Jump to content

Sign in to follow this  
ComSimPilot

P3Dv5 CPU utilizes all cores - an observation

Recommended Posts

Posted (edited)

That's the thing though, HT Off solves problems in a no-brainer way, whereas with HT On you have to take a lot of care how it's set up or will give you problems. On all my systems HT On requires proper setup of AMs with the simulator AND at the same time the add-on exe apps that run all the time along with the sim to get better results. How the testing is done is another thing that can fool the tests, you need to run at least three after rebooting so that all the data is cached in the same way,. Reboot the PC and any setup will look bad off the bat. It's best to have some kind of software to record and show the performance. Also what I see is one of the biggest mistakes - guys looking for a hike in FPS. A good smoother setup usually reduces FPS by one or two frames at 30 since it has better traction with the background task.

Edited by SteveW

Steve Waite: Engineer at codelegend.com

Share this post


Link to post

...So with any doubts, or when starting out, use HT Off and see how that goes. Have a think about this: Watching those graphs in Task Manager we can see they all rise to high use during loading the simulator and when new scenery is encountered. With more than 8 cores try using just the first 8 cores. Example with ten cores HT Off try 0011111111 = 255. That way when the data collecting is intense, there's a couple of cores not being hit by that. Since the simulator relies on good system resources throughput as well, those last two cores will be attracting that activity and there will be less affect on core zero.


Steve Waite: Engineer at codelegend.com

Share this post


Link to post

There are considerations we must take care of when testing back to back setups: As time of day progresses we might see a change in weather and traffic schedules. So in other words we don't want those changes we want as near to the same test as possible. I have been using a custom built test harness from way back with FSX to obtain more accurate comparisons back to back when I first started to post about this subject, weather and traffic are the same because no randoms are in use, instead this uses a custom Fibonacci sequence that although looks random it always repeats the same values until the seeds are altered. We can use a saved flight and always start that to help get closer comparisons. Any program that can keep a file of fps results can be used to compute the variance, perhaps a spreadsheet can be programmed for that. But watching for changes can take hours off your day. Looking for a hike in fps can be a mistake because with less traction with the background task frees up the renderer to produce more fps as we see when lowering the FFTF.


Steve Waite: Engineer at codelegend.com

Share this post


Link to post
Posted (edited)

Thanks Steve. I really appreciate your skills here and the time you take to get us on your train.

But I have a question:

When do we have time to fly aroound in the sim? Because THAT exactly is my hobby. 😉

After hours, days and depends on your "primary job, maybe weeks of installing, setting up and testing, a new release comes out again and "here we go "again".

Well, I am surely joking somehow. But I did actually asked myself WHY is all of this necessary? Isn´t there a way to keep it more "userfriendly"? Bacause this is not userfriendly at all.

Isn´t it?

Marcus

Edited by mpo910

Share this post


Link to post

Hi Markus,

That goes without saying.

No need to spend large parts of the day testing when there are faster ways to test. and that is directed only to those that really want to devote time to test.

Remember that there are very varied ways to enjoy the simulator, some love testing and extracting the best performance, that in itself is also a game.

Fly and enjoy the flying, no one is suggesting to spend any time at all testing if you don't want to.

If you read the posts you can set HT off and maybe reduce the core count in the Simulator AM by one or two and get flying. Setting of AMs and HT is not going to extract more performance in shovel loads.

What you do for increased performance is simple - avoid big settings in the simulator 

Regards

Steve

  • Upvote 1

Steve Waite: Engineer at codelegend.com

Share this post


Link to post
Posted (edited)
35 minutes ago, SteveW said:

Hi Markus,

That goes without saying.

No need to spend large parts of the day testing when there are faster ways to test. and that is directed only to those that really want to devote time to test.

Remember that there are very varied ways to enjoy the simulator, some love testing and extracting the best performance, that in itself is also a game.

Fly and enjoy the flying, no one is suggesting to spend any time at all testing if you don't want to.

If you read the posts you can set HT off and maybe reduce the core count in the Simulator AM by one or two and get flying. Setting of AMs and HT is not going to extract more performance in shovel loads.

What you do for increased performance is simple - avoid big settings in the simulator 

Regards

Steve

Steve,

yes of course. I stated I was somehow joking. I did also take a lot of time to proceed like you suggested! Step by step and always use one "base testsetup and related settings".

What I really didn´t liked is the status of V5 as it came out at it´s first release. That took a lot of time and brought some nasty taste to us/me. 

With HF 2 it is much better. Allthough I am suffering from some issues wich are related to EA (BETA) but also VR (HP Reverb). Water reflections not drawn simultaniously in both eyes (seems a frame lag or viewpoint issue), Single pass rendering is faulty too (without EA mode faulty), 2 overlapping views in each eye with EA off). Also I am thinking there is still some room for core handling improvement, VRAM usage (inconsistant VRAM usage only by switching between VR on and off - VRAM is increasing slowly).....

With 5.1 these things have to be fixed. I am counting on LM here. I actually like the SIM a lot. 30 to 45 FPS in VR stereo mode on EGLL and EDDF........with all of it´s glory addons in PMDG/QW and Aerosoft planes). THAT is remarkable good! 4xSSAA with 150% set in Steam VR to sharper images!

Also I would expect 5.1 to make usage off the capabilities of Simul TrueSky. The way it is implemented is a good first "try" but there is a LOT of room for improvement.

Sim is stable in my case. NO CTD, and if I have one, mostly it is not P3D but something on my rig! 

Regards Marcus

System specs

Prepar3D_v5_Academic_5.0.31.35253

Edition Windows 10 Home
Version 2004
Installed on 20200501
OS Build 19041.388
Experience Windows Feature Experience Pack 120.2212.31.0

Latest Nvidia Drivers

i9 10900K @ 5.0 Ghz | HT ON | 1.32V | AffinityMask=87381
Mainboard ASUS ROG Maximus XII Formula Z490
32 GB RAM 3600 Mhz Corsair Vengeance | 4x8GB
RTX Titan 24GB
32" Samsung UHD Monitor using 4k
Custom Water Cooling | 2x 360MM | 1x 240MM Radiator
HP Reverb v2 VR HMD | Steam VR

Main Addons

ASP3D | 7514 Open Beta
ChasePlane | Via ORBX
FSReborn Prof.
GSX 2
FSUIPC6
AIGAIM | Latest Version
SODE | Latest Version
MultiCrewExperience | Latest Version
Airbus A330 | Latest Version
PMDG 747/748 | Latest Version
QW787 | Latest Version
FSLABS A320 | Latest Version
Navigraph Charts
ORBX Base
ORBX LC
ORBX TE NL
ORBX TE NCA
ORBX Regions
ORBX HD Buildings
ORBX HD Trees V1 & V2
REALTURB | All Continents
+200 Addon Aiports | via .xml method added
 

Edited by mpo910
  • Like 1

Share this post


Link to post

Sure, I am OK with that.

I used a test harness originally because since I was asked to be giving advice on these forums I wanted to be sure of what changes are useful or not. So with passing that on to others there's a fair chance others might be better helped.

P3D v5 handles HT fairly well. However from the OP post at the top we can see that the HT mode shares two major tasks on core zero and that's an area for improvement. So as I mentioned have a go with HT off and that will solve that issue. Then look at if you can improve things with HT on.

In the example for the ten core 0011111111=255, the HT equivalent would be 00,00,01,01,01,01,01,01,01,01=21845.

If you went for no AM HT Off = 1111111111, that in HT mode would be 01,01,01,01,01,01,01,01,01,01=349525.

Staying with HT and looking in Task Manager graphs I see activity on LP1 00,00,00,00,00,00,00,00,00,10, the second in from top left. That is taking throughput from that LP0 since these two LPs are sharing the same core zero. Reducing the core count to 8 LPs = 21845 as I showed above, I notice that this activity on LP1 has decreased,  that seems to be because the free cores have attracted that activity. Remember also that we have stages of what the sim does, it might be running in a settled state or it might be pulling in new data.

It is with these ideas that we might be able to improve things, and it is by nature a very awkward process. Also remember that we can only improve a little with HT and AMs, and that AMs can only reduce the LPs used. Since there is an optimum count with many cores available, we might be able to achieve some useful results.

  • Like 1
  • Upvote 1

Steve Waite: Engineer at codelegend.com

Share this post


Link to post
2 hours ago, SteveW said:

Sure, I am OK with that.

I used a test harness originally because since I was asked to be giving advice on these forums I wanted to be sure of what changes are useful or not. So with passing that on to others there's a fair chance others might be better helped.

P3D v5 handles HT fairly well. However from the OP post at the top we can see that the HT mode shares two major tasks on core zero and that's an area for improvement. So as I mentioned have a go with HT off and that will solve that issue. Then look at if you can improve things with HT on.

In the example for the ten core 0011111111=255, the HT equivalent would be 00,00,01,01,01,01,01,01,01,01=21845.

If you went for no AM HT Off = 1111111111, that in HT mode would be 01,01,01,01,01,01,01,01,01,01=349525.

Staying with HT and looking in Task Manager graphs I see activity on LP1 00,00,00,00,00,00,00,00,00,10, the second in from top left. That is taking throughput from that LP0 since these two LPs are sharing the same core zero. Reducing the core count to 8 LPs = 21845 as I showed above, I notice that this activity on LP1 has decreased,  that seems to be because the free cores have attracted that activity. Remember also that we have stages of what the sim does, it might be running in a settled state or it might be pulling in new data.

It is with these ideas that we might be able to improve things, and it is by nature a very awkward process. Also remember that we can only improve a little with HT and AMs, and that AMs can only reduce the LPs used. Since there is an optimum count with many cores available, we might be able to achieve some useful results.

Steve,

thank you and really appreciate your knowledge.

but basically you need need a lot of time on your hands to tinker with HT on and affinity masks?

but for us that don’t have all the time

just switch HT off and adjust sliders  and  bobs your aunty?

 

thanks

mike

  • Like 1

Share this post


Link to post

Exactly, and remember what I said earlier: Start out with HT off and with many cores you can use an AM to restrict that core count to free up one or more to give breathing room for the system resources when the simulator is pulling in scenery.

  • Like 1

Steve Waite: Engineer at codelegend.com

Share this post


Link to post

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this  
  • Tom Allensworth,
    Founder of AVSIM Online


  • Flight Simulation's Premier Resource!

    AVSIM is a free service to the flight simulation community. AVSIM is staffed completely by volunteers and all funds donated to AVSIM go directly back to supporting the community. Your donation here helps to pay our bandwidth costs, emergency funding, and other general costs that crop up from time to time. Thank you for your support!

    Click here for more information and to see all donations year to date.
  • Donation Goals

    AVSIM's 2020 Fundraising Goal

    Donate to our annual general fundraising goal. This donation keeps our doors open and providing you service 24 x 7 x 365. Your donation here helps to pay our bandwidth costs, emergency funding, and other general costs that crop up from time to time. We reset this goal every new year for the following year's goal.


    31%
    $7,965.00 of $25,000.00 Donate Now
×
×
  • Create New...