[SOLVED] Is my i9-10900k working properly?

[XCUT]

I have this processor hooked up with MSI MEG Z490 ACE and Arctic Liquid Freezer II 360.

The all-core workloads seem to be working fine - 4,9 ghz on each core.

The problem begins with one-core tests, where I'd expect to see 5,2-5,3 Ghz. Most of the time, however, it sits around 4,9 - 5,1 Ghz, and doesn't even hit 5,2 Ghz (Turbo 3.0). 5,3 Ghz sometimes pops up for a second in HWInfo 64 when sitting on desktop. It also very rarely appears when running Cinebench R15 single-core tests -- then it also pops up for no longer than a second. In video games, 5,1 Ghz was the highest I've seen. I even ran Total War Shogun 2, which utilizes only one core, but to no avail. And I'm running below 70C degrees.

I tried different options in Bios, including updating it to the latest version or imposing Intel's power limits, but still nothing. Even on motherboard's default settings, which blatantly ignore Intel's guideline, the 5,2-5,3 Ghz frequencies are extremely rare.

Is this thing running as it's supposed to be? Or is my processor flawed?

Thanks in advance.

 

[--SID--]

What RAM at what speed?

 

[XCUT]

Crucial Ballistix 32gb 3600, 16 cl, 1,35 v. The results are the same both with and without XMP enabled.

 

[Phaaze88]

You most likely have to go into bios and disable all but one core to see those speeds regularly - 2 cores for TBM 3.0.
Even though you're running single core tests, you still have Windows running and utilizing other cores.

Intel's advertised single core turbo and TBM 3.0 may as well be a marketing gimmick. PCs today seldom use only 1-2 cores/threads, save for brief periods.
The 10900K's 'real' operating boost frequency range is 4.9 - 5.1ghz.

I'd say the cpu is operating normally.

 

[SlavfromBulgaria]

You most likely have to go into bios and disable all but one core to see those speeds regularly - 2 cores for TBM 3.0.
Even though you're running single core tests, you still have Windows running and utilizing other cores.

Intel's advertised single core turbo and TBM 3.0 may as well be a marketing gimmick. PCs today seldom use only 1-2 cores/threads, save for brief periods.
The 10900K's 'real' operating boost frequency range is 4.9 - 5.1ghz.

I'd say the cpu is operating normally.

Fact, if you go into specifics you will see that Intel chips can hit their marketed turbo frequency only on 1-2 cores and see how much it can turbo on the rest of the cores.

 

[Phaaze88]

TBM 3.0 was really funky when it was first introduced though the X299 platform. The 3rd party software didn't work on my 7820X most of the time, so I ended up uninstalling it.
With the newer Core-i models, it's a Windows driver - probably still trips up sometimes, but I doubt it's as broken as the former version was.

 

[XCUT]

So, it's working as "intended", huh?

 

[SlavfromBulgaria]

So, it's working as "intended", huh?

Yes, CPU's usually have 2 states - working or dead. I wouldn't be worried of a slight difference in clocks compared to the marketed ones.
My Ryzen 5 1600 AF is marketed to turbo to 3.6GHz, but without an overclock it stays around 3.4GHz all-cores. I have a manual overclock to 3.7GHz on all cores without using more voltage. If you want to push higher clocks you will have to do it manually.

 

[uWebb429]

one-core tests

The problem is that most monitoring apps do not do a great job reporting Intel turbo boost when a CPU is lightly loaded. The recommended monitoring procedure was documented by Intel when the first Core i CPU was released in 2008. Inaccurate monitoring data has convinced almost everyone that Intel Turbo Boost is not working correctly.

To test what is going on, I set my 10850K to use the exact same turbo ratios that a 10900K uses when it is running at default specs. ThrottleStop uses high performance timers within the CPU and closely follows the Intel recommended monitoring method. When idle, ThrottleStop reports the CPU multiplier as high as 52.95.

What does this mean? The CPU must have been using the 53 multiplier almost all of the time. This monitoring method is extremely precise. The 53 multiplier is alive and well.

Some monitoring apps do not do too bad when the CPU is idle but their monitoring methods and results start to become suspect as soon as there is any load on the CPU.

Let's compare HWiNFO to ThrottleStop while a TS Bench stress test is fully loading 2 threads. The CPU has scheduled one thread so it is mostly running on Core 2 and the other thread is running on Core 3.

ThrottleStop reports a multiplier of 52.63 for both cores. This is equivalent to the CPU using the 53 multiplier 81.5% of the time. Any Windows background tasks that need processing will need to wake up a third core. This will force the CPU to switch to the 51 multiplier or less the other 18.5% of the time.

(53.00 X 0.815) + (51.00 X 0.185) = 52.63

The ThrottleStop results make sense. Intel Turbo Boost is working exactly as advertised.

HWiNFO shows that the two cores doing all of the work are using the 50 multiplier. That is not accurate and it is misleading. Anyone looking at HWiNFO multiplier data would be convinced that Intel Turbo Boost does not work correctly.

The other problem is inefficient programs running in the background 24 / 7. These apps will keep cores in the active state. The more active cores, the lower the turbo multiplier.

The C0% data that ThrottleStop shows is a very accurate measure of background activity. When idle, my 10850K is usually at 0.1%. Anything under 0.5% is fairly normal. If you tend to run a lot of apps in the background then I would suggest increasing the Turbo Group Core Count numbers in the BIOS. Instead of

Ratio Limit 53 - Core Count 2
Ratio Limit 51 - Core Count 3

change these to

Ratio Limit 53 - Core Count 3
Ratio Limit 51 - Core Count 4

The other two entries can stay the same. This will encourage your 10900K to use the 53 multiplier more often when lightly loaded. Any 10900K is quite capable of running at this speed.

If you want to see what your CPU multiplier is doing while gaming, turn on the ThrottleStop Log File option. The on screen MHz data that most users go by when gaming is not as accurate as everyone assumes.

The other thing to look into is reducing the CPU voltage. If you are not going to be overclocking, you should be able to offset the CPU voltage by up to -100 mV. You can drop 15°C off your full load temps when you do this. Here is what I need to run Cinebench R20 at 4900 MHz. A 10900K should be able to run reliably while needing less voltage than that. I have seen some 10900K running this speed at 1.35 V and beyond. This much voltage at this speed is not necessary.

 

[XCUT]

Now, that's informative, thanks.

The other thing to look into is reducing the CPU voltage. If you are not going to be overclocking, you should be able to offset the CPU voltage by up to -100 mV. You can drop 15°C off your full load temps when you do this. Here is what I need to run Cinebench R20 at 4900 MHz. A 10900K should be able to run reliably while needing less voltage than that. I have seen some 10900K running this speed at 1.35 V and beyond. This much voltage at this speed is not necessary.

Instead of setting up voltage offset, can I simply choose a lower CPU Lite Load mode? This should also reduce voltage.

I have also one more question. Does installing ThrottleStop means that the program will automatically apply some of its schemes/parameters? I'm asking because I only would like to use it as a monitor program, and I wouldn't like it to override CPU/BIOS settings.

 

[uWebb429]

When you first run ThrottleStop, it will read all of its default settings from the CPU and it will store this information in the ThrottleStop.INI configuration file. However the BIOS setup the CPU, those are the values that ThrottleStop will show.

You should be able to run ThrottleStop 9.3.1 without it making any automatic changes. If you want ThrottleStop to change your CPU turbo ratios or voltages, you have to specifically tell ThrottleStop to do that. If you make any changes in the BIOS, I would suggest removing the ThrottleStop.INI configuration file. This will force ThrottleStop to re-read your CPU state when it starts. Same thing if you make any changes with ThrottleStop. Just delete the ThrottleStop.INI configuration file, reboot, and everything should be back to normal.

Try running HWiNFO and check the VCore voltage when your CPU is loaded. Some MSI boards tend to use a lot of voltage. If you ever want to reduce your CPU temperatures, look into using a negative offset undervolt.

a lower CPU Lite Load mode?

What is a CPU Lite mode? Do yo mean slow your CPU down? Why would you want to do that?

Turn on the ThrottleStop Log File option and go play some Total War Shogun 2 for at least 15 minutes. When you are done testing, exit the game and then exit ThrottleStop so it can finalize your log file. Upload the file or copy and paste the data to www.pastebin.com so I can see how your CPU is running. Hopefully it shows a higher multiplier than what the onscreen data is showing you.

 

[XCUT]

What is a CPU Lite mode? Do yo mean slow your CPU down? Why would you want to do that?

https://forum-en.msi.com/faq/article/cpu-lite-load

CPU Temperature spikes Immediately following Feature Update to Windows - Windows 10 Help Forums

Immediately after installing the latest (big) feature update to Windows 10 1903, my CPU temperatures started to spike. CPU is NOT overclocked CPU is Water Cooled, all indications are that pump is running and fans run Case, fans, radiator are all clean Tem

www.tenforums.com

If I understand correctly, it's an option which defines how much voltage the BIOS will pump to the CPU. By default, my BIOS sets it to Mode 9, which makes i9 10900k to peak sometimes at 1,35V. Now I put it on mode 6, and the highest I've seen was 1,28V. In Assassin's Creed Origins, it was staying around 1,26V on mode 6. I don't think this affects the CPU's speed at all.

Later, I will try this ThrottleStop, and see how Turbo Boost fares.

 

[uWebb429]

Thanks for the CPU Lite Load info. I use a negative offset voltage on my Asus board. The MSI Lite Load feature seems to accomplish the same thing. I do not think this setting will change the CPU speed. It only seems to lower the voltage.

It is worth playing around with this setting, especially if you have a good quality 10900K. The default voltage curves for all Intel CPUs are always on the high side. Intel does this so they can guarantee long term stability. If you are willing to do some stability testing, most Intel CPUs will be fine with up to 100 mV less voltage.

If you use ThrottleStop, post a picture of the FIVR window. It might show what offset voltage the CPU is set to.

 

[XCUT]

I ran ThrottleStop, but I'm confused about the number of cores it shows. It also doesn't specify which core is which.

EDIT: Nevermind the above.

I ran ThrottleStop with Ratio Limit 53 with core count 2 and then with core count 3. The results were as in HWInfo 64 (I also tested without HWInfo enabled). 5,2 or 5,3 Ghz did not show up neither in Cinebench nor in ThrottleStop benchmark, with one-core test in Cinebench and 1 and 2 threads in ThrottleStop benchmark. The 5,3 Ghz shows up most frequently when idle. I'm afraid Phaaze88 was right -- Intel's 5,3 Ghz is more like a gimmick.

Also, ThrottleStop doesn't show any Voffset enabled.

 

[uWebb429]

If you double click on the monitoring table in ThrottleStop, it will show you all 20 threads and the middle column will show you the ID number for each thread. Click on the Mod heading to cycle between Mod and ID.

Intel's 5,3 Ghz is more like a gimmick

I already posted pictures that show Intel Turbo Boost works exactly as advertised.

On the main screen when your computer is idle, what does ThrottleStop report for average C0%? The problem most people have is there are too many tasks running in the background. These tasks will constantly wake up additional cores which reduces the multiplier.

The best thing about a K series CPU is that you can adjust the turbo ratios so they work the way you want them to work. Try increasing Turbo Group 1 so it uses the 53 multiplier when up to 4 cores are active.

Does the FIVR window show Adaptive or Static and is the Voltage slider set to Default or is it set to a fixed value?

 

[XCUT]

During idle, the whole CPU usage is usually below 5%. After some time, GData kick in with its idle scanning, but that only works when the PC is completely idle and when I don't move my mouse curser, so the antivirus is not the problem. Thus, increasing Turbo Core Count to 4 probably won't change much. The TVB is indeed alive, but it works mostly during GData idle scanning or during the first few minutes after the startup. In 2-thread / 1-core benchmarks it shows rarely.

My voltage is set to Auto (adaptive).

 

[uWebb429]

the whole CPU usage is usually below 5%

Do you mean the C0% or what the Task Manager is showing? It takes about 0.1% in the C0 state to run all of Windows when it is setup properly. Some apps have a bad habit of constantly waking up additional cores. Windows 10 is really good at not doing this.

 

[XCUT]

I meant the Task Manager.

 

[uWebb429]

Task Manager

If you are going to use the Task Manager, use the Details tab. The Processes tab that most people look at does not show CPU usage. The data the Processes tab shows is not accurate when a CPU uses a high percentage of turbo boost like the 10900K does.

 

[XCUT]

There is only 99 CPU on System Idle Process. Everything else is 00.