many people realize. Keeping a close eye on the HWiNFO Effective Clock data is an easy way to spot this problem.
Previously before I read the thread you linked on LTT I had options Disable and Lock Turbo Power Limits checked in FIVR and unchecked Short Power PL2 in TPL but it seems like it's enough to JUST run the program to solve that throttling. Many people are propably not aware of that issue as not everyone even knows what is effective clock or if they should look at it at all.
 
1 - 7 of 7 Posts
Top