Why when “Limited by MainThread” I see in the task manager that CPU cores is not 100% utilized ?

 

 

 

Edited September 20, 20241 yr by SMN204

6.05 GHz on a 10600K??

The sim will always be limited by something. Think of it as the slowest part of the computing/redenring process. It it weren't, your FPS would always increase, ad infinitum.

The Main Thread limitation just means that the processing of the main thread instructions is what is the determining factor in the performance you see. As to why the game doesn't consume 100% of the CPU at all times, I cannot say, but I prefer that to not be the case.

14 minutes ago, Bob Scott said:

6.05 GHz on a 10600K??

No, overclocked to 5.0 😃
I don't know why the number in the manager is 6.05.....

 

Edited September 20, 20241 yr by SMN204

Taskmanager is not the ideal tool to monitor such things, use e.g. MSI Afterburner for this. In addition, with todays multicore CPUs, utilization percentages of the whole CPU doesnt tell you anything meaningful, but utilization of the GPU does. If it is significantly below 95%, you are CPU limited, no matter what the overall CPU utilization percentage tells you. Reason: think of a eight-core CPU, one core at 100%, seven cores at 0%. The overall load is only 12.5%, yet you are fully CPU limited due to the one core (usually the main thread on it) being at 100%.

3 minutes ago, AnkH said:

Taskmanager is not the ideal tool to monitor such things, use e.g. MSI Afterburner for this. In addition, with todays multicore CPUs, utilization percentages of the whole CPU doesnt tell you anything meaningful, but utilization of the GPU does. If it is significantly below 95%, you are CPU limited, no matter what the overall CPU utilization percentage tells you. Reason: think of a eight-core CPU, one core at 100%, seven cores at 0%. The overall load is only 12.5%, yet you are fully CPU limited due to the one core (usually the main thread on it) being at 100%.

The OP shows a screenie of the per-core utilization, and that pic makes it clear that this is not an issue of averaging core ute rates.  It looks like the core running the main thread (core 5) is running in the low 80% range.

My first guess would be memory/IMC bandwidth limitations.

8 minutes ago, AnkH said:

Reason: think of a eight-core CPU, one core at 100%, seven cores at 0%. The overall load is only 12.5%, yet you are fully CPU limited due to the one core (usually the main thread on it) being at 100%.

In that case, it's only the speed of the processor in single-core mode that matters, not multi-core mode?
And a CPU with a higher frequency on one core is preferable to a multi-core CPU with a lower frequency of cores?

 

That's not how CPU usage works. Limited by a core doesn't necessarily mean it's pegged at 100%, because it's also dependent on the work of the other cores finishing their workload, and working in tandem with the RAM and GPU. So you never see a perfect utilization of the CPU unless it's a synthetic / render workload. It's all a ballet of utilization.

4 minutes ago, Bob Scott said:

My first guess would be memory/IMC bandwidth limitations.

I have a DDR4 3200.   32GB.
Is that the reason?

No, memory is NOT the issue here. As I said, your GPU is only at 63%. As I said, this is sufficient to tell that you have a CPU Limit. And yes, better single core performance is still preferable in MSFS. Might change (hope so) with MSFS2024, considering the published system requirements.

20 minutes ago, SMN204 said:

I have a DDR4 3200.   32GB.
Is that the reason?

It's probably a good part of it.  3200 CAS 16 memory is "Joe Average" RAM with ~10ns latency, and the 10600K has a pretty limited cache on top of that.  So your CPU is having to wait for data from the memory subsystem where computers with faster RAM and CPUs with much more cache don't.

The memory speed is how fast the clock ticks, the CAS (column address strobe) latency is a measure of how many clock cycles the CPU has to idle from the time a memory read request is made until the data is available on the data bus.  Faster clock speeds and lower CAS latency means faster access to the data--less CPU cycles spent idling while waiting for data.  The large increase in performance with the AMD X3D CPUs, which have massive L3 cache memory capacity, illustrates how sensitive MSFS actually is to memory bandwidth.

6 minutes ago, Bob Scott said:

It's probably a good part of it.  3200 CAS 16 memory is "Joe Average" RAM with ~10ns latency, and the 10600K has a pretty limited cache on top of that.  So your CPU is having to wait for data from the memory subsystem where CPUs with faster RAM and much more cache don't.

And which memory (DDR4) would be the best choice in my situation?
i5-10600k (5GHz) + MSI MPG Z490

Edited September 20, 20241 yr by SMN204

Forget about this memory discussion please. Fast memory provides you with max. 10% FPS if you are lucky. In the scenario shown here we speak about 2-3 FPS. Besides the fact that 3200MT/s is already above specs for the 10600K IMC...

23 hours ago, AnkH said:

Forget about this memory discussion please. Fast memory provides you with max. 10% FPS if you are lucky. In the scenario shown here we speak about 2-3 FPS. Besides the fact that 3200MT/s is already above specs for the 10600K IMC...

Pretty cheeky of you there, Bucko.

We're not talking about getting a massive boost in performance here.  We're discussing why the CPU is not running closer to 100% on the main core.

I run MSFS on a 10900K at 5 GHz, with 3600 MHz CAS 15 RAM, which is about 17% faster than the OP's 3200/16 RAM, and I have 20MB of shared L3 cache vs the 10600K's 12MB.  With that I see around 90% utilization on the main thread's core in MSFS.

P.S

I have 3 monitors 2560 x 1440, all screenshots are when running 3 monitors....

 

 

Edited September 20, 20241 yr by SMN204