7800x3d thread usage issue

[Dragam1337]

This issue could probably be fixed if you just upgrade to Windows 11. Windows 10 doesn't have proper support for the Ryzen 7000 series - while yes it is still a single CCX / CCD, the intricacies of the CPU can still change, requiring a better CPU scheduler, which Windows 11 has. Windows 10 IIRC lacks the improved CPU scheduler of Windows 11. If you try upgrading to Windows 11 your thread usage issue would more than likely go away. Using Windows 10 with new hardware is strongly discouraged because it has older hardware support - its relying on older CPU code and other various stuff that won't be fully optimized for newer hardware, while Windows 11 is actually optimized for the newer hardware.

Just no.

If it was an os issue, then it wouldn't work correctly in any game.

The fact that amd obviously hasn't prioritized to do much work on the win10 drivers is NOT indicative of win10 not being able to use ryzen 7000.

But you would have known that if you had bothered to actually read the thread, rather than just playing the parrot "win10 bad, win11 good, gwoaaar"

Then you are left with OS or hardware, if you eliminate all driver possibility. OS's obviously need to support a CPU in order to function, but mostly its done at hardware.
A simple example is the need to use a certain motherboard bios version in order to use a newer build of a CPU with the same socket.

In the past the issues I have had on various PC's with parking tended to be due to bios, even cases where an OS workaround was needed to bypass the bios.

Dude, read through the thread.

 

[P4-630]

If you have a spare disk, you could at least "try" a windows 11 install for confirmation...

 

[Ferather]

Win11 isn't-wasn't immune to AMD related issues, including 7000 series (see here). It seems AMD added a workaround/OS update to power settings.
If its a case of power configuration, perhaps more can be done with Win10, I am not sure. Good luck though.



@P4-630, good idea.

 

[neatfeatguy]

If you have a spare disk, you could at least "try" a windows 11 install for confirmation...

He doesn't give a shit. He won't try 11 because he's stubborn. If he refuses to follow the next logical step in his troubleshooting process there's nothing else you can do to help nor convince him.

 

[Dragam1337]

Win11 isn't-wasn't immune to AMD related issues, including 7000 series (see here). It seems AMD added a workaround/OS update to power settings.
If its a case of power configuration, perhaps more can be done with Win10, I am not sure. Good luck though.

View attachment 292777

@P4-630, good idea.

That seems very plausible that they need to make that specifically for ryzen 7000 on win10 aswell.

 

[Ferather]

Well I have ADHD and don't do well with concentrating on text for too long, so sorry for not reading everything.

----

@Dragam1337, I just looked through the registry quickly, I suggest you backup Windows so you can fully restore.
You will need to Google what to change, but I guess you might be able to fix the problem (not sure).

Computer\HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Power\PowerSettings

Else try and find a suitable power profile (.pow) file, or make your own?

 

[A&P211]

Newest one. And no offense intended, but you aren't on the same platform (neither is the other guy), so not the same thing at all.

Have you tried windows 11

 

[Ferather]

Maybe the thread name should be amended: '.... looking for fixes for Windows 10, Windows 11 not an option currently.'
Other members with similar issues with Win10 can then discuss it and hopefully find a solution.

 

[Dragam1337]

Well I have ADHD and don't do well with concentrating on text for too long, so sorry for not reading everything.

----

@Dragam1337, I just looked through the registry quickly, I suggest you backup Windows so you can fully restore.
You will need to Google what to change, but I guess you might be able to fix the problem (not sure).

Computer\HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Power\PowerSettings

Else try and find a suitable power profile (.pow) file, or make your own?

Fair enough, it just gets tiresome having to repeat stuff everytime someone new joins the thread, rather than them just reading through the thread themselfs - cause all the details are already here (and this obviously isnt just aimed at you).

Will look into it at a later point - a bit burned out on troubleshooting atm.

 

[Zubasa]

Well i tried the new bios, and it didn't fix the threading - i reckon it's the amd mobo drivers for win10 that is the issue.

In regards to the voltage... with stock memory settings it runs soc voltage at 1.040v and vddio / mc at 1.120v.

When you enable memory profile 1, and voltage settings are set to auto, soc voltage gets increased to 1.270v and vddio / mc to 1.360v... which seems a "little" excessive.

Lowered them manually to 1.140v and 1.230v... so far it runs ok.

Good that you took care of it before things goes wrong.
Strong evidence that the issue is most likely Motherboards feeding too much voltage into the core chiplet, the burnt pin are exactly where the VRM supplies power to the cores and are directly under the core chipet.
Pin analysis of the destroyed Ryzen 7800X3D - All burned pins supply the VDDCR (CPU Core Power Supply) | igor'sLAB (igorslab.de)

 

[Shou Miko]

Sorry for me to be rude but @Dragam1337 stop being like the people around that work that refuses to upgrade because the software ain't optimized for the hardware you have.

I dealt with people when it was windows xp to windows 7, windows 7 to 10 and not that much windows 10 to 11 there are some selected people if you don't like Windows 11 don't upgrade your hardware or go Linux that's the choices you have accept them or live with the issues you encounter.

There is a third options but I doubt you have the money and resources to convince Microsoft to optimize Windows 10 for AMD Ryzen 7xx0X3D series processors.

 

[Dragam1337]

Good that you took care of it before things goes wrong.
Strong evidence that the issue is most likely Motherboards feeding too much voltage into the core chiplet, the burnt pin are exactly where the VRM supplies power to the cores and are directly under the core chipet.
Pin analysis of the destroyed Ryzen 7800X3D - All burned pins supply the VDDCR (CPU Core Power Supply) | igor'sLAB (igorslab.de)
View attachment 292881

Ironically the latest bios update removes the option to touch the vcore... so no minus offset... but the pbo profile 3 keeps it at a max of 1.120v which i would consider very conservative, along with a 70c thermal limit.

I also disabled expo, just in case, until all of this gets resolved.

 

[Aquinus]

Edit : i have noticed that it's always the same 6 threads not getting any usage.

Aside from any hardware issues that you may have, this is the nature of multi-threading for any application. Not all concurrent tasks are made equal and not all of them require the same amount of compute power in a given period of time. Just because an application is using something like 32 or 64 threads doesn't mean that every thread needs to run full tilt all of the time. The result is seeing less than 16 threads worth of usage. What you're seeing is a limitation of the game and the engine's ability to implement concurrent multiprocessing within the game engine. What people don't consider is that a lot of times threads need to either block or be parked because they're waiting on data or the next "tick" for processing. There is likely a single threaded bottleneck somewhere in the engine that limits how far the game can scale to the number of threads available.

From a purely theoretical standpoint, please see Amdahl's Law and obviously, that's an ideal model for this problem. It's likely worse in the real world due to architecture deficiencies with a given piece of software.

Edit: Also keep in mind that switching cores that threads are running on isn't free. Context switching between cores is pretty expensive due to cache coherency (or rather, lack thereof.) Certain threads are probably sticking to certain cores to improve cache locality, so if the hardware isn't fully loaded, the scheduler is going to try and pick the best place for those threads to run.

 

[Dragam1337]

Aside from any hardware issues that you may have, this is the nature of multi-threading for any application. Not all concurrent tasks are made equal and not all of them require the same amount of compute power in a given period of time. Just because an application is using something like 32 or 64 threads doesn't mean that every thread needs to run full tilt all of the time. The result is seeing less than 16 threads worth of usage. What you're seeing is a limitation of the game and the engine's ability to implement concurrent multiprocessing within the game engine. What people don't consider is that a lot of times threads need to either block or be parked because they're waiting on data or the next "tick" for processing. There is likely a single threaded bottleneck somewhere in the engine that limits how far the game can scale to the number of threads available.

From a purely theoretical standpoint, please see Amdahl's Law and obviously, that's an ideal model for this problem. It's likely worse in the real world due to architecture deficiencies with a given piece of software.

Edit: Also keep in mind that switching cores that threads are running on isn't free. Context switching between cores is pretty expensive due to cache coherency (or rather, lack thereof.) Certain threads are probably sticking to certain cores to improve cache locality, so if the hardware isn't fully loaded, the scheduler is going to try and pick the best place for those threads to run.

Another person that obviously didn't bother reading the thread...

 

[Aquinus]

Another person that obviously didn't bother reading the thread...

No, I did. I just don't think you understand the problem. A faster CPU with same concurrency limitations of the game engine will result in less usage, particularly if you're GPU bound. What was your previous CPU?

 

[Dragam1337]

No, I did. I just don't think you understand the problem. A faster CPU with same concurrency limitations of the game engine will result in less usage, particularly if you're GPU bound. What was your previous CPU?

You obviously didn't.

 

[Aquinus]

You obviously didn't.

Instead of being hostile, you could have answered my question instead of being rude about it. I still think that your understanding of the problem is impoverished.

A faster CPU with same concurrency limitations of the game engine will result in less usage, particularly if you're GPU bound. What was your previous CPU?

 

[Dragam1337]

Instead of being hostile, you could have answered my question instead of being rude about it. I still think that your understanding of the problem is impoverished.

As i have written on this very page, im tired of having to answer the same stuff over and over and over, because people are too lazy to read through the thread...

It's an issue with threading in most games on win10 with this cpu, but as shown in for instance post 82, it is fixable... so no, it aint just "the nature of mulit-threading"...

As an fyi to other people posting, im not gonna reply to anything that has already been covered in this thread going forward.

 

[Aquinus]

As i have written on this very page, im tired of having to answer the same stuff over and over and over, because people are too lazy to read through the thread...

It's not on this page and there was only an off-hand reference to an Intel chip a couple pages back, but you definitely didn't say what your previous build was. Stop being rude or nobody is going to want to help you. I'm more than happy to let you figure this out on your own.

By the way, have you tried Windows 11 yet?

 

[Dragam1337]

It's not on this page

Post #134

 

[Aquinus]

Post #134

You mean this one?

Fair enough, it just gets tiresome having to repeat stuff everytime someone new joins the thread, rather than them just reading through the thread themselfs - cause all the details are already here (and this obviously isnt just aimed at you).

Will look into it at a later point - a bit burned out on troubleshooting atm.

Where is the mention of your previous build in that? Stop being a tool, man.

 

[R0H1T]

Well you still haven't established if there's an actual problem with the hardware, even if due to OS/drivers, or just that you want it to run in a (similar) way that you saw with 11th gen Intel. There was very little change between SB to RL so around 8-10 gens of high IPC/clocks & not much changing elsewhere! This is also why Intel has had a consistent run of predictable scores/performance in games/workstation ~ the biggest change came in ADL & then to a much lesser extent RPL. AMD is constantly changing lots of things with Zen, to add to that multiple dies/chiplets/3DV cache et al.

And to add to that every system is different, so even on a fresh windows install if you have different set of hardware you can see different behavior/performance levels.

 

[Aquinus]

Well you still haven't established if there's an actual problem with the hardware, even if due to OS/drivers, or just that you want it to run in a (similar) way that you saw with 11th gen Intel. There was very little change between SB to RL so around 8-10 gens of high IPC/clocks & not much changing elsewhere! This is also why Intel has had a consistent run of predictable scores/performance in games/workstation ~ the biggest change came in ADL & then to a much lesser extent RPL. AMD is constantly changing lots of things with Zen, to add to that multiple dies/chiplets/3DV cache et al.

And to add to that every system is different, so even on a fresh windows install if you have different set of hardware you can see different behavior/performance levels.

People also underestimate how much of a difference that the CPU scheduler can make with heavy concurrent workloads and can widely vary depending on the CPU architecture. This is a case where newer OS kernels can actually help. That's why trying Windows 11 isn't a stupid idea.

 

[the54thvoid]

I'm closing this thread down due to an inflexible OP.