Discussion Regret Intel 13th Gen Build (mini-Rant)

[wyliec2]

After using my new 13th gen build for several weeks, I wish I’d gone with AMD….

For the last 6-7 years I’ve used Hanbrake for video encoding (primarily BD and UHD) to MKV files.

Over this time I have used several processors – Intel 5960X, 10600K, 10700K as well as AMD 3950X and 5950X.

While the 3950X/5950X platform in my basement is the primary encoding platform, I occasionally will run some encodes on my office PC which was previously a 10700K.

With all of these platforms, I’ve been able to do keyboard activities – Word, Excel, browser concurrent with Handbrake encodes – the CPU would run at 90%+ with Handbrake regardless of whether I was using the machine for additional activities or not. My very light keyboard activity would minimally impact the ongoing Handbrake encode.

I recently upgraded the 10700K to an i5-13500 for my daily office type work machine. The 13500 significantly outscores the 10700K in Cinebench R23 with multi-core scores of 20906 to 12795 respectively. The 13500 does much better with Handbrake encodes IF LEFT UNTOUCHED.

The problem arises when I try to open a browser window while an encode is running. The system quickly shifts the encode processes to the E cores (CPU use goes from 90%+ to 45%) and leaves the P cores doing virtually nothing since web browsing isn’t CPU intensive. This is very apparent in Task Manager displaying graphs of the individual CPU threads. Additionally, browsing will periodically freeze for many seconds, again while the P cores are idle (the E cores are at 100%).

I’d been leery of the whole P core/E core architecture for just this reason. The 13500 is great for daily use and for Handbrake when letting the machine run with no other activity. For combined use, the operating characteristics are largely broken compared to all of the other platforms I’ve used. This experience has been using Windows 11 on the 10600K, 10700K, 5950X and 13500 platforms.

 

[white.a.drew]

After using my new 13th gen build for several weeks, I wish I’d gone with AMD….

For the last 6-7 years I’ve used Hanbrake for video encoding (primarily BD and UHD) to MKV files.

Over this time I have used several processors – Intel 5960X, 10600K, 10700K as well as AMD 3950X and 5950X.

While the 3950X/5950X platform in my basement is the primary encoding platform, I occasionally will run some encodes on my office PC which was previously a 10700K.

With all of these platforms, I’ve been able to do keyboard activities – Word, Excel, browser concurrent with Handbrake encodes – the CPU would run at 90%+ with Handbrake regardless of whether I was using the machine for additional activities or not. My very light keyboard activity would minimally impact the ongoing Handbrake encode.

I recently upgraded the 10700K to an i5-13500 for my daily office type work machine. The 13500 significantly outscores the 10700K in Cinebench R23 with multi-core scores of 20906 to 12795 respectively. The 13500 does much better with Handbrake encodes IF LEFT UNTOUCHED.

The problem arises when I try to open a browser window while an encode is running. The system quickly shifts the encode processes to the E cores (CPU use goes from 90%+ to 45%) and leaves the P cores doing virtually nothing since web browsing isn’t CPU intensive. This is very apparent in Task Manager displaying graphs of the individual CPU threads. Additionally, browsing will periodically freeze for many seconds, again while the P cores are idle (the E cores are at 100%).

I’d been leery of the whole P core/E core architecture for just this reason. The 13500 is great for daily use and for Handbrake when letting the machine run with no other activity. For combined use, the operating characteristics are largely broken compared to all of the other platforms I’ve used. This experience has been using Windows 11 on the 10600K, 10700K, 5950X and 13500 platforms.

you also have to take into consideration the factor that all the systems are running in different settings, timings, bios layouts, and if you upgraded the cpu in the system you are having issue with, without clearing all the old cpu's id info out it will cause the new cpu to run improperly. causing the problems you are explain.

 

[Eximo]

Presumably you could set the processor core affinity to change that behavior.

 

[wyliec2]

you also have to take into consideration the factor that all the systems are running in different settings, timings, bios layouts, and if you upgraded the cpu in the system you are having issue with, without clearing all the old cpu's id info out it will cause the new cpu to run improperly. causing the problems you are explain.

10700K and 13500 are different sockets. The upgrade consisted of an entirely new build including mobo, DDR5 RAM and clean install of everything.

 

[wyliec2]

Presumably you could set the processor core affinity to change that behavior.

Thanks for the suggetion but that would essentially hardcode a solution to a situation that used to (prior to P/E cores) work just fine automatically. Such a solution would at times be beneficial and at other times detrimental.

If I knew what I know now, I would have definitely gone AMD...good ol' hindsight!!

 

[white.a.drew]

10700K and 13500 are different sockets. The upgrade consisted of an entirely new build including mobo, DDR5 RAM and clean install of everything.

I know but again like everyone else who builds pc and never thinks ahead of time. Mobos are pre posted to make sure they work before sold, so they have a cpu ran in them that cpu is never cleared from the history prior to sale. So you have to clear the test cpu I'd from the mobo or it will cause the new cpu to read wrong.

 

[white.a.drew]

Thanks for the suggetion but that would essentially hardcode a solution to a situation that used to (prior to P/E cores) work just fine automatically. Such a solution would at times be beneficial and at other times detrimental.

If I knew what I know now, I would have definitely gone AMD...good ol' hindsight!!

To clear a old cpu ID from a mobo, remove the mobo lithium battery with the psu unplugged well both the psu and lithium battery removed from the mobo hold the power button on the case for 30s x2s, after this reseat the battery the psu cable boot to the bios And reconfigure the bios to how you normally run the pc and see if this helps the cpu run right.

 

[TerryLaze]

Thanks for the suggetion but that would essentially hardcode a solution to a situation that used to (prior to P/E cores) work just fine automatically. Such a solution would at times be beneficial and at other times detrimental.

If I knew what I know now, I would have definitely gone AMD...good ol' hindsight!!

Try putting handbrake on high priority and see if it uses all cores, if that makes the encode run normally then you can make that permanent with several methods.

The last two methods here would work, if priority is the issue.

3 Ways to Change Priority Levels for Programs on Windows 10 and Windows 11

Check out these few different ways to change the priority level for an app or program on both Windows 10 and Windows 11.

www.guidingtech.com

 

[wyliec2]

Try putting handbrake on high priority and see if it uses all cores, if that makes the encode run normally then you can make that permanent with several methods.

The last two methods here would work, if priority is the issue.

3 Ways to Change Priority Levels for Programs on Windows 10 and Windows 11

Check out these few different ways to change the priority level for an app or program on both Windows 10 and Windows 11.

www.guidingtech.com

I have considered that but that but concerned it may block the other things I'm doing. Again, until the 13th generation, everything ran fine when performing light foreground activities concurrently with CPU-intensive tasks.

The 13500 has 6 P cores with two threads and 8 single thread E cores for a total of 20 threads.

Without any ancillary activity, Handbrake alone runs with 20 threads in use (observing logical processor activity in Task Manager) - all 20 processor graphs are at/near 100% utilization.

As soon as I open a browser, all 12 P core threads drop to idle with only the 8 E core threads running at 100%.

This is very poor processor utilization compared to how all of the other platforms just "worked".

Fortunately, I have the 5950X which I'm using to post this while concurrently running two Handbrake encodes and ripping a BD with MakeMKV - processor consistently running at 90%+

I used to be able to do this with the 10700K in my office - the 13500 is much faster but doesn't deal with mixed workloads efficiently.

 

[white.a.drew]

I have considered that but that but concerned it may block the other things I'm doing. Again, until the 13th generation, everything ran fine when performing light foreground activities concurrently with CPU-intensive tasks.

The 13500 has 6 P cores with two threads and 8 single thread E cores for a total of 20 threads.

Without any ancillary activity, Handbrake alone runs with 20 threads in use (observing logical processor activity in Task Manager) - all 20 processor graphs are at/near 100% utilization.

As soon as I open a browser, all 12 P core threads drop to idle with only the 8 E core threads running at 100%.

This is very poor processor utilization compared to how all of the other platforms just "worked".

Fortunately, I have the 5950X which I'm using to post this while concurrently running two Handbrake encodes and ripping a BD with MakeMKV - processor consistently running at 90%+

I used to be able to do this with the 10700K in my office - the 13500 is much faster but doesn't deal with mixed workloads efficiently.

This is because you don't have the system setup right for the encoding your trying to do.... the cpu is fine from the sounds of it, your issues lay in figuring out how to properly run the newer type of cpus.

 

[wyliec2]

This is because you don't have the system setup right for the encoding your trying to do.... the cpu is fine from the sounds of it, your issues lay in figuring out how to properly run the newer type of cpus.

I respectfully disagree. I have built and optimized systems both professionally and as a hobby for decades.

I don't think the problem is specifically with the processor but with the architecture which demands that the OS scheduler determine how to shift processes among the asymmetrical cores. Windows 11 was supposed to have improvements in this regard over Windows 10.

At the end of the day, processors without asymmetrical cores would not exhibit this behavior.

 

[Nighthawk117]

I respectfully disagree. I have built and optimized systems both professionally and as a hobby for decades.

I don't think the problem is specifically with the processor but with the architecture which demands that the OS scheduler determine how to shift processes among the asymmetrical cores. Windows 11 was supposed to have improvements in this regard over Windows 10.

At the end of the day, processors without asymmetrical cores would not exhibit this behavior.

If you havn't already done so, go to Task Manager while the Handbrake encode is running, find it on the Details tab. Right click and set priority to High. See if that makes any difference, on my machine Handbrake defaults to a priority of Below Normal. Not sure why, but if it's like that on yours then that may explain why it's being pushed to the E cores.

 

[Nighthawk117]

My bad, I didn't see that @TerryLaze had already suggested it. I would be amazed if this was not your problem:

https://imgur.com/E7QQEG4

View: https://imgur.com/E7QQEG4


By default it's set to Below Normal, I bet this combined with the new thread director is what's causing your problem. Usually applications default to Normal, I would see if it works and if so just change the default behaviour on Tools>>>Preferences>>>Advanced. To be clear I would change it to Normal first and try High if it doesn't work.

 

[white.a.drew]

I respectfully disagree. I have built and optimized systems both professionally and as a hobby for decades.

I don't think the problem is specifically with the processor but with the architecture which demands that the OS scheduler determine how to shift processes among the asymmetrical cores. Windows 11 was supposed to have improvements in this regard over Windows 10.

At the end of the day, processors without asymmetrical cores would not exhibit this behavior.

The more system are changed and the architecture are optimized the less like its Predecessors it becomes optimization for an older system won't be the same as a new one. So you could have been like me and 100s more and built 100's of PC in your past doesn't mean you can properly calibrate a newer system.... not to be rude. I have a lot of issues with newer builds now as well. It takes time to relearn how the system is calibrated for different tasks now. Not everything will be as simple as normal

 

[wyliec2]

My bad, I didn't see that @TerryLaze had already suggested it. I would be amazed if this was not your problem:

https://imgur.com/E7QQEG4

View: https://imgur.com/E7QQEG4


By default it's set to Below Normal, I bet this combined with the new thread director is what's causing your problem. Usually applications default to Normal, I would see if it works and if so just change the default behaviour on Tools>>>Preferences>>>Advanced. To be clear I would change it to Normal first and try High if it doesn't work.

Actually changing it to Normal was one of the first things I tried.

I've always had it set to Below Normal which allowed other activities to take place with Handbrake utilizing any unused processor resources. This worked great since the other activities took minimal resources letting Handbrake run largely unimpeded.

Even with Handbrake priority of Normal, it still shifts Handbrake entirely to the 8 E cores leaving 50% of the CPU resources unused. It's the "all or nothing" effect that thread director applies that's the problem. I'm leery about setting it to Above Normal because it may not allow anything else to run.

I've read several uses complaining about similar effects - many involving Handbrake.

Truthfully, I wasn't expecting to do any encodes on this machine until I found out how well it executed with multi-core workloads. After seeing the performance I started trying to use it the way I had earlier platforms only to find that minimal alternative tasks wholly cripple the multi-core performance.

I'm not sure there's a solution with current P/E core architecture that will allow Handbrake to run consistently using 80=90% CPU while allowing other activities to take place quickly.

 

[wyliec2]

The more system are changed and the architecture are optimized the less like its Predecessors it becomes optimization for an older system won't be the same as a new one. So you could have been like me and 100s more and built 100's of PC in your past doesn't mean you can properly calibrate a newer system.... not to be rude. I have a lot of issues with newer builds now as well. It takes time to relearn how the system is calibrated for different tasks now. Not everything will be as simple as normal

Given this is a community to provide input and suggestions, your commentary which has essentially been to repeatedly suggest I don't know what I'm doing is rude and of zero value. Likewise, you have not proffered any information to suggest any specific expertise on your part.

If you have a constructive suggestion on how to set 12th or 13th gen P/E core architectures to run consistent 80-90%+ CPU utilization when presented with a mixed workload, please share.

Failing such pertinent information, I will not continue further pointless discourse with you.

 

[Nighthawk117]

Actually changing it to Normal was one of the first things I tried.

I've always had it set to Below Normal which allowed other activities to take place with Handbrake utilizing any unused processor resources. This worked great since the other activities took minimal resources letting Handbrake run largely unimpeded.

Even with Handbrake priority of Normal, it still shifts Handbrake entirely to the 8 E cores leaving 50% of the CPU resources unused. It's the "all or nothing" effect that thread director applies that's the problem. I'm leery about setting it to Above Normal because it may not allow anything else to run.

I've read several uses complaining about similar effects - many involving Handbrake.

Truthfully, I wasn't expecting to do any encodes on this machine until I found out how well it executed with multi-core workloads. After seeing the performance I started trying to use it the way I had earlier platforms only to find that minimal alternative tasks wholly cripple the multi-core performance.

I'm not sure there's a solution with current P/E core architecture that will allow Handbrake to run consistently using 80=90% CPU while allowing other activities to take place quickly.

I've never heard anyone ever mention this in any review ever. If you set it to High does it change anything? If you havn't I would try just to see what it does. I'm just running Handbrake with priority High right now, it hasn't made the PC unresponsive.

You can of course adjust process affinity manually, I believe you can even execute Handbrake encodes from the command line and tell it what to do. Never tried that myself though, I'm a point and click kind of guy.

 

[wyliec2]

I've never heard anyone ever mention this in any review ever. If you set it to High does it change anything? If you havn't I would try just to see what it does. I'm just running Handbrake with priority High right now, it hasn't made the PC unresponsive.

You can of course adjust process affinity manually, I believe you can even execute Handbrake encodes from the command line and tell it what to do. Never tried that myself though, I'm a point and click kind of guy.

I will try setting it higher - I'm in the middle of an encode with 12+ hours remaining. I'm not sure when a change takes effect - I don't know if it's dynamic such that you can change it on the fly or have to close and reopen the application.

What platform are you using with Handbrake??

 

[Nighthawk117]

I will try setting it higher - I'm in the middle of an encode with 12+ hours remaining. I'm not sure when a change takes effect - I don't know if it's dynamic such that you can change it on the fly or have to close and reopen the application.

What platform are you using with Handbrake??

:O That's one hell of an encode. I think it should change it on the fly. My system is an i9 10850K, 32GB RAM with Windows 10.

 

[Zerk2012]

I have considered that but that but concerned it may block the other things I'm doing.

So you never even set it to high priority to see if it solved your problem ? Why not?
If your not happy with what you have and have the other PC/s then just sell the new PC and call it a day, everything is not perfect except your ability to optimize a PC.

Your words.
I respectfully disagree. I have built and optimized systems both professionally and as a hobby for decades.
This is a newer generation not just cores and threads or hyperthreading but your not trying to optimize this new platform.

Not a rant, nothing personal, just from reading this post!

 

[wyliec2]

So you never even set it to high priority to see if it solved your problem ? Why not?
If your not happy with what you have and have the other PC/s then just sell the new PC and call it a day, everything is not perfect except your ability to optimize a PC.

Not a rant, nothing personal, just from reading this post!

Actually the default is Below Normal which has worked perfectly with Windows 10/11 on 5960X, 10600K, 10700K, 3950X and 5950X platforms.

It's only the 12th and 13th Gen with asymmetrical CPU cores that the issue occurs - this is likely as much an issue with thread director as with the CPU itself.

You may not understand clearly that it is not about contention for resources, it's about when you have an intensive task like Handbrake using 90% and you simply open a browser which uses virtually no resources, Handbrake is thrown to E cores leaving 50%+ of the CPU doing absolutely nothing.

The priority shouldn't really matter when you have a job that can use resources and most of your CPU resources are sitting idle - this is somewhat nonsensical given the concept for P/E cores was improved efficiency. he basis for my rant.

I explained (if you really read) how I arrived at this point.

I like being able to fully utilize all the performance I have available, with the 13500, I'm not able to do that. Possibly changing Priority to Above Normal or to High will change something - but it's something that the Windows thread director would have to recognize. From what I read, it mostly looks at which windows have focus at the moment.

I wanted to initiate a discussion so that others that may have a similar use case would be aware of the pitfalls of the P/E core architecture.

Here is a conversation from the Intel community posted April 2022 (it seems like nothing has changed). Intel support basically throws it on Windows.

Intel communities question:

I am using a 12th Gen Intel(R) Core(TM) i9-12900K with Windows 10 pro.

When I am using 3D software to render images, the CPU is running a total of 16 cores using both P-cores and E-cores. That's good.

However, if I open a website or open a different program (e.g. Photoshop or After Effects) during rendering, only half of the cores are used (8 cores) and the computer becomes very slow. It is very inconvenient to use only 8 cores in total.

And back to the 3D software, 16 cores are used, which means that nothing else can be done during rendering. I liked my previous i9 10980XE because I could do various tasks while rendering, and I want to use it the same way.

How can I disable the switching of P-cores and E-cores so that I can always use the maximum number of cores, as the CPU used to?

Response:

Thank you for posting on the Intel️® communities.

Regarding your question, the process of utilizing the P-cores or E-cores depends on the operating system and the task running at the moment; there is no way to change the configurations of the CPU to run the maximum number of cores. However, you can try changing the CPU affinity. It is based on the Operating System, and you need to contact the developer of the operating system seeking help on how to do it.

Best regards,
Jean O.
Intel Customer Support Technician

 

[wyliec2]

:O That's one hell of an encode. I think it should change it on the fly. My system is an i9 10850K, 32GB RAM with Windows 10.

I am now on my 13500 - I made the Priority change to High. With this browser window open, only the 8 E cores are running Handbrake; the 12 threads associated with the 6 P cores are largely idle. Either the priority has no interaction with thread director or it doesn't take effect until the application is re-launched.

Note - with your 10850K, YOU WILL NOT SEE THE ISSUE I"M DESCRIBING. The 10850K should be a great platform for Handbrake.

FWIW - I only use the Handbrake GUI and encode in H265 10-bit preset Slower RF 19-21. My priority is quality followed by filesize - All my content is played on various 4K TVs (65" or larger) with 5.1 and higher audio systems from a media server. I'm happy to get an output file that is one-third the size of the original content. This fits my needs of never having to worry about quality and having 1200+ movies online.

 

[Zerk2012]

Actually the default is Below Normal which has worked perfectly with Windows 10/11 on 5960X, 10600K, 10700K, 3950X and 5950X platforms.

It's only the 12th and 13th Gen with asymmetrical CPU cores that the issue occurs - this is likely as much an issue with thread director as with the CPU itself.

You may not understand clearly that it is not about contention for resources, it's about when you have an intensive task like Handbrake using 90% and you simply open a browser which uses virtually no resources, Handbrake is thrown to E cores leaving 50%+ of the CPU doing absolutely nothing.

The priority shouldn't really matter when you have a job that can use resources and most of your CPU resources are sitting idle - this is somewhat nonsensical given the concept for P/E cores was improved efficiency. he basis for my rant.

I explained (if you really read) how I arrived at this point.

I like being able to fully utilize all the performance I have available, with the 13500, I'm not able to do that. Possibly changing Priority to Above Normal or to High will change something - but it's something that the Windows thread director would have to recognize. From what I read, it mostly looks at which windows have focus at the moment.

I wanted to initiate a discussion so that others that may have a similar use case would be aware of the pitfalls of the P/E core architecture.

Here is a conversation from the Intel community posted April 2022 (it seems like nothing has changed). Intel support basically throws it on Windows.

Intel communities question:

I am using a 12th Gen Intel(R) Core(TM) i9-12900K with Windows 10 pro.

When I am using 3D software to render images, the CPU is running a total of 16 cores using both P-cores and E-cores. That's good.

However, if I open a website or open a different program (e.g. Photoshop or After Effects) during rendering, only half of the cores are used (8 cores) and the computer becomes very slow. It is very inconvenient to use only 8 cores in total.

And back to the 3D software, 16 cores are used, which means that nothing else can be done during rendering. I liked my previous i9 10980XE because I could do various tasks while rendering, and I want to use it the same way.

How can I disable the switching of P-cores and E-cores so that I can always use the maximum number of cores, as the CPU used to?

Response:

Thank you for posting on the Intel️® communities.

Regarding your question, the process of utilizing the P-cores or E-cores depends on the operating system and the task running at the moment; there is no way to change the configurations of the CPU to run the maximum number of cores. However, you can try changing the CPU affinity. It is based on the Operating System, and you need to contact the developer of the operating system seeking help on how to do it.

Best regards,
Jean O.
Intel Customer Support Technician

To sum up what you said the old platforms work with the default settings but your not willing to try new things with a new platform.

This looks more like a troll post instead of a real question seeking answers.

 

[wyliec2]

To sum up what you said the old platforms work with the default settings but your not willing to try new things with a new platform.

This looks more like a troll post instead of a real question seeking answers.

If you don't get it, you don't get it. That's OK. No need to be a jerk Zerk....

That's not exactly what I said....Intel platforms though 11th gen and present day AMD platforms work fine with settings that make sense. Those same settings produce nonsensical results on 12th and 13th gen CPUs that leave half the CPU dormant when there's CPU-intensive jobs running.

Do a search and you'll see a variety of scenarios with people trying to work around the P core/E core issue. Some people have suggested disabling all the E cores to force everything to run on the P cores - throwing away a significant chunk of the computing power you paid for.

Basically, it takes a hack of sorts to get Windows to fully utilize both P and E cores.

 

[wyliec2]

Good news - I've found a workaround that solves most of the issue.

I have a 27" 4K monitor so I can fit quite a bit on the screen.

If I keep the Handbrake Queue window partially visible, the system works normally.

Basically, if I open a browser full screen, Handbrake is immediately shuffled off to the E cores leaving the P cores idle and overall CPU usage drops to 45%.

If I reduce the browser window so most of the HB Queue window is visible, the system ramps back up to 90%+ CPU utilization.

It doesn't matter what keyboard centric work I'm doing in the alternative screens (browser, Word, Excel, Outlook, etc), Handbrake keeps humming along so long as a good portion of the Queue window is visible - doesn't need focus, just to be visible.

Quirky yes but I'll take it.