Best CPU for Prepar3dv4.2

[Bert Pieke 7,585]

Sure is sounding like a topic for the Hardware forum to me...

[ssair1 92]

How do you determine which cores Prepar3dv4 is using? Does anyone know what is the maximum number of cores, according to LM, that P3d utilizes? Does threaded optimization make a difference? Ok... I guess I need to find some good resource materials on this subject? Any recommendations? 

[Bert Pieke 7,585]

1 hour ago, ssair1 said:

How do you determine which cores Prepar3dv4 is using? Does anyone know what is the maximum number of cores, according to LM, that P3d utilizes? Does threaded optimization make a difference? Ok... I guess I need to find some good resource materials on this subject? Any recommendations? 

I am not sure there is a maximum.. it is more a case of the benefit you get from farming out work to ever more cores..

Just like 9 women cannot create a baby in one month, there is a practical tradeoff in terms of coordinating the effort.

From what I have read on these forums, a 4-6 core CPU is plenty for scenery loading which is really all the additional cores do. The raw speed of the main P3D core is ultimately what determines actual rendering performance (alongside the other critical components like motherboard, memory and GPU, all of which are part of the effort).

Edited May 20, 2018 by Bert Pieke

[Nemo 1,305]

I had always the opinion that it is better to have all cores at the same speed. I seems that I am wrong. But how much difference between main core (core 0) and other cores is "allowed"?

Edited May 20, 2018 by Nemo

[Pete Dowson 558]

4 hours ago, ssair1 said:

How do you determine which cores Prepar3dv4 is using?

With only P3D running, open up the Windows Task Manager and check the "Performance" tab. Then check with P3D closed. The difference must be all P3D (the rest Windows and other background services).

There's an "always on top" option in "Options" if you want to check with P3D in its "pseudo" full screen mode.

Pete

 

[Pete Dowson 558]

3 hours ago, Bert Pieke said:

From what I have read on these forums, a 4-6 core CPU is plenty for scenery loading which is really all the additional cores do.

Yes. Whilst I let P3D use any or all cores as it wishes, I set the core affinity for other processes (and for FSUIPC's own internal threads, such as for Lua plug-ins and WideServer) to cores other than those which Task Manager shows P3D actually using most. Thus, for me at least, whilst not all the cores are fully loaded, at least I'm making good use of them all! 😉

Pete

 

[GSalden 4,657]

Quote from Beau Hollis , LM P3D Architect :

”The default behavior is to use all cores. The only purpose of applying an affinity mask would be to disable one or more cores. We don't recommend this, but we recognize that it may be beneficial in some cases. For example, there could be an add-on that communicates with another application. There could also be other unrelated windows applications running, that users don't want to close when they run Prepar3D. Higher fidelity training devices often rely on additional applications to perform simulation work. In that case, the system might be configured to allocate specific cores to specific applications. This can be done via OS settings, but the AM setting in the cfg makes it a bit easier.

The other use case that is often mentioned is to disable the hyper-threaded core that shares a physical processor with the primary thread. Sometimes doing a full workload on the HT core can slow down the primary. This is really chip and workload specific. We updated our job scheduler in v2 or v3 such that it typically uses core 1 without overloading it. In our tests, even with HT on, using all cores was typically better than masking off core 1 after the scheduling changes were made. Since then, we have not recommended custom AM settings. 

One last thing I should note is that add-on dlls are considered to be part of the Prepar3D process. The AM settings will be applied to those dlls as well. Add-on developers may be creating their own threads and assigning heavy work loads to them.”

 

 

[Pete Dowson 558]

52 minutes ago, GSalden said:

One last thing I should note is that add-on dlls are considered to be part of the Prepar3D process. The AM settings will be applied to those dlls as well. Add-on developers may be creating their own threads and assigning heavy work loads to them.”

Yes, this is why FSUIPC's individually-specifiable AMs must be a subset of those set for P3D (which is best left default therefore), and why I check which cores P3D is using mostly so I can specifically assign other processes and threads to less-used ones.

This check seems only to be needed once as P3D seems to choose (or be allocated) the same cores each time. I should probably check this again on every new P3D release, though.

Pete

 

[G-RFRY 1,712]

I'm running the same CPU clocked at 4.9 in software with smart fan it did get to 5ghz but I limited t to 4.9, on a 28inch 4k panel, I noticed when I ran it on RivaTurner  for tweaking it ran mostly at 4.6-4.7 and only hit max when you do a view change at high detail area then drops back,

I have my sim locked at 30 fps and stays there again only on very high detail airports it will drop but it will stay above 20 fps with PMDG aircraft.

    

[robbroadbent 0]

Hi all,

I was wondering if someone could offer some advice please.

I have a new system similar to those mentioned here, Asus X299 Deluxe Prime motherboard, i9 7900x CPU running at 4.6Ghz, 32gb Ram and 3 x 1080ti cards with SLI bridge installed.

I am very unfamiliar with setting up SLI.

I also have 3 projectors and use Immersive Display Pro for warping. I was running a 6700K CPU with 2 980ti's and it ran fairly well. It had 1 projector on one GPU and 2 projectors on the other in non SLI because the MB did not support it.

For the life of me I cannot get decent frame rates with the new system. I have tried one GPU for each projector. (Woeful!) All projectors on 1 GPU in SLI (better but still not great) and the best FPS I get is the 3 projectors on one GPU not running SLI. (ie 2 GPU's sitting there doing nothing)

In the first 2 scenarios, as soon as I create a new window (for testing purposes) and undock it, the FPS plummets. There is only a small FPS hit in the 3rd configuration.

I have not changed anything in the Nvidia Control Panel except to enable SLI so I don't think that is the problem unless there is another setting I don't know about.

I have read many forum posts and tried a multitude of suggestions but nothing seems to work for me.

Any suggestions would be appreciated and thank you for your time.

Best regards

Rob Broadbent

[him225 51]

On ‎5‎/‎21‎/‎2018 at 12:13 PM, GSalden said:

Quote from Beau Hollis , LM P3D Architect :

The other use case that is often mentioned is to disable the hyper-threaded core that shares a physical processor with the primary thread. Sometimes doing a full workload on the HT core can slow down the primary. This is really chip and workload specific. We updated our job scheduler in v2 or v3 such that it typically uses core 1 without overloading it. In our tests, even with HT on, using all cores was typically better than masking off core 1 after the scheduling changes were made. Since then, we have not recommended custom AM settings. 
 

This is an interesting point, is there any further info available as to better decide in which cases disabling the core may be beneficial and when not?

[GSalden 4,657]

Hi Rob,

Here I have a triple view setup too : 7940x ( 4.8-4.7- 12x 4.4 ) , 32 Gb and 1 1080 Ti.

I was able to test with a second card, Titan X ( almost like a 1080 Ti ) and also on my system 1 card was smoother with slightly better fps.

Since P3Dv4 I remarked that with V4.1 and V4.2 , each extra undocked view costs more performance then before. Before it was 25-30% loss per undocked view , while now it is more like 40% per undocked view. 

In your case start with no AM and no HT as my CPU goes to 100% on almost every core when trying to use an AM entry.

Also set your core 0 to 4.8 , core 1 to 4.7 and leave the rest to 4.4. Core 0 performance is what you are after. Windows chooses whatever core it wants but with a core 1 at 4.7 it probably will start there.

Edited May 27, 2018 by GSalden

[westman 257]

SLI always give a footprint sligtly lower max fps 2-3fps, 3way SLI not supported by LM as lot of other apps work most for syntetic benchmarking like 3Dmark.

I not run multimonitor anymore Gerald and Pete have all answers there.

Run 4K locked 30hz there a have not need the extra max fps single card give but on the other hand you dont bottleneck 2X1080Ti as you do with single card.

after lot of testing with differnt setups for P3D V4-2 monitor 55inch 4k TV set at 30hz.

Retired my X299 7920X and 7980XE for a 8700k my personal thougts is FRQ is king , doing tests from 5ghz up to 5.4ghz with ht on and off the 8700k performs like the 7980XE at same FRQ only drawback loadtimes. 

i run the 8700k 5.5ghz HT off with ultra fast mems 4266mhz C17 2x1080TI SLI , this setup is much smother then for example 7980XE at 5ghz or 8700k at 5.3ghz HT on .

This is my personal thoughts after testing, the way i want to have the sim runing with my settings

here is a link to  firestrike ultra hall of fame result with 3X1080TI , my aka is gubben only to know that  i run 3X180TI

https://www.3dmark.com/hall-of-fame-2/fire+strike+3dmark+score+ultra+preset/version+1.1/3+gpu

 

Edited May 27, 2018 by westman

[Pete Dowson 558]

3 hours ago, westman said:

SLI always give a footprint sligtly lower max fps 2-3fps, 3way SLI not supported by LM

Oddly I found that the fps was about the same with SLI enabled or disabled on my 3 x 1080Ti's, but noticeably smoother with SLI than without. As this isn't the doing of P3D4 I can only assume it is down to the nVidia driver.

There are some experiments I can make BIOS-wise, like upping the Core 0 frequency and lowering the others -- currently they are all on 4.8GHz. The other thing I'd like to do is use the full rated speed of my 32Gb of memory -- 3400 Hz. Currently it is wound down to 2800 Hz because the system seemed a bit unstable even at 3000!

I have the projectors all set to 30Hz (actually 29Hz because one of them wouldn't set at 30 despite all three being identical!)

I'm actually achieving around 20 fps even at the densest airports, but with drastically reduced AI Traffic (limit of 80 at frame rates less than 25). It does dip to 15-16 occasionally, but it is reasonably smooth even at seemingly terrible rates.

If I could wind things up to get 20 fps minimum with more traffic, I'd be very happy. But I think this may need some action by L-M to make P3D4's performance better with three separate scenery virews open.

Incidentally (or perhaps it is quite relevant?), none of my windows are "undocked". They are three opened windows on the same stretched P3D4 background. I suspect doing it by undocking would be even poorer in terms of performance.

Pete

 

[robbroadbent 0]

3 hours ago, westman said:

SLI always give a footprint sligtly lower max fps 2-3fps, 3way SLI not supported by LM as lot of other apps work most for syntetic benchmarking like 3Dmark.

I not run multimonitor anymore Gerald and Pete have all answers there.

Run 4K locked 30hz there a have not need the extra max fps single card give but on the other hand you dont bottleneck 2X1080Ti as you do with single card.

after lot of testing with differnt setups for P3D V4-2 monitor 55inch 4k TV set at 30hz.

Retired my X299 7920X and 7980XE for a 8700k my personal thougts is FRQ is king , doing tests from 5ghz up to 5.4ghz with ht on and off the 8700k performs like the 7980XE at same FRQ only drawback loadtimes. 

i run the 8700k 5.5ghz HT off with ultra fast mems 4266mhz C17 2x1080TI SLI , this setup is much smother then for example 7980XE at 5ghz or 8700k at 5.3ghz HT on .

This is my personal thoughts after testing, the way i want to have the sim runing with my settings

here is a link to  firestrike ultra hall of fame result with 3X1080TI , my aka is gubben only to know that  i run 3X180TI

https://www.3dmark.com/hall-of-fame-2/fire+strike+3dmark+score+ultra+preset/version+1.1/3+gpu

 

Thanks Gerard,

I'll turn HT off tomorrow and see if that improves things. (Its 9:50 PM here in Oz) I'll also try the O/C suggestions. 

When you had the second card installed, did you run in SLI? If so, was it better with SLI on or off?

I don't quite understand why when I run the 3 GPU's not in SLI (1 to each projector) that it is the worst scenario with lowest frame rates. It seems P3D doesn't like to operate on more than 2 GPU's.

You are spot on about the frame rate hits when you undock a view.I'll report back with my findings.

7 minutes ago, Pete Dowson said:

Incidentally (or perhaps it is quite relevant?), none of my windows are "undocked". They are three opened windows on the same stretched P3D4 background. I suspect doing it by undocking would be even poorer in terms of performance.

Pete, could you please explain how you achieved this?

I thought the only way to achieve a non distorted view was have the 3 separate windows.

Best regards

Rob