reported One or more cores reaching 100%

[104th_Money]

@Flappie @Special K

I can confirm that this is still happening for me, even with DCS main window open and it is same in game.

[Special K]

vor 2 Stunden schrieb 104th_Money:

@Flappie @Special K

I can confirm that this is still happening for me, even with DCS main window open and it is same in game.

Hey mate, you sent me a DM that it went away with removing GPU Tweak III?

Can you get me a new dcs.log please?

Cheers

K

[Thomas_GER]

I can confirm that I have the same phenomenon with an AMD 9800X3D. But it does not occur everytime I start DCS. However if it occurs then it even shows up when I am just in the mission editor. And when start a mission and jump into a slot it takes forever to get into the cockpit.

Usually when I change the Core Affinity in Process Lasso when this happens - it kind of "unblocks" or speeds up DCS and I find my self in the cockpit immediately.null

dcs.20250308-094914.crash dcs.20250308-094914.dmp dcs.log.old

[The_Nephilim]

7 hours ago, Special K said:

Do you have Python on your PC? If yes, I can send you a small script later and we can see if we figure which ones are the LPE cores.

Well I can always get python if you could can you please send the script..

[MoleUK]

4 hours ago, Thomas_GER said:

I can confirm that I have the same phenomenon with an AMD 9800X3D. But it does not occur everytime I start DCS. However if it occurs then it even shows up when I am just in the mission editor. And when start a mission and jump into a slot it takes forever to get into the cockpit.

Usually when I change the Core Affinity in Process Lasso when this happens - it kind of "unblocks" or speeds up DCS and I find my self in the cockpit immediately.null

dcs.20250308-094914.crash 1.98 kB · 0 downloads dcs.20250308-094914.dmp 198.61 kB · 0 downloads dcs.log.old 2.58 kB · 1 download

Ditto here with a 5800X3D on Win 11. Most of the time DCS starts fine with no core pegged to 100%, but sometimes it does start with only 1 core stuck at 100%. 

Using task manager to disallow the DCS process from using that core then re-allowing it seems to fix it and distribute the load properly afterwards. No idea why it only happen on some startups and not others, I can literally start the .exe 10 times in a row and see it 2 times out of 10 with no changes inbetween.

On some boots it seems to also hop a load between two cores, spiking one CPU to 100% then switching that load entirely to the other core in a neverending hot potato loop. I wouldn't have noticed except for keeping an eye out for the 100% load problem.

I've said before that sometimes mouse movement is causing CPU behavior changes in these instances as well, again only on some boots not others.

Edited March 8 by MoleUK

[MoleUK]

^ From 104thMoney's post a few posts above mine, this is the same 'hot potato' behavior that happens on some boots for me as well. No idea if it's actually a problem or related to the 100% load on 1 CPU issue or not.

Edited March 8 by MoleUK

[MoleUK]

I'll put up some more later if it helps.

The first log/pic here is of a core 100% being locked solely due to mouse movement. I kept the mouse moving in a small circular motion and it kept 1 core pegged to 100%, when mouse movement stopped it simply started passing that 100% load from 1 core to another over and over again, which is what I typically see.

I rarely boot into the state where mousemovent keeps it pegged at 100%, but it's repeatable if I boot the game enough times. Hence why I think that old mouse polling problem might be related at least tangentially.

2nd pic/log is of the more typical behavior I see, where 1 CPU is pegged at 100% at all times but the load simply hops from 1 core to another over and over again, while core #14 stays steady.

If I start DCS enough times, I will get an instance where that 100% load simply stays on 1 core instead of being passed from 1 core to the other hot potato style.

dcs 1 core locked to 100% usage only while mouse is moving.log more hot potato.log

[felipesm]

The problem only disappears when you set affinity to the CPU cores, through the task manager on DCS.exe

[Special K]

vor 1 Stunde schrieb felipesm:

The problem only disappears when you set affinity to the CPU cores, through the task manager on DCS.exe

I have not seen a dcs.log of you yet.

vor 2 Stunden schrieb MoleUK:

I'll put up some more later if it helps.

The first log/pic here is of a core 100% being locked solely due to mouse movement. I kept the mouse moving in a small circular motion and it kept 1 core pegged to 100%, when mouse movement stopped it simply started passing that 100% load from 1 core to another over and over again, which is what I typically see.

I rarely boot into the state where mousemovent keeps it pegged at 100%, but it's repeatable if I boot the game enough times. Hence why I think that old mouse polling problem might be related at least tangentially.

2nd pic/log is of the more typical behavior I see, where 1 CPU is pegged at 100% at all times but the load simply hops from 1 core to another over and over again, while core #14 stays steady.

If I start DCS enough times, I will get an instance where that 100% load simply stays on 1 core instead of being passed from 1 core to the other hot potato style.

dcs 1 core locked to 100% usage only while mouse is moving.log 46.6 kB · 1 Download more hot potato.log 46.6 kB · 1 Download

Ignore the main menu. That is not multi-threaded.

vor 4 Stunden schrieb The_Nephilim:

Well I can always get python if you could can you please send the script..

Try this and let it run. Curious about the results though. It should only work with Core Ultra CPUs. If it does anyway.

 

cores.py

Edited March 8 by Special K

[Thomas_GER]

vor 3 Stunden schrieb felipesm:

The problem only disappears when you set affinity to the CPU cores, through the task manager on DCS.exe

No it is the opposit - at least for me: If I do NOT setup any affinity I see this behavior. Then I jump in and set affinity (where I leave 1 or 2 cores that I do not designate to DCS) and it works. 
Had exactly the same today in a MultiPlayer mission. Tried to slot and it took for eeeeeever. Then changed affinity and bam....I was in the slot.
Please find the log here. I could re-produce this issue 3x now.
It is usally one of the render Cores, according to process lasso it is Core 1 (so not "0" but "1")

 

dcs.log

Edited March 8 by Thomas_GER
Update

[The_Nephilim]

test

Edited March 9 by The_Nephilim

[The_Nephilim]

9 hours ago, Special K said:

Try this and let it run. Curious about the results though. It should only work with Core Ultra CPUs. If it does anyway.

 

cores.py 1.01 kB · 5 downloads

ok it just says there all e cores?? what is this program supposed to do?

[104th_Money]

18 hours ago, Special K said:

Hey mate, you sent me a DM that it went away with removing GPU Tweak III?

Can you get me a new dcs.log please?

Cheers

K

That was for new stutter that appeared after I had messed with the settings in GPU Tweak III, the 100% core issue is still present. Ill send you a new DCS log when i get a chance.

[Special K]

vor 9 Stunden schrieb Thomas_GER:

No it is the opposit - at least for me: If I do NOT setup any affinity I see this behavior. Then I jump in and set affinity (where I leave 1 or 2 cores that I do not designate to DCS) and it works. 
Had exactly the same today in a MultiPlayer mission. Tried to slot and it took for eeeeeever. Then changed affinity and bam....I was in the slot.
Please find the log here. I could re-produce this issue 3x now.
It is usally one of the render Cores, according to process lasso it is Core 1 (so not "0" but "1")

 

dcs.log 173.8 kB · 2 Downloads

Have you disabled SMT? Your CPU is missing 50% of your (logical) cores.

vor einer Stunde schrieb The_Nephilim:

ok it just says there all e cores?? what is this program supposed to do?

Yeah I feared it would do that. 

That's what Intel said how you could figure the LPE cores. Unfortunately it doesn't work. They haven't used any default mechanism that people were using since years, so you need to access the CPU directly, which this program does. Unfortunately the information isn't where they told it should be.

From the looks of it, I can't tell you, which of your E cores are the LPE cores. So can't DCS. The only way is using all E cores as common cores and hope that not many tasks get assigned to the LPE ones. Rendering shouldn't be affected from it.

Edited March 9 by Special K

[Thomas_GER]

vor einer Stunde schrieb Special K:

Have you disabled SMT? Your CPU is missing 50% of your (logical) cores.

Yeah I feared it would do that. 

That's what Intel said how you could figure the LPE cores. Unfortunately it doesn't work. They haven't used any default mechanism that people were using since years, so you need to access the CPU directly, which this program does. Unfortunately the information isn't where they told it should be.

From the looks of it, I can't tell you, which of your E cores are the LPE cores. So can't DCS. The only way is using all E cores as common cores and hope that not many tasks get assigned to the LPE ones. Rendering shouldn't be affected from it.

I double checked - SMT is activated (or on AUTO, there are only 2 options either AUTO or DISABLED)
I will have to dig into deeper cause there is also a X3D Turbo Mode where it is stated that the CCD function gets disabled….whatever this means.

Edited March 9 by Thomas_GER

[Thomas_GER]

Ok….good to know: In the 9800X3D enabling the X3D Turbo Mode DISABLES SMT, although SMT is set to Auto.

I will test now with X3D Turbo Mode OFF

Update:

O - M - G……thank you for pointing me on that. Turning off the Turbo Mode shows now physical and logical cores and the balancing works buttersmooth.

[104th_Money]

@Special K Here is my latest DCS log file.

dcs.log

[Special K]

vor einer Stunde schrieb 104th_Money:

@Special K Here is my latest DCS log file.

dcs.log 496.17 kB · 2 Downloads

Thanks mate. Is this happening during the main menu or during flight?

[DA_FLY]

This problem come to me again as well)

Micro stutters...

[104th_Money]

@Special K It starts as soon as the main menu opens and stays that way even when flying. As soon as I close DCS completely, core usage drops down to normal. Setting affinity or priority does not change the core usage for me either. And at this point I have just about every single recommended setting implemented to stop stuttering, lag and get best fps.

Edited March 9 by 104th_Money

[The_Nephilim]

On 3/4/2025 at 4:14 PM, Special K said:

I can show you an example, how DCS handles the cores. I have an Intel 14900HX here, which is a CPU with 8 P-cores and 16 E-cores:

P-Cores:
PC2 (4): 8, 9, 12, 13
PC1 (12): 0, 1, 2, 3, 4, 5, 6, 7, 10, 11, 14, 15

E-Cores:
EC0: 16-31

This means, I have 2 P-cores (logical cores 8/9, 12/13) in performance class 2 and 6 P-cores (logical cores 0/1, 2/3, 4/5, 6/7, 10/11 and 14/15) in performance class 1.
DCS now assigns my cores as follows:

2025-02-28 15:14:22.797 INFO    EDCORE (Main): common cores: {4, 5, 6, 7, 10, 11, 14, 15}
2025-02-28 15:14:22.797 INFO    EDCORE (Main): render cores: {8, 9, 12, 13, 0, 1, 2, 3}
2025-02-28 15:14:22.797 INFO    EDCORE (Main): IO cores: {16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31}

So - my fastest cores get assigned to the rendering threads, which is correct. The rest of the P-cores are assigned to typical other tasks, where my least performant cores, the E-cores get assigned to IO tasks.

----
Looking at your CPU, it looks like so:

P-Cores:
PC1: 0, 1, 10, 11, 12, 13, 18, 19

E-Cores:
EC1: 2, 3, 4, 5, 6, 7, 8, 9, 14, 15, 16, 17

You do not have different performance classes, which means that your P-cores should be all about the same performance. This is likely, as your CPU is optimized for a more balanced power / thermal distribution. Now, what does DCS do to your CPU?

2025-03-04 19:55:19.543 INFO    EDCORE (Main): common cores: {12, 13, 18, 19}
2025-03-04 19:55:19.543 INFO    EDCORE (Main): render cores: {0, 1, 10, 11}
2025-03-04 19:55:19.543 INFO    EDCORE (Main): IO cores: {2, 3, 4, 5, 6, 7, 8, 9, 14, 15, 16, 17}

Now, from the looks of it, your CPU does not have hyperthreading anymore, which makes your logical and physical cores even. 
So you only use 4 cores for rendering and 4 cores for common tasks, which is a bit of a pity, as you can only run 4 rendering threads in parallel, where my CPU can run 8.

In your former example with E-cores disabled, DCS looked like so:

2025-03-04 03:04:11.354 INFO    EDCORE (Main): common cores: {12, 13, 18, 19}
2025-03-04 03:04:11.354 INFO    EDCORE (Main): render cores: {0, 1, 10, 11}
2025-03-04 03:04:11.354 INFO    EDCORE (Main): IO cores: {}

This means, that your common cores had to do the IO tasks also.

All in all, I am not 100% sure if these new Core Ultra CPUs are really a great idea for gaming. At least, DCS tries to distribute the tasks as good as it goes from the schema that works for all other CPUs.

If you see a high load on the E-cores btw, you might have a lot of IO tasks to be done. Would be interesting what that could be in your case.

Hey I looked into this and these ultra cores do not have LPE cores, they are just straight up E Cores.. here is a source I found:

 

https://en.wikipedia.org/wiki/Arrow_Lake_(microprocessor)

 

So I am guessing I will need to experiment a bit more but for know I am using PL and put the game on All P Cores and I used 8 E Cores for the game. now All E Cores goto 100% and it might slightly run a bit better will need to further test that theory.. but Other then the e Cores all being maxed out DCS Runs fine for me..

 

Now I seen conflicting info on if I should use the E COres for gaming but I will try and dig around intels site and see what they say on the matter.. I was wondering tho does the new version of Win 11 the 24H2 run the E/P Cores better. I also will keep an eye out for improvemnets to the Core Ultra..

[Special K]

vor 7 Stunden schrieb The_Nephilim:

Hey I looked into this and these ultra cores do not have LPE cores, they are just straight up E Cores.. here is a source I found:

 

https://en.wikipedia.org/wiki/Arrow_Lake_(microprocessor)

 

So I am guessing I will need to experiment a bit more but for know I am using PL and put the game on All P Cores and I used 8 E Cores for the game. now All E Cores goto 100% and it might slightly run a bit better will need to further test that theory.. but Other then the e Cores all being maxed out DCS Runs fine for me..

 

Now I seen conflicting info on if I should use the E COres for gaming but I will try and dig around intels site and see what they say on the matter.. I was wondering tho does the new version of Win 11 the 24H2 run the E/P Cores better. I also will keep an eye out for improvemnets to the Core Ultra..

Ah very good, didn't know that. Thought they all go with LPE cores.

Well then it is indeed only the wrong scheduling class assignment. The guys are working on it already.

vor 15 Stunden schrieb DA_FLY:

This problem come to me again as well)

Micro stutters...

Please, always provide a dcs.log. There are multiple reasons why you can stutter. But it is only possible to say something, if we see a log.

[The_Nephilim]

7 hours ago, Special K said:

Ah very good, didn't know that. Thought they all go with LPE cores.

Well then it is indeed only the wrong scheduling class assignment. The guys are working on it already.

Please, always provide a dcs.log. There are multiple reasons why you can stutter. But it is only possible to say something, if we see a log.

OK that is good, I will keep an eye out for the scheduling class fix thank you

[scommander2]

On 2/21/2025 at 4:59 AM, Special K said:

=> For the guys using E-Core CPUs, please make sure that you are on Windows 11 and that core parking is disabled.

A quick question about core parking disabling on Windows 11:  does it mean dcs.exe run on P and E cores with no restriction for the render cores (or, okay with the common cores.  My understanding the IO cores should be in E-cores)? Thanks.

Edited March 12 by scommander2

[Special K]

vor 2 Stunden schrieb scommander2:

A quick question about core parking disabling on Windows 11:  does it mean dcs.exe run on P and E cores with no restriction for the render cores (or, okay with the common cores.  My understanding the IO cores should be in E-cores)? Thanks.

Core parking and core assignment is something different. 
So for a general E-core CPU (at least the old ones), the E-cores are used as IO-cores, yes.
The core parking is more or less something where Windows decides to not use a specific core, which might be right, if you were the Windows scheduler and think you are the boss in town, but with DCS assigning tasks to cores, the scheduler must not do the parking anymore, as this interferes with what DCS does. So you do not want to have the parking enabled.