P3Dv5 CPU utilizes all cores - an observation

[Bob Scott 7,716]

Quote

That is often referred to as supercilious - it  only serves to make your post appear to be saying that anther is wrong. But they were not.

No, it doesn't "only" serve to make my post mean what *you* think it means.  And don't presume to think you have license here to publicly critique the writing style of others.  If you wish to discuss it further, take it to PM.

[SteveW 1,792]

In conclusion we must understand that the simulator is not the same as a video stream, since it has one primary monolithic task - we don't allow it to share a core with that task, and to go with that primary task we have some tasks running in the background that are not time dependent and they might be completed more quickly with two tasks per core. If we have plenty of cores we probably need not share those tasks, with few cores we might be able to use that method to gain performance in the background secondary tasks.

[Dirk98 354]

14 minutes ago, w6kd said:

No, it doesn't "only" serve to make my post mean what *you* think it means.  And don't presume to think you have license here to publicly critique the writing style of others.  If you wish to discuss it further, take it to PM.

I like your English language style and usage discussion not less than HT On/Off!!

Very educational )) 

Cheers,

Edited August 13, 2020 by Dirk98

[Bob Scott 7,716]

One other interesting difference between P3Dv5 and its predecessors is that the main rendering task appears to be shared now between LPs 0, 2, and 4, with much of the workload still on LP0.  I'm not seeing the same hit to LP0 loading when I push up CPU-intensive options like autogen in P3Dv5...but it does appear to result in heavier loading of LPs 2 and 4.

[ComSimPilot 623]

Hi guys,

No need to create tension over this. For me, 10900K stock runs P3D super smoothly. There is no comparison with 7700K overclocked. Motion is super smooth now which results in a completely more realistic feeling especially when landing. So, performance wise I don't see any problem.

Obviously Steve has an in-depth knowledge in how processors work and I thank him very much for taking the time to explain. I must admit, I will need to re-read a few times Steve's previous posts later tonight and come back if I have more questions. 

[Nyxx 6,993]

Well since my next CPU will be a 10900k with HT ON, if steve could just give a AM for v5 and another for addons thats all I would need.

So Steve what's the best AM's for this CPU  please.

Edited August 13, 2020 by Nyxx

[Bob Scott 7,716]

18 minutes ago, ComSimPilot said:

For me, 10900K stock runs P3D super smoothly. There is no comparison with 7700K overclocked. Motion is super smooth now which results in a completely more realistic feeling especially when landing. So, performance wise I don't see any problem.

Undoubtably--the difference between my 6-core 5.3 GHz 8086K and the 10-core 10900K at 5.2 GHz is really pretty impressive in both P3Dv4 and v5.  Going from a 7700K to the 10900K must be a real eye-opener--4 cores to 10...8MB level 3 cache to 20MB...and with a program that will use those cores when it needs them. 

I still have playing with the 10900K's per-core HT on my to-do list.  I'm thinking of turning on HT with cores 5-9, leaving the main P3Dv5 rendering task on cores 0/2/4 to their own unshared physical cores.  It's more a curiosity than anything else to me...both 64-bit versions of P3D rock with 10 cores HT off.  My experience with the 8086K was that P3Dv4 ran smoother without HT...I find myself wondering if per-core HT and P3Dv5 will look any different.

 

[mikeymike 357]

Can someone please explain this to me

i have an i9 9900k not overclocked HT off

now when I enter v5 and pull up task manager I notice the core all fluctuating.

i then right click on P3d exe select all cores hit ok.

then all cores level out and the sim is smooth.

why is that? And should I be doing that?

 

thanks

mike

[Bob Scott 7,716]

1 minute ago, mikeymike said:

Can someone please explain this to me

i have an i9 9900k not overclocked HT off

now when I enter v5 and pull up task manager I notice the core all fluctuating.

i then right click on P3d exe select all cores hit ok.

then all cores level out and the sim is smooth.

why is that? And should I be doing that?

What you're doing is making the OS override the thread affinities that P3D established when it loaded.  Though the load appears more balanced across the cores when you do this, it may well introduce issues with thread scheduling, cache loading, etc that actually slow things down and/or cause microstutters.  I'm of the opinion that LM set it up purposefully to run as it does without intervention, with the sim's core threads and data loaded to and remaining on LP0. 

IMHO, you need to articulate to yourself a good reason for interfering with the processor loading...if you can't, probably best to just leave it be.  Of course, who can't help but to tinker?  😉

[Cruachan 705]

Hi,

I too am experimenting with turning HT off in P3Dv5 via the Affinity Mask on my i7-8 core 5960X. Seems to be working and, as an added bonus, it’s lowering max temps. Previously 2-3 cores were approaching 80C, now all are mostly sub 70C and that’s despite current summer room ambient temps being uncomfortable at times. Good for the cpu which is overclocked dynamically on all cores to 4.6Ghz and good for my peace of mind as well! This processor has served me well over the past 4-5 years and I would like to keep it for a bit longer.

01,01,01,01,01,01,01,01 = 21845 Decimal

All Addons I am placing on the unused LPs 9,11,13 and 15. I’m using Process Lasso for this purpose rather than having to create bat files as it works well. The AM is applied automatically and consistently when each Addon is started.

So far, results are looking good.

Mike

 

[Dirk98 354]

On my i9 9900K with HT On I'm having great results with the following AM:

11.11.01.01.01.01.01.01

Some of SteveW's recent posts and direct answers got me started playing around this particular Affinity. And so far this is the best one. Btw it works very well in XP-11.5 b17 too. Even the scenery loading test time is very much indicative of the improvement as well. You guys, with 8-core CPUs may really want to try it.

Thanks.

PS: hex F555

Edited August 13, 2020 by Dirk98

[Cruachan 705]

7 minutes ago, w6kd said:

I'm of the opinion that LM set it up purposefully to run as it does without intervention, with the sim's core threads and data loaded to and remaining on LP0.

Yes, Bob, but it’s more than likely that their logic applied more to their uncluttered development setups rather than our heavily burdened installations. We have the capability and so we tune (or try to tune) accordingly while avoiding the heat penalties incurred from unfettered use of HT. My setup does not perform well with HT enabled and no user AM. Following Steve’s wisdom and guidance I experimented and found 10 active LPs (4085) to be optimal for most of the time in v4.5. V5 is a different animal but so far, as detailed in my previous post, results are looking good. That could change 😉

Regards,

Mike

[mikeymike 357]

38 minutes ago, w6kd said:

What you're doing is making the OS override the thread affinities that P3D established when it loaded.  Though the load appears more balanced across the cores when you do this, it may well introduce issues with thread scheduling, cache loading, etc that actually slow things down and/or cause microstutters.  I'm of the opinion that LM set it up purposefully to run as it does without intervention, with the sim's core threads and data loaded to and remaining on LP0. 

IMHO, you need to articulate to yourself a good reason for interfering with the processor loading...if you can't, probably best to just leave it be.  Of course, who can't help but to tinker?  😉

I guess I’ll leave it alone then lol.

i will possibly try HT on.

what affinity mask would you recommend for a i9 9900k

thanks  

mikE

[Bob Scott 7,716]

13 minutes ago, Cruachan said:

Yes, Bob, but it’s more than likely that their logic applied more to their uncluttered development setups rather than our heavily burdened installations. We have the capability and so we tune (or try to tune) accordingly while avoiding the heat penalties incurred from unfettered use of HT. My setup does not perform well with HT enabled and no user AM. 😉

 

HT on without an appropriate Affinity Mask is a big no-no...sharing LP0's main workload with anything else really gets in the way. 

mikeymike said he was running HT off, hence my recommendation to leave affinities alone.  Running with HT on is a whole different ball game...and per-core HT is likely to be yet another one still.

[Dirk98 354]

28 minutes ago, mikeymike said:

I guess I’ll leave it alone then lol.

i will possibly try HT on.

what affinity mask would you recommend for a i9 9900k

thanks  

mikE

miKe, try my affinity above, it is like from the horse's mouth.

How do you set it btw?

Edited August 13, 2020 by Dirk98