Quoted examples like "AffinityMask=65533 (1111111111111101)" need clarification.

Doesn't show any commas so it's denoting a 16 core CPU, a pair of 8 core CPUs on a board 11111111,11111101=65533

a four core HT OFF 1111=15 = four full on cores

a four core HT ON 01,01,01,01=85 does the same job four full on cores

 

 

8 hours ago, SteveW said:

Using an affinity mask 65533 leaves one LP which you would need to study to be sure it's not filled with the sim process.

If by "sim process" you mean some internal process of Prepar3D.exe then that is not clear there, Steve. Why would he need to study if it (LP1) is filled with the sim process when we know AM=65533 just bars P3D from using that LP1 on his specific 8-core/16-LPs processor at that AM? Unless you meant some sim-related add-on's exe (outside exe.xml), of course, but that would not be a sim process in context.

Thanks,

Edited January 18, 20188 yr by Dirk98

8 hours ago, SteveW said:

In reality the sim works perfectly within the AM we give it.

Not in my case. Some AMs I tried ruined the performance and fluidity.

I'm at a loss here

Thanks.

7 hours ago, Cruachan said:

In the end, all this confirmed was that I was getting the best overall fluid performance with only occasional micro-stuttering while using 8 LPs:

8 = 21845

So, 8 LPs is confirmed as being the sweet spot while using an i-7 5960X in my rig (see sig.)

From previous suggestions mentioned in this thread you should also try AM=21846 / (0101 0101 0101 0110), I still don't know why specifically, perhaps, as a good practice, if I got them right ))

Edited January 18, 20188 yr by Dirk98

7 hours ago, Cruachan said:

Pegging at 30fps in-sim seems to produce best results for most flights (dawn, day, dusk or night) and Unlimited in more taxing areas like busy hubs.

Still Unlimited frames, VSYNC On and 20-25Hz Monitor refresh is unbeatable for me.

41 minutes ago, Dirk98 said:

If by "sim process" you mean some internal process of Prepar3D.exe then that is not clear there, Steve. Why would he need to study if it (LP1) is filled with the sim process when we know AM=65533 just bars P3D from using that LP1 on his specific 8-core/16-LPs processor at that AM? Unless you meant some sim-related add-on's exe (outside exe.xml), of course, but that would not be a sim process in context.

Thanks,

You're making this hard work and not keeping up. It was you suggested the AM and suggested it was filled with activity of the sim even though it was barred. So I am saying how do you know that - you would have to study those processes.

 

27 minutes ago, Dirk98 said:

Not in my case. Some AMs I tried ruined the performance and fluidity.

I'm at a loss here

Thanks.

You can't just muck around with AMs like you are and get results.

5 hours ago, SteveW said:

The gain from HT enabled is that since inevitably there will be many threads per LP, that switching time between them is considerably reduced with HT enabled. 

 

5 hours ago, SteveW said:

...so simply throwing more LPs at it might not work out. Adding more LPs can work with background tasks of the sim but the extra jobs divides the memory performance down. Again we need to rationalise it especially on many-cored machines.

I've added the above 2 abstracts to my P3Dv4 Setup doc. ))

Thanks!

8 minutes ago, SteveW said:

You're making this hard work and not keeping up. It was you suggested the AM and suggested it was filled with activity of the sim even though it was barred. So I am saying how do you know that - you would have to study those processes.

 

 

Sorry, it was somebody else, not me, Steve. 

Ok, you may want to ask AdvancedFollower.

Anyway I can see sometimes you mention "sim process" which means add-on exe  in context.

Thanks.

Well someone suggested it that's the answer.

 

Look what we get from HT:

More work done on concurrency, like it or lump it a load of threads distributed properly finish sooner than on the same CPU with HT disabled. Stuff runs more efficiently together. More heat, there's more work done, work=heat. If we use overclocking we set the overclock to accommodate HT.

 

What we don't get from HT:

More fps. The core runs at the same speed HT on or off. If there's no interruptions to the main sim process (rendering) then we don't see an improvement in fps.

 

Using fps as an indication of making good progress with any AM configuration is useless. When the sim is running properly fps is not an issue, changes to the scene and smoothness of transitions are key. The amount of change to fps I during a repeatable scenario that does not flat-line the sim somehow or we won't see any changes changing AMs. Anyway it's more important to ensure the other apps are kept off the main sim processes unless we got big CPUs like Mike.

9 hours ago, SteveW said:

In reality the sim works perfectly within the AM we give it.

Ok, you must have referred to a specific AM mentioned by AdvancedFollower

 

27 minutes ago, SteveW said:

More work done on concurrency, like it or lump it a load of threads distributed properly finish sooner than on the same CPU with HT disabled. Stuff runs more efficiently together. More heat, there's more work done, work=heat. If we use overclocking we set the overclock to accommodate HT.

Yes, that's what I settled on since started digging HT and going through this thread. I was good with [email protected] no HT before. But after reading valuable info from you and others and after some more of my own observations I quite happily ended up with 4.6GHz HT On and AM=253. I thought I was done with my fine tuning, but then I saw you also mentioned AM=254 11,11,11,10 as a feasible option. I don't expect to see any improvement with it, but I'll just try for the heck of it.

My main concern is my coming AM for i9 7900x HT On. Unlike before I think now that I may start testing with 00,00,01,01,01,01,01,01,01,10 AM and see if the total of 20LPs won't bog down overall performance compared to no HT and some AM that would give Prepar3D 8-10 LPs. Other add-on will also be taken care of.

Thanks.  

Edited January 19, 20188 yr by Dirk98

25 minutes ago, Dirk98 said:

Ok, you must have referred to a specific AM mentioned by AdvancedFollower

 

Sorry was referring to his AM with one zero, yes.

1 hour ago, Dirk98 said:

Still Unlimited frames, VSYNC On and 20-25Hz Monitor refresh is unbeatable for me.

Hi Dirk,

Ah, well, you will note from my signature that my monitor refreshes at 120Hz. I suspect the reason why I see good results may have something to do with the sim's displayed frames (30/s) matching the monitor's output exactly every 4th screen refresh (most of the time). You would think VSync would produce similar results, and it does, but, for whatever reason, those perceived results are not as good. Believe me, I've tried...many times! Also VSync just gets in the way when the sim's frame rate dips below 30/s for extended periods and regular screen updating suffers as a consequence.

I use G-SYNC for everything else that runs in a true full screen environment. This switches on automatically as the application leaves the desktop and off when it returns on exiting. Sadly, with P3D G-SYNC falls down when frame rates drop below 30/s and the sim starts to stutter so, instead, I configure the Monitor Technology as Fixed Refresh in the NCP for P3D and results are quite acceptable.

Mike

Yes MIke, seems you're observing the extra bit of smoothness - that there's very little time between each refresh, irrespective of the GPU output rate. Because when a complete frame is received to the monitor there's little waiting time to display it, the next refresh is coming up in half that of a 60Hz monitor.

When using 30fps or more there's so little time between frames the main thing we notice is rate of change in the flow. I remember 80, 90, 120Hz tvs coming out - I could easily make it out. With the sim we get fps down to 20. As we go below 30 we see the rate of change far more obviously as there's more time between frames. That's where we tune the sim. I can turn scenery object count down and keep 60fps all day on a relatively old machine.