My version on how to get smooth and stutter free P3D 3.1

[jrs182p 23]

Here's food for thought. I can't get my head around this. I don't know if this will be the same for everyones system but I noticed it immediately. If I set the AM 212 (11010100) in the prepar3d config I can watch the first thread on core 2 pinned at pretty much 100%, core 4 doing pretty much nothing with quiet a bit of activity on 6 and 7.

 

However if I remove the AM from the config and then set the affinity to 212(11010100) in Process Lasso after starting Prepar3d, cores 2/4/6/7 all seem to be distributing the load evenly and all reach 100% in dense heavily populated areas and all evenly come down to <50% when in more remote areas. Some how this almost feels smoother to me.

 

I'm sure there is some testing to be done here but there is definitely a difference in letting Process Lasso set the affinity which confirms Steve's information but is it for the worse? Its hard to tell but it does feel smoother to me using Process Lasso to set the affinity.

[Nyxx 6,993]

Joe

That's very interesting, because after setting up PL last night I had P3D in window mode to watch like you did. I set mine up like the beginning of the topic. So P3D was on it's own.

 

Like you a saw, core 2 @ 90-95% and the other cores etc it used not doing much at all, I was hoping to see like you after you set it in PL.

 

I was using AM 116

[SteveW 1,792]

if I remove the AM from the config and then set the affinity to 212(11010100) in Process Lasso after starting Prepar3d, cores 2/4/6/7 all seem to be distributing the load evenly and all reach 100% in dense heavily populated areas and all evenly come down to <50% when in more remote areas. Some how this almost feels smoother to me.

Your assumption would seem to suggest the sim works incorrectly under it's own devices. However, this is not the case. Unfortunately what is happening is, the sim makes 8 jobs and Process Lasso bunches all 8 onto 4 cores, leading to the loss in performance shown in my graphs with 4 v 5 cores.

 

Remember...

It is best to allow the sim to start with its own affinity configuration in the .cfg. FSX and P3D do not reorganise in the same way if the AM is altered or task manager is used to move jobs.

 

If the application handles it's own affinity then it is best to let it do it, other apps can be run with the .bat I provided, or utilising such apps as Process Lasso.

 

not doing much at all, I was hoping to see like you after you set it in PL.

No you don't want that. That was overloading the core with a second job that if it is job 5 or above, is wasted, and adds heat and a VAS overhead for no reason.

[jrs182p 23]

I see. I wasn't really assuming anything, I was just merely pointing out what I see my CPU doing setting the AM both ways. To be honest all this HT jargon blows my mind and I don't really understand it however your information/knowledge is a great help and I have progressed from using HT=Off with no AM to setting HT on with AM 85 or 212 and it works beautifully for which I have you to thank!.

 

As you stated why spend more on an i7 with HT if your not going to use it when you could just get an i5 without it much cheaper.

 

I currently use AM 212(11010100) 2/4/6/7 in the sim and set all my other addons to use (00101001) 0/3/5. Seems to work nicely.

[SteveW 1,792]

I currently use AM 212(11010100) 2/4/6/7 in the sim and set all my other addons to use (00101001) 0/3/5. Seems to work nicely.

I see, you have a 4 core HT enabled, not an 8 core?

 

But anyway, you missed the point about 116 which has been compared to 212 in the 4 core HT enabled CPU.

 

The reason to use 116 over 212, is that there are no two maxing jobs on two LPs of one core at the same time when the sim is loading terrain as there is with 212. You may see better weather performance as well with 116.

[GHarrall 494]

Are you guys just using the free version of process lasso?

[Nyxx 6,993]

I am Glenn.

[pincopanco 5]

I see, you have a 4 core HT enabled, not an 8 core?

 

But anyway, you missed the point about 116 which has been compared to 212 in the 4 core HT enabled CPU.

 

The reason to use 116 over 212, is that there are no two maxing jobs on two LPs of one core at the same time when the sim is loading terrain as there is with 212. You may see better weather performance as well with 116.

116 and 212 are not similar?

 

116 = 01-11-01-00

212 = 11-01-01-00

[SteveW 1,792]

116 and 212 are not similar?

 

116 = 01-11-01-00

212 = 11-01-01-00

 

Quite. With 212 the LPs 6 and 7 contain jobs that demand maximum throughput. So if they are on the same core together they compete for that bandwidth. Whereas, with 116, LPs 4 and 5 don't compete, since when 5 demands maximum throughput together with 6 they are on separate cores. The job on LP 4 won't be competing with the job on 5 anything like the way 5 and 6 would compete.

[jrs182p 23]

Yes Steve I'm using a 4790k 4 core with HT enabled. As the above post I'm confused about the difference between 116 and 212.

 

Edit, i jsut read your above post and will try 116! thanks again for the advice.

[SteveW 1,792]

If we are grouping 4 jobs on three cores, it is best to put jobs 1, 3, and 4 on separate cores. Two of the jobs have to pair up on one of the HT cores, so which one?

 

Well, we don't want to share the core with job 1 on it, that's the User Interface, and the Renderer. We don't want to share cores with jobs 3 and 4, since they can max out collecting data. So what of job 2? Job 2 can be thought of as managing jobs 3 and 4 and reporting to job 1. Well job 2 can get busy, but when job 3 gets going, job 2 relents. So jobs 3 (nearly) and 4 get full bandwidth. With 116 we put three cores to best advantage.

[pincopanco 5]

Quite. With 212 the LPs 6 and 7 contain jobs that demand maximum throughput. So if they are on the same core together they compete for that bandwidth. Whereas, with 116, LPs 4 and 5 don't compete, since when 5 demands maximum throughput together with 6 they are on separate cores. The job on LP 4 won't be competing with the job on 5 anything like the way 5 and 6 would compete.

Too complicated for me.

I know that two LPs of one core creates stutters and that core 0 disabled creates blurry in my system.

[SteveW 1,792]

If you read through this discussion you'll see that it's mainly interference with other apps that cause stutter. However, irrespective of understanding theory and priciples, 116 has been tested against 212 and the theory holds out well in practice, in 116 favour.

 

Instead as has been said, use 4 cores 4 LPs with 85, and place the addons away from the first sim job. The second sim job can be shared on a core with less impact than jobs 1, 3, and 4 with 4 LPs unmasked.

[GHarrall 494]

I am Glenn.

 

It looks like it does a lot. 

 

Is there a setup guide anywhere? 

[Guest]

My goodness... I came back to this topic to see what settings I had to use to test this all but you all completely lost me... AM 116 seems to be the way to go (i7 4790) but in the cfg or not? Or using the Windows affinity mask option? Or Process Lasso? Or all? Or not even AM 116?

 

Maybe someone could summarize it all in a single and easy to follow post...?