I've tried the various settings but I still see the most performance with HT enabled using  AM=224

AM=224=(11,10,00,00), so on a four core with HT disabled it is equivalent to no JOBSCHEDULER section or AM=0=(1111). But with HT enabled allows only three threads on two logical processors and could not somehow perform well.

Good Morning all!

 

what an interesting reading! Thanks for all your efforts.

 

I can't get my head around all the information we have here. I am currently running an i7 6700k at 5GhZ. I am running HT OFF and AF14. 

 

What changes would you suggest me to try, so I can post the results back to you guys?

 

HTON AF85? AF84?

 

Thanks a lot.

 

Gio

dude, wow what vcore & cooling are you using to get the 6700k to 5ghz

Hi,

 

I was under the same impression. Until I made a quantitative analysis. This is

 

3) HT off, no AM, frames unlimited

 

 

Compare this to the upper diagram 1) in my post #93. Same settings (except fps), same circuit (KSFO), Frames are (much) higher, but jitter is as well. I am still surprized by the result, but have to accept it.

 

Of course your system may behave in a different way.

 

Kind regards, Michael

Your post 93 definitely shows locked or limited frames set to 20, but now your graph shows no limited or locked frames. How can any kind of useful information be garnered from those comparisons?

 

I would say from the unlimited graph is that the system will show drops even at 20fps fixed. I would certainly look for more performance from this system by reducing settings and set 20, or leave as-is and set 15 locked.

Your post 93 definitely shows locked or limited frames set to 20, but now your graph shows no limited or locked frames. How can any kind of useful information be garnered from those comparisons?

I wanted to compare just this aspect: the difference of size of fluctuations between locked frames and unlocked. I am very sorry if I didn't make this clear.

 

And, yes, I agree, the system struggles towards the end (landing at FB KSFO HD) where it drops below 20 and develops fluctuations. I agree, this could be ameliorated (i) by either lowering settings, or (ii) locking to a lower value, no objection here.

 

Kind regards, Michael

With locked settings the graphs generally stick to the line showing various dropouts because it's based on the visual frames, but the actual render frame rate is fairly close to unlimited as it renders the look ahead buffer. The best way is to compare Unlimited graphs. Nothing wrong with your comparison if it compares locked vs unlimited of the same flight section with no randoms. What I do is save a flight at some point on AP, and let that set off for each test period.

Hey,

 

I also have to ask which is the best AM setting for my 5820K.

 

What i see is when i use all threads and only mask the virtual core on cpu zero via AM i get a very low fps variance in p3d v3 only between 1-2 %. When i disable all "virtual" cores via AM i get about 15 % fps variance.

Yesterday i had a saved flight over London in the PMDG 777 and i always like to do landings so i tried with different settings but forgot to try without vsync internal. Will try this morning. Now about smooth experience the fourth try was the best and also the most demanding with LOD maxed and extreme complexity and dense, very dense autogen. HD2048.LOTS of shadows and ultra. High water and butter smooth to 27L from LAM.

 

Strange thing is that the third attempt was with normal settings and i got an oom.

 

But the sim seems to like a lot of restarts. Also loading gets faster as the cache is build up.

 

Michael Moe

Hello Simmers;

 I need to find out the best config. for Nvidia Inspector for my Prepar3d V2.5 and V3

 

I can`t find a good config for it and I think because of that I`m getting instability issues.

 

 My PC specs and tweaks are:

 

Windows 10 64x pro

 

Intel i7 2600K OC to 4.3Ghz using Corsair Hydro liquid cooler

8GB of RAM 1866

SSDs

GPU: 2 x ASUS GTX 660 Ti DirectCu "SLI"

 

Maybe some has the same config.

 

I`ll be so grateful for the assist and if you need my .CFG info, let me know.

 

Thiago 

​

AM=224=(11,10,00,00), so on a four core with HT disabled it is equivalent to no JOBSCHEDULER section or AM=0=(1111). But with HT enabled allows only three threads on two logical processors and could not somehow perform well.

I used it on my 8 core AMD and it made a huge performance difference and then I used several other AM's with my Intel CPU but still got the best performance with 224, I was sitting in the mid 30s with 85 and 84 but I get about 55 with 224

Cool, 3 cores should be enough but some systems could use more. But I'm saying a 4 core with HT disabled would see an AM of 224 as 0, and a four core with HT enabled is still only four cores. How many real cores is that AMD?

Getting rock solid 30 with 20% traffic, FlyTampa OMDB, PMDG 777. AM=85, FFTF=0.01

i7 4770k 3.5ghz, GTX 970 4GB.

Presume you mean with HT enabled since 85=(01,01,01,01), should give a good result.

 

For a 4 core HT=On system 85 should work OK, one LP masked of each core, assuming there's no big addon process runs up before the main sim thread and results in sharing the core running the main sim thread. On any system it's best to look at the sim starting up with Task Manager and check which cores or LPs come alive, core zero is shown top left in the performance graphs.

 

Often we see with certain addons they state an AM of 14=(1110) to leave the first core free (for the program that starts before the sim). The 8LP (HT enabled) equivalent value would be 84=(01,01,01,00) or 168=(10,10,10,00). Trouble with that is the sim can work slightly better with four rather than three processors, although three are fine in most cases.

 

I show affinity masks with the binary value following in braces, I also include commas to separate out HT pairs, to assert the fact that HT is enabled or not.

 

14=(1110) - core zero on the right.

84=(01,01,01,00) - core zero HT pair on the right showing Logical Processors 1 and 0.

 

Hopefully it is starting to sink in that we cannot simply state a claim for some AM, since they are explicitly tied to the simulator PC system and addon types. Whether HT is enabled or not is extremely important when discussing AMs with Intels.

 

 

My advice is to first get the main sim threads working well, with an AM if HT is enabled, or perhaps on an 8 core CPU we might also use an AM to mask 2 or 4 cores.

 

Then get the sim quality vs frames settings right with all vsync modes disabled and Unlimited fps setting to test performance. Run the sim in Unlimited for at or above the monitor refresh rate, but below 40fps as the time between frames becomes erratic, use locked fps. With locked fps choose a division of the monitor refresh rate, on a 60Hz display that's 30,20,15, or 1/2, 1/3, 1/4 refresh. If going for locked frames choose around half the value of the minimum unlocked frames performance.

Tried Jobscheduler at 1364 for my i7 5930k and this is the result (see screenshot) should I not be getting more action on the other cores because core 2 or whatever that second one is called shows all the action. 

 

 

Also jobscheduler was not in my config so I added it,,,is that correct?

 

Yes that looks correct, your Logical Processors are arranged LP0 Top Left, LP1 Lower left, LP2 (the main sim thread) next from left top, and the masked LP3 below. All the upper ones show more activity which is to be expected.

 

1364=(01,01,01,01,01,00) should be OK for 6 core HT enabled.

AMD FX8350 is 4 modules , 1 modul =2 cores with shared resources.