P3D multicore usage anomoly

[SledDriver 39]

7 minutes ago, SteveW said:

So AM=63=111111 six cores. what can that do better than the four core PC AM=15=1111?

It can load the scenario faster. And with that shows that it can load whatever, faster.

That's it. There's no hike in fps although there can be a small improvement with an optimised back end.

I'm still undecided whether maintaining one core for all my weather engine stuff is a good allocation of resource. I've watched my system monitors carefully while the sim flies and the core which is doing weather rarely gets above 15% utilisation.

 

[SteveW 1,792]

I don't see much tweaking or messing with AMs, I don't see any disagreement in my mail. I can do a good reliable setup quick, very little to do in my opinion. Details on my site.

If you push the sim to near max then expect to have to alter sliders, that's the choice we accept or turn it down. When the system can more than match the requirements of the sliders, will the next sim be able to max that system out?

[SteveW 1,792]

And on the AMs HT cores debate.

Ask yourselves what your mate does with a four core sim, he most likely uses no AM which means AM=15=1111 (we always have an AM it's not turn-offable, unless software uses AM=0 for a special case other than all ones).

What does your other mate do with 32 cores? What is the result with AM=15=1111 on his 32 core monster? Is it the same as your other mate with the four core?

What if he uses AM=30=11110? Still four cores but one core along. Any difference there?

 

 

[SledDriver 39]

3 minutes ago, SteveW said:

And on the AMs HT cores debate.

Ask yourselves what your mate does with a four core sim, he most likely uses no AM which means AM=15=1111 (we always have an AM it's not turn-offable, unless software uses AM=0 for a special case other than all ones).

What does your other mate do with 32 cores? What is the result with AM=15=1111 on his 32 core monster? Is it the same as your other mate with the four core?

What if he uses AM=30=11110? Still four cores but one core along. Any difference there?

 

 

Not quite sure what you are driving at here.

[SteveW 1,792]

16 minutes ago, SledDriver said:

I'm still undecided whether maintaining one core for all my weather engine stuff is a good allocation of resource. I've watched my system monitors carefully while the sim flies and the core which is doing weather rarely gets above 15% utilisation.

 

Sorry just noticed. Use two or more cores, or two LPs even if on the same core is better.

[SteveW 1,792]

Just now, SledDriver said:

Not quite sure what you are driving at here.

Not completely straightforward?

[SteveW 1,792]

...With even small % showing, they can affect the wrong task no matter what % shows in Task Manager. Often an improvement can be had when we corral addons away from certain cores and use two at least per addon. Apps restricted to one core periodically wait for themselves.

[SteveW 1,792]

AMs and cores? Those were examples in reply to Bert's post.

Answers a few questions.

Example

Q. Do I use no AM?

A. You always use an AM. If you have a four core and do not specify an AM in cfg, then you use 1111, an AM=15 four cores. If you have a six core then that is 111111=63.

With enough know how you can determine that with HT enabled on the four core you need AM=01,01,01,01 or AM=85 to get the same result in the sim as no AM with HT off.

The difference being the way the other apps roll out across double the CPU count. Probably appearing on Mythbusters some time.

 

[SledDriver 39]

16 minutes ago, SteveW said:

...With even small % showing, they can affect the wrong task no matter what % shows in Task Manager. Often an improvement can be had when we corral addons away from certain cores and use two at least per addon. Apps restricted to one core periodically wait for themselves.

>and use two at least per addon. Apps restricted to one core periodically wait for themselves.

...which is exactly the opposite theory proposed for P3D.exe by many here, which some suggest is much worse off on more than one core...

It's like people are just making this stuff up.

[Noel 4,483]

27 minutes ago, SledDriver said:

😀 None of you are hearing what I am really saying.

Anyway. Apparently no-one agrees with me and you all think its fine to spend so much time tweaking.

I gave up tweaking for 2 years Sled...until I visited this thread 

It doesn't seem hard anymore as it all is very predictable.   With my hardware I can't linger in dense metro areas in super complex planes, so I have to be careful to choose my aircraft, departure and arrivals carefully so I can maintain very high sliders and not suffer from lack of headroom.  I have thrown down the gauntlet to developers to try to develop a realtime modulator for all of those features that have potential to be modulated to maintain a certain level of performance because you're right LM certainly has not done this with P3D.  I know things like shadows can be modulated w/o a complete redraw, but I'm not sure what else can.   It would be so nice to be able to configure the utility to maintain a frame rate of 30 and a CPU utilization maximum of 90% and a maximum GPU utilization of 90% at all times by modulating a palette of attributes, prioritized by the user:  shadow distance, autogen density and LOD radius, AI traffic, etc.  This would then, if done well, become essentially a set it and forget it situation.  I'm shocked there isn't one already but alas perhaps the core engine just isn't cut out for it.

[Noel 4,483]

4 minutes ago, SledDriver said:

>and use two at least per addon. Apps restricted to one core periodically wait for themselves.

...which is exactly the opposite theory proposed for P3D.exe by many here, which some suggest is much worse off on more than one core...

It's like people are just making this stuff up.

Sled do you have ultra smooth performance?  What exactly are you looking for still I've forgotten.

[SledDriver 39]

15 minutes ago, Noel said:

I gave up tweaking for 2 years Sled...until I visited this thread 

It doesn't seem hard anymore as it all is very predictable.   With my hardware I can't linger in dense metro areas in super complex planes, so I have to be careful to choose my aircraft, departure and arrivals carefully so I can maintain very high sliders and not suffer from lack of headroom.  I have thrown down the gauntlet to developers to try to develop a realtime modulator for all of those features that have potential to be modulated to maintain a certain level of performance because you're right LM certainly has not done this with P3D.  I know things like shadows can be modulated w/o a complete redraw, but I'm not sure what else can.   It would be so nice to be able to configure the utility to maintain a frame rate of 30 and a CPU utilization maximum of 90% and a maximum GPU utilization of 90% at all times by modulating a palette of attributes, prioritized by the user:  shadow distance, autogen density and LOD radius, AI traffic, etc.  This would then, if done well, become essentially a set it and forget it situation.  I'm shocked there isn't one already but alas perhaps the core engine just isn't cut out for it.

Exactly.

There is NO reason the software cannot maintain and modulate a set of weighted criteria to make its' best effort at achieving a reasonable target set by the user, and thereafter why the software should not learn from it's failures and update those criteria.

Further, there is no reason that software cannot also maintain data for each geographical area visited and get better at delivering those areas each time it is re-visited.

Same for weather scenarios, different planes, etc etc etc.

This is what good programming does.

LM have at best been entirely lazy in their first 10 years with the product.

I say that as an experienced developer and senior consultant on commercial graphics (including 3D) software design and implementation for many years.

Having only recently personally moved from FSX to P3D, I was initially pleased with some of the improvements in P3D, but ultimately a bit shocked and surprised at the chasm of development which has just been ignored. It's a great shame, and unfortunately, the perhaps somewhat technically inexperienced acceptance of what little real work has been provided by LM in now 10 years of supposed development is at best, very disappointing.

 

Edited April 23, 2019 by SledDriver

[SledDriver 39]

15 minutes ago, Noel said:

Sled do you have ultra smooth performance?  What exactly are you looking for still I've forgotten.

I can achieve virtually smooth performance in most situations, but I have yet to do that with traffic enabled, and I like traffic as it adds greatly to the sense of immersion.

And it's not the traffic itself, which I can display smoothly, but the horrid load lags as I approach a new scenery area, which basically stall the sim for up to 2 seconds.

And added to that, it's the inability of the sim to adapt to even the simplest change of circumstances (like a different plane or some clouds) - and changing circumstances are exactly what a sim should be able to deliver.

Do bear in mind this is LM we are talking about, not some back street, unfunded sweat shop.

Edited April 23, 2019 by SledDriver

[itsjase 52]

4 minutes ago, SledDriver said:

Exactly.

There is NO reason the software cannot maintain and modulate a set of weighted criteria to make its' best effort at achieving a reasonable target set by the user, and thereafter why the software should not learn from it's failures and update those criteria.

Further, there is no reason that software cannot also maintain data for each geographical area visited and get better at delivering those areas each time it is re-visited.

Same for weather scenarios, different planes, etc etc etc.

This is what good programming does.

LM have at best been entirely lazy in their first 10 years with the product.

I say that as an experienced developer and senior consultant on commercial graphics (including 3D) software design and implementation for many years.

Having only recently personally moved from FSX to P3D, I was initially pleased with some of the improvements in P3D, but ultimately a bit shocked and surprised at the chasm of development which has just been ignored. It's a great shame, and unfortunately, the perhaps somewhat technically inexperienced acceptance of what little real work has been provided by LM in now 10 years of supposed development is at best, very disappointing.

 

Gonna say from all your other posts you really don't sound like an "experienced developer", or if you are, you seem to be way out of the loop.

Having dynamic adjustments in real-time graphics isn't the easiest thing to do, otherwise all 3D games would be using them. Sure you can implement dynamic resolution, LOD, shadow quality etc to lower details, but this will be a reactive process, which will induce some stuttering or frame-time spikes. To implement this in a proactive way would require a re-write of the whole rendering pipeline and even then it isn't simple or it would be widespread in games today.

Maintain data? Sure if there were no addons installed, and no weather, and a fixed time of day, then yeah maybe this would be feasible.

You make it seem like LM is the only dev that doesn't implement all these features you ask for, when literally nobody does. And those that have tried haven't been very successful.

I don't know who you've been consulting and what their budget is, but no offense I don't think you have any idea what you are talking about.

 

[Noel 4,483]

7 hours ago, itsjase said:

Maintain data? Sure if there were no addons installed, and no weather, and a fixed time of day, then yeah maybe this would be feasible.

You make it seem like LM is the only dev that doesn't implement all these features you ask for, when literally nobody does.

P3D, FSX and prior versions and Xplane all allow users to set sliders such that the sims become un-flyable.  This is obviously a very good thing because it accommodates hardware upgrades, while at the same time setting folks up for disappointment and frustration!  I still believe that a real-time modulator (you're only changing parameters already created and user-controlled thru sliders etc) seems doable, at least could have been had the ESP engine been created w/ this in mind.  This is what Sled appears to be asking for and not a whole lot more!  All of you folks out there including SteveW and itsjase:  do you understand why something like this is out of the question?  Is tweaked ESP engine simply unable for technical reasons we don't appreciate?  

One more time itsjase, SteveW, RobA, Luke, all of you devs answer this absolutely reasonable question as no one here has really touched it, simply put:  why the heck isn't a real-time user-configurable utility possible?

"It would be so nice to be able to configure the utility to maintain a frame rate of 30 and a CPU utilization maximum of 90% and a maximum GPU utilization of 90% at all times by modulating a palette of attributes, prioritized by the user:  shadow distance, autogen density and LOD radius, AI traffic, etc."

????????!!!!!