Ray Proudfoot

P3D running but CPU not running at max speed

Recommended Posts

1 minute ago, killthespam said:

Steve,

Would you be so kind and point me again to those pages, please.

Sure:

http://www.codelegend.com/forums/viewforum.php?f=2

 

I'm not presenting ideas there i'm saying how stuff actually works and why we do stuff the way we do.

 

Since the simulator alters its demand on the CPU as the view changes, flying or panning, we need to set the sim up with overhead. Just like when you're in a climb, if you are at maximum throttle and you get a downdraught, your climb rate decreases and you can't increase the throttle. That's how the sim works when you set it up with overhead.

Share this post


Link to post
Help AVSIM continue to serve you!
Please donate today!

So Ray, by setting VSync the GPU output is steered to your refresh frequency so that there is maximum coincidence of frames to monitor rate. Setting VSync=On to 'fix the flicker problem' is not what we are doing. If you study the link I sent it is exceptionally useful to know how to profile the simulator in such a simple way rather than simply add stuff remaining in the dark as you go.

Share this post


Link to post
1 minute ago, killthespam said:

Thank you Steve, much appreciated.

You are very welcome Alex.

Share this post


Link to post

...while we're talking about CPU use, we have the other thing, core usage. I read such stuff in lengthy discussions about careful setting of memory timing, multiplier steps, voltage increases and bus frequencies and then what do we find? What to do with AMs and HT switch off HT and leave out the AM. It's good to understand that side of the equation as well since as far as I'm concerned you can't have one without the other.

The major metric I've explained in the past: Use the AM to add cores until you see no more increase in scenario loading speed, no big decreases in loading time means you need not apply more cores.

With big CPU's those apps providing the AM setting like FSX and P3D do that because they can't determine the partition required by our setup - we all have unique systems.

With a stopwatch time how long the scenario takes to load to a certain point as the sim starts up. Do it a few times to remove the delay before caching.  Add cores and note the times. Do that with your regular setup starting at a major airport.

Every core or LP will queue up for the resources and disk drives and the like can't be spread ever more widely and keep the earlier cores fed - so enable only those needed. Extra work is done and lost by the excess cores utilised so scrapping any chance of a proper performance test.

 

Edited by SteveW

Share this post


Link to post

...then with HT (hyperthreading) it's definitely not cool to mis-understand this technology:

When a thread waits for those resources I just mentioned, that thread is said to go to sleep. In reality the CPU keeps clocking to an extent. those cycles are not fully used, shifting binary words and filling up data structures for matrixes you name it, the core temperature settles down (ignoring CPU throttling) since there's less work done.

Now there's always more than two threads on a core even though they mention two for HT. But using that analogy, the other thread on an HT core still shares those cycles with the other thread, there is after all only one core. However the HT core enables the other thread to start and stop with no fuss, halving the context switching time used as threads are set up on the core for continued execution.

Also, importantly, since the free cycles are now getting used fully, and the workload is being eaten up more quickly, more heat is produced. So instant heat increase with HT enabled is simply that more work is done and less gaps, less sleeping.

With an app like Pant 3D or a video frame renderer, the workload can be more or less continuously split down onto more and more cores. For example a 1000 frame movie can be rendered in an instant on a 1000 core PC, so takes twice as long to run on 500 cores.

P3D has parts only similar to that, and also has a major part that can be thought of as the game-loop that is critically timed. That is, everything happening inside that loop must be accountable for each CPU cycle or performance is lost pretty quickly within the application. Add in a dll or gauge and that critical time is interfered-with.

Remember also when testing and profiling that certain add-ins can take such a proportion from that time in the loop that tests of settings of CPU and GPU make little difference.

Recently desktop CPUs coming out more widely allow certain cores to run at a higher speed. With the AM we can dictate to P3D where to put the first task - that major game-loop task. Then we allow the required amount of Logical Processors to chew on the rest. Just avoid overeating.

 

 

 

Edited by SteveW

Share this post


Link to post

...Leaving HT off so to advantage overclocking is a mistake:

Since that maximum overclock setting is keeping cool because there's sleeping time on the cores, right?

it's only going to be waiting for you to run an app or an app to start a sub-process that utilises more fully that core without sleeping, so that it heats up like an HT core. That's right - crash.

An overclocked system needs to be set in its heat curve to accommodate HT.

Share this post


Link to post

@SteveW, I appreciate your posts and your efforts to educate me on the finer points of getting the best out of P3D. I will look at your forum in slower time when things here have calmed down a bit. I've spent a lot of time on P3D in the last couple of weeks and my ability to absorb new information is just not there at the moment.

My priority is trying to solve the problem of why P3D crashes on first load but not on subsequent ones.

Share this post


Link to post
11 hours ago, SteveW said:

The app is pretty good and with a robust simconnect and other solid features, only really pressing buttons in the sim so in itself is unlikely the source.

If you mean UTLive, yes, I know it isn't that. I can only think one of the aircraft liveries from FLAi is corrupted or bad in some way. To check this I'll restore the UTL repaints.xml and test again.

The other thing which may have contributed to the crash is that the number of AIaircraft at and around EGLL has increased to nearly 600. Far too much for my system. I'm whacking down the settings which must have somehow got too high.

11 hours ago, Ray Proudfoot said:

I almost never run in full screen mode.

Why? With P3D in full screen mode you can still have other things displayed on a separate monitor, as I do. Just make sure the blackout option in the P3D options is off.

Really the only difference then between full screen and windowed is the border. It looks better without a border! 😉

Pete

 

 

Share this post


Link to post
6 minutes ago, Pete Dowson said:

Why? With P3D in full screen mode you can still have other things displayed on a separate monitor, as I do. Just make sure the blackout option in the P3D options is off.

Really the only difference then between full screen and windowed is the border. It looks better without a border! 😉

Pete

It goes back a long time when if you were in full-screen mode and a program crashed the popup wouldn't show because effectively you were in DOS mode. Old habits and all that.

Share this post


Link to post

Pete is right. Some addon AI aircraft may be corrupt.

I had the same. No crashes but a framerate dropping over time...

Share this post


Link to post

I understand UT live is no longer being developed.

Edited by IanHarrison
Underline on.

Share this post


Link to post
2 hours ago, IanHarrison said:

I understand UT live is no longer being developed.

Wrong.

Share this post


Link to post
2 hours ago, IanHarrison said:

I understand UT live is no longer being developed.

Hi Ian,

While this unexpected revelation may be true, I’m sure that I’m not alone in wanting further details before arriving at a judgement. For example, where did you source this info and can you confirm the veracity of the content?

Regards,

Mike

Share this post


Link to post
Just now, Cruachan said:

Hi Ian,

While this unexpected revelation may be true, I’m sure that I’m not alone in wanting further details before arriving at a judgement. For example, where did you source this info and can you confirm the veracity of the content?

Regards,

Mike

He can't because it is not true.

Share this post


Link to post
14 minutes ago, pgde said:

He can't because it is not true.

That may be correct. However, until we have more detail from Ian or you can support your contra assertion then we all remain in a state of limbo. Lack of evidence does not make it true.

Mike

Share this post


Link to post

Unlike stated on codelegend forum P3Dv4 works much better with 11.11.11.01 affinity on a 4-core CPU.  😊

Share this post


Link to post
5 hours ago, Dirk98 said:

Unlike stated on codelegend forum P3Dv4 works much better with 11.11.11.01 affinity on a 4-core CPU.  😊

Anyone can tell you that's a typical AM as mentioned in my many posts and pages related to it. Only someone that doesn't understand would say something like that Dirk. Also your test method is not clear and the result of performance could be from any kind of set of factors - lol.

Edited by SteveW

Share this post


Link to post
15 hours ago, Ray Proudfoot said:

@SteveW, I appreciate your posts and your efforts to educate me on the finer points of getting the best out of P3D. I will look at your forum in slower time when things here have calmed down a bit. I've spent a lot of time on P3D in the last couple of weeks and my ability to absorb new information is just not there at the moment.

My priority is trying to solve the problem of why P3D crashes on first load but not on subsequent ones.

Cheers Ray, I understand. However, I still feel you may missed my point entirely in that by applying VSync you limited the CPU throughput giving you rise for concern that you no longer saw near 100%. Because you saw the flicker and found applying VSync stops that behaviour.

To your other concern - did P3D load OK every time before you made any adjustments or added in any stuff?

Share this post


Link to post

I know you're only on a wind up Dirk lol, but let's go over the rules:

Make sure the AM excludes one LP from the first core met by the app:
 

so 11,11,11,01 seems good with 01 on the right - check.

Use the rest of the LPs to limit them to what's needed.

OK well you're limited by the chip with only three more cores left so with a bigger CPU you might do that.

 

So where's the problem? I see none.

Edited by SteveW
  • Upvote 1

Share this post


Link to post
5 hours ago, Dirk98 said:

P3Dv4 works much better with 11.11.11.01 affinity on a 4-core CPU.

By the way - doesn't this say at least that the effort to check out the use of AMs was worth it? At least with that setup. it is possible that can be improved on but it would be on a per system basis. Otherwise that's typical of the recommendation I've been making from when multicore arrived.

  • Upvote 1

Share this post


Link to post
5 hours ago, Dirk98 said:

Unlike stated on codelegend forum P3Dv4 works much better with 11.11.11.01 affinity on a 4-core CPU.  😊

Yes - pretty obviously this comment is very annoying - since you well know that almost single handedly I've been trying to champion this kind of understanding into the community. The AM's on the site don't represent the best or worst setting - they represent examples of setting. So I wouldn't mind if you retracted this statement.

  • Like 1
  • Upvote 1

Share this post


Link to post
7 hours ago, Cruachan said:

Hi Ian,

While this unexpected revelation may be true, I’m sure that I’m not alone in wanting further details before arriving at a judgement. For example, where did you source this info and can you confirm the veracity of the content?

Regards,

Mike

As far as I know the project has not been terminated but sometimes things get in the way of continuous development cycles.

  • Upvote 1

Share this post


Link to post
8 hours ago, SteveW said:

Cheers Ray, I understand. However, I still feel you may missed my point entirely in that by applying VSync you limited the CPU throughput giving you rise for concern that you no longer saw near 100%. Because you saw the flicker and found applying VSync stops that behaviour.

To your other concern - did P3D load OK every time before you made any adjustments or added in any stuff?

I monitored CPU usage using CPU-Z and on all cores the max was around 4.4 or thereabouts. Are you saying that is throttled back with VSync enabled? I temporarily disabled VSync and Triple Buffering expecting CPU usage to rise but it didn't. Why? Turbo Boost is enabled in the BIOS so why didn't CPU speed rise to 5.0?

I wonder what will happen later this week when I swap my 1920*1080 display for a 3840*2160 one. I'm not expecting fps to remain around 60 (airborne) but I wonder how much they will change. The monitor (BenQ PD3200U) does have a 30Hz option so that could be helpful.

To answer that question it's hard to recall because I had such a horrendous time trying to get MyTraffic6 to run and it would cause P3D to crash before the free flight screen. But I did have GoFlight, FSUIPC5 and ChasePlane entries. Oh, plus the one AS16 adds. But that's it. Nothing else.

I removed all entries in DLL.XML and EXE.XML bar the default single entries and P3D loaded okay. I rebooted and added FSUIPC5 to  DLL.XML and P3D crashed before the free flight screen. So logically FSUIPC5 was the cause. But I know it isn't. I restored my original XML files and rebooted. Crash before free flight screen.

I waited for everything to close and launched it again. Free Flight screen shows fine and P3D runs fine. I'm struggling for a solution and for now just accept P3D will crash on first launch but thereafter will be fine.

Share this post


Link to post
9 hours ago, SteveW said:

Anyone can tell you that's a typical AM as mentioned in my many posts and pages related to it. Only someone that doesn't understand would say something like that Dirk. Also your test method is not clear and the result of performance could be from any kind of set of factors - lol.

I'll try it then, sounds intriguing!! :biggrin:

 

Edited by Dirk98

Share this post


Link to post

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now