Prepar3D V4.5 is behaving very "Stuttery" and "Laggy"

[Evros 310]

20 minutes ago, SteveW said:

So with HT disabled have a look at how many cores are in operation:

A. during loading the scenario

B. during settled in the flight

C. fly across to a new area so that the sim has to load more data

We can add more than four to improve the loading time, but that can be at a cost of managing the extra bandwidth when they all work together to do not so much. So needs some homework to be sure about a particular machine and the preferences and scenery and so on.

 

 

Right, so I just shoved the shelved 8600K to my FS rig to test the scenario. Without AM, all cores were loaded at around 85% usage. Sim was smooth and everything went well. With AM=30, all assigned cores were running 100%, creating excruciating stutters and causing blurries.

 

21 minutes ago, SteveW said:

What have you got showing in task manager there then, probably four very obvious tasks, but any more showing up in there?

Four tasks? What? You might be confusing the terminology here.

 

15 minutes ago, SteveW said:

Also your point about avoiding core zero in my AM 30; well my AM 30 does in fact miss core zero agreed 011110.

But it is not, so as to miss core zero, it is just simply to show that any four cores will be the same as far as P3D and FSX are concerned.

We can try 110011 if we want AM 51. Makes no difference.

What makes a difference is how the simulator is used and what other exe apps are running beforehand and so on. So the best way to extract the most is to use an optimum amount of cores, usually 4 or 6 and leave the rest for the system. We can go further and corral exe apps onto a subset of cores. Remember that with HT on or off some programs will occupy every core they find, that 4 in a 4 core and 8 in a four core + HT.

I don't have a point in avoiding core zero. My point is that no core should be avoided, period. We don't need to spoon feed hardware management. Kernel is more than capable of balancing the load when different processes are fighting for resources on same execution pipeline. See first quote for example.

 

18 minutes ago, Rogen said:

Sounds like the more threads spun out onto cores equals the more work the managment thread (running on core 0) has to perform.

So more cores = good up to a point at which the management of those cores starts to cost perfomance.

 

53 minutes ago, Evros said:

Unless you are running HEDT with "too many cores", I see no logical explanation to do so

You see, I understand the case. It's just that AM should not be suggested as a blanket statement for better performance. You are just taking performance off the table with it when you apply it to average consumer CPU.

 

[SteveW 1,792]

26 minutes ago, Evros said:

You see, I understand the case. It's just that AM should not be suggested as a blanket statement for better performance. You are just taking performance off the table with it when you apply it to average consumer CPU.

 

No one suggested that. Show me your proof of horrendous stutter etc. I'm not seeing anything like that.

I think you are unfairly looking for discrepancies where there are none. You need also to refer to all posts on a subject by an author. Don't just pull out one where it can't say everything about something to cover folk that want to find a problem with it.

AMs have been in use for 20 years to corral an app that will spread.

[SteveW 1,792]

If in doubt maybe check in with Gerard. He's moved onto a 14 core from a 6. He said with the six he could not improve on HT Enabled AM 340. And now with the 14 core he is using 8.

[SteveW 1,792]

...Gerard has unlocked all the achievements and attained Yoda level.

[Evros 310]

33 minutes ago, SteveW said:

No one suggested that. Show me your proof of horrendous stutter etc. I'm not seeing anything like that.

Perhaps tomorrow if I have time. I'm about to go to bed now. But it should be self evident that when pipeline is backlogged, stutters will happen.

 

33 minutes ago, SteveW said:

You need also to refer to all posts on a subject by an author. Don't just pull out one where it can't say everything about something to cover folk that want to find a problem with it.

I don't understand what you mean by this. Actually I have hard time deciphering half of your posts is this thread, so if possible, maybe you could put more effort in to composure of your posts. This is where misunderstandings come from.

Edited December 7, 2019 by Evros

[SteveW 1,792]

 

8 minutes ago, Evros said:

I don't understand what you mean by this. Actually I have hard time deciphering half of your posts is this thread, so if possible, maybe you could put more effort in to composure of your posts. This is where misunderstandings come from.

I do, all the time. But I think you didn't read them properly. I've said the same thing for a decade rearranged a million ways.

[SteveW 1,792]

So the Q. was how does AM 85 on a 4 core+ HT make a difference to a four core set to HT disabled?

It doesn't, it is the same as HT off no AM.

01,01,01,01=85 HT on

is the same as

1111=15 HT Off

The difference is in the way the system performs.

And if a four core is such a bad thing, how come so many have those without complaint?

That's surely understandable?

Edited December 7, 2019 by SteveW

[newtie 533]

2 hours ago, Sethos said:

the masks that don't use the first core. I've seen it on multiple systems and on a colleague's system that P3D will then struggle to render terrain properly.

Not on my system.

[newtie 533]

2 hours ago, SteveW said:

You definitely 'get it' sir.

Cool.

Steve and I had a conversation quite some time ago (you may not remember) about digging a hole.

Two guys can't dig a hole at the same time; one jumps in the hole while while the other gets out of the way.

Can only get one guy in the hole at a time 😀

This is what Rogen is talking about.

Cheers,

Mark

[Jim Young 2,356]

On 12/5/2019 at 1:45 PM, BrendanMan33 said:

i'm quite stumped as to why my simulator is performing so poorly.

You have a modern and powerful system to run P3Dv4.5 and many add-ons so P3D is not causing the problems with the stuttering and lags (believe stuttering and lags are the same).  If you mean flickering, you get flickering if you turned off v-sync and triple buffering.  The problem you are having is with an add-on or utility you have installed.  P3Dv4.5 will run great by itself and with your P3D settings all set to the max and everything turned on like autogen, reflections, shadowing, dynamic lighting, etc.  It is when you install just one add-on that things start to go bad. 

Looking at your long list of added software programs, the only thing I see that might be causing your stuttering is the Orbx FTX programs like the FTX Global Base and other elements that are installed after that.  That and another scenery combination may not be compatible or your settings for the Global Base are too high (understand you can adjust this setting).  If you have the ability, you should disable the Orbx/FTX Global Base programs or uninstall them.  Run P3D and see if this fixes the problem(s).  If not, then disable all of your scenery add-ons by going into Lorby-Si's Add-on Organizer and disabling all of the add-ons.  Run P3D and see if this fixes the problem.  If you enable all of your photoscenery (and some airport scenery uses photoscenery), you should disable it.  Photo scenery will cause stuttering and lags if too much of it is loading.  As I originally reported a few years ago, photoscenery will load if enabled even if it is not located in your current flight plan.  So, if you are flying in Europe, photoscenery enabled for, say, California, will load too.  I believe Flightbeam and FSDT scenery have photoscenery in it.  Looking at your long list of products, I do not see any photoscenery like MegaScenery Earth or Orbx/FTX Northern/Southern Germany or Aerosoft's Austria.  Just many of the airports. 

Another add-on that takes up a lot of resources is AI but only if set too high. Keep  I see you use a pirated AI program (starting with SP) and that is not allowed here at AVSIM and I have removed your link to the list of add-ons.  You must remove this from your list.  This pirated program uses intellectual property developed by others and they never gave the pirate permission to use the aircraft in their program.  You should also remove all of the links to the software programs you own as advertising on AVSIM is not allowed and you have links directly to the stores.  You can post only links to products in the AVSIM Bargain Hunter's Forum only.

So these are things you need to investigate.  It is impossible for others to figure out why you have the stutters and lags as there are so many settings used by members and so many add-ons not used by others that could be the cause of your problems.  Stuttering has been discussed so much on AVSIM over the years and I believe only the "blurries" beats it out (the blurries can mostly be fixed by turning off hyper-threading in the BIOS). Only i7 and i9 CPU's have hyper-threading so you won't have this option in your BIOS which is good.  Intel is getting rid of hyper-threading in the near future.  More info at the following link - https://www.techspot.com/article/1850-how-screwed-is-intel-no-hyper-threading/.

Hope this helps.

Jim

 

[SteveW 1,792]

If we look at that HT AM85=01,01,01,01 we can improve the loading time by using 01,11,01,01

On the pair '11' we see two tasks loading in HT mode. What that means is that their threads take turns to use that core, as Mark says.

However, with HT on that core their context (situation within the core) is not lost and when the threads reconvene they don't have to load up a saved away state.

Also those two LPs (logical processors) don't have to endure any waiting as the other threads fill in on the Sister LP.

So with that '11' we see an increase in loading speed, decrease in time to load the scenario, and it's not even an additional core on the bus. So that way HT improves the system throughput by more or less halving the switching time between threads and gives us options to extract more performance. If we know what we are doing. Also the shared cache is used more effectively with HT enabled.

That's in effect all the time. But as I said when the sim is in normal operation they don't go mad unless we set too high a setting.

 

[SteveW 1,792]

Good one Jim.

I recommend installing the sim and before adding in addons, check out the performance first. Sort it all out first so that when we add in some scenery, or aircraft, we can see the effect it has in relation to the sim and the performance capacity of the machine.

 

Do not set high settings and bring them down, will be all day doing it that way.

Instead set sliders low at first and check out fps (SHIFT+Z). Bring up sliders without dropping fps below the refresh rate of the monitor. Use Display Settings Unlimited + VSync = Off for that test.

Next turn VSync=On or use locked fps (for low fps around 19-24).

When that is understood, keep the setting profile. Now introduce something and retest.

 

 

Edited December 7, 2019 by SteveW

[SteveW 1,792]

Another one I see is that P3D is different than FSX.

P3D has lots of xml instructions abnout scenery and aircraft and other items, that are all pathed in a very complex setup.

This has moved most development onto testing thoroughly with the stock Program Files location and a drive other than  C.

Beware of installing in a non stock location or private folder as it can't easily be certain that all things are pathed correctly.

When someone speaks about massive stutter, (and behaviour changing with AMs), that could be due to an error in the build.

 

 

 

[SteveW 1,792]

33 minutes ago, Jim Young said:

 Intel is getting rid of hyper-threading in the near future. 

 

For games with HT enabled all the tasks bunch up onto half the cores so it's not recommended unless they are built like P3D and FSX. Also most games use four cores effectively, especially now with SSDs and fast motherboards and GPUs. So for games get a four or six core so it can scream, and put more money onto the GPU.

Just testing out a few AM's no difference in behaviour of P3D as one would expect. SO change of AM makes no difference. Even fps is more or less the same with only two cores instead of four. Takes twice as long to load the scenario with only two. With 6 cores knocks off a couple of seconds than four cores. That's reflected every time the sim gathers more tiles and scenery objects.

[SteveW 1,792]

Remember that FSX ESP and P3D which inherited that behaviour, these were not built as games. ESP had to fit inside another system, that might run HT or many cores for other reasons than just performance tweaking. So the affinity is controllable on those. Whether Intel intend to remove HT from games machines or not is not the point here with P3D. In any case a newly designed game can be built to handle HT and many cores without overfilling the boat.