27 minutes ago, Ray Proudfoot said:

Mike,

I didn't have the enthusiasm to read through 13 pages. But glancing through there doesn't appear to be much input from LM support.

I can live with my situation. I may even be able to increase one or two settings. I think having the 1050 card benefits me because I can revert to earlier drivers which don't have a VAS leak. Owners of the 1080 can't use those as the card is not supported.

Hi Ray,

Oh they are aware of the problem... big time! Infrequent posts from LM are the norm, I'm afraid, but they do from time to time when they have something useful to say to us.

Regarding the 1050 vs 1080, yes that is my understanding as well. However, your sig is introducing an element of confusion regarding the GPU actually in use.

And now, to lighten the mood a little, for something completely different

My wife spotted this image and sent it to me knowing I would appreciate it. Unusual lenticular cloud formation resembling the shape of......

Regards,

Mike

Thanks Mike. Silly me. Sig altered. I think the problem at P3D Forum is people don't always give enough info to help with the problem.

Yes, I saw that cloud shot today too. Eerie. Just missing the livery. 

I did a flight with Concorde-X in P3D v3.4.22 along the same route as in the video from KIAD to KJFK, this time with:

LOD=6.5
Driver 376.33
Everything activated in ORBX Vector (except frozen lakes)
ORBX Global + OpenLC US
Washington X from Drzewiecki Design
Other settings are the same as in the video

Here is my VAS profile (the "buckle" in the middle captures the flight over the Washington X area). The peak VAS was around 3650 MB after landing at KJFK, which I think is a very good value considering the settings and add-ons used. Personally, if you use a driver that doesn't leak VAS, I don't see how there is a problem with VAS.

Thanks ab. 376.33 is key to decent performance. Cutting it a bit fine with only 350Mb free at JFK. Is that the FSDT one? How many Ai do you have in FSUIPC?

I am using the default KJFK airport with AFCAD from MyTraffic 5.4b (which explains why the VAS consumption was not higher, but even with the default airport, the NYC area is quite demanding). I don't use FSUIPC to limit the air traffic. The maximum AI traffic per session recorded by FSUIPC log was ca. 540.

You can save a fair bit of VAS by limiting Ai to a lower number. Where there any free gates at JFK? 

2 hours ago, Ray Proudfoot said:

376.33 is key to decent performance.

You can´t say that in general as the recent drivers work properly on other systems like they do on mine. And I´m one of those who stuck with 376.33 for half a year for well known reasons.

I´ve now just completed a flight from Dubai (FlyTampa) to Vienna (FlyTampa) with a 777-300 (PMDG) using 100% AI traffic, LOD radius 6.5, and maximum scenery and autogen density.
This is how my VAS has been handled:

This graph was recorded from entering the virtual plane until leaving it at the gate with 1GB free VAS after a flight of more than 6 hours.

I honestly never had a sim working that smooth and I´m on latest GeForce 382.05. As I wrote before I have HT switched off, disabled the shadercache, disabled multithreading of the nvidia-driver, disabled optimize_parts ans so on... I already wrote most of the settings I use in previous posts.

You´re complaining that P3D doesn´t release VAS well enough. Well... I suppose it does...  Let me explain my guess...
I think that the P3D-engine simply doesn´t unload objects which anyway most likely have to be loaded again later on a flight and that behavior might be for performance reasons. Why unloading and reloading objects from the harddisk (especially autogen!) if there is anyway plenty of memory available? A constantly unloading and reloading of exactly the same objects would not be efficient and definetly at the cost of performance if there anyway was enough memory available.
After I left Dubai across the persian gulf about 300MB where released from memory. That´s a least not too bad and considering that FlyTampa´s Dubai is mainly based on customized scenery objects (which are very unlikely to be loaded again on the same flight) instead of repetitive autogen (which you have plenty of in the area of London btw.) that could be exactly the according amount of memory released after departure.

On the other side if the sim is starting to run closer to the risk of an oom my experience is that P3D 3.4 starts releasing VAS much more aggressively at this point. It still does that far away from perfect but it seems to me that the system only unloads common objects like autogen, ground textures, vegetation or whatsoever when an imminent lack of memory starts forcing it to do so.

All I actually wanted to say... That insufficient release of VAS you´re complaining about (although you still have plenty of VAS left in your scenarios) is eventually not a bug... It might be a feature ;-)

Just some quick advise, all the talk of this VAS issue, as I'm still running V3.4 with none of the hotfix's would it be a good idea to update to the latest hotfix? I mean would it be advantageous or not? as I've ever only updated the client since 3.2, I don't really recall getting any OOM's except for using the FSLABS A320-X which happen's all the time (I don't think I've ever completed a flight in it, but thats another story).

Cheers Rod.

"All I actually wanted to say... That insufficient release of VAS you´re complaining about (although you still have plenty of VAS left in your scenarios) is eventually not a bug... It might be a feature ;-)"

not a feature , As Designed...I said in page 3 of this post...

"I would say VAS management is very good , it never OOM when the VAS usage was getting high..."

at page 6 we are still not having any OOM's..LM has done a great Job , even when impaired by bad Nvidia drivers...

Jorge

On 7.5.2017 at 2:31 PM, Ray Proudfoot said:

No, not yet. But it seems awfully close to doing that. Maybe something magical happens when VAS < 200Mb.

First, if i say you in the following lines, i dont mean only YOU as the one i replied to, i mean you guys that seek hours for solutions.

So you spend so much time for something that never happened to you? Are you so scared of an OOM?

Sometimes i wonder why people dont see the limitations, really. They rather deactivate stuff after the start midflight, a total immersion breaker imo (im not kidding, that was the last idea someone had in a famous streamers chat last week, "deactivate the airport one started on, after the start, to save VAS) than just LOWER your scenery settings. Is this such a hard thing to do? And that can be the only thing that really affects VAS. That and maybe traffic, i dunno.

Since i dont have a monster machine i have to play with lower scenery settings and guess what, i NEVER saw a VAS usage over 2.5 GB in the projectfly overlay. And guess more, i have tons of payware airports, planes like PMDG 737, 777, Majestics Dash8 and what not, running ORBX global with things like openLC Europe, trees, a weather engine (FSXWX), among others running at the same time.

Added ontop are the other tools like FScloud, smartcars and all the stuff needed, what of course doesnt go into the 4 GB anyway, as long one doesnt only have 4 GB ram installed.

What i want to say with that. It only the users own fault, not p3d developer nor other developers. Of course, some addons are better on the subject than others. But thats not the point. Just reduce your damn scenery settings or are the 5 more buildings so important that you have to spend hours and hours and hours to solve a problem that is basicaly not even existing if you just pull that triggers to the left....

Why have to have scenery settings on max? Why you have to put AI traffic on airports when on Vatsim anyway, there are so many things one could do to NOT have this problem. I understand you want immersion, but is it really immersion to have OOMs (something that very few ppl experience anyway, its just Paranoia, like we see here) or talking endless hours about VAS? Really? 

Sometimes this makes me not want to watch twitch streamers anymore. Only VAS VAS VAS VAS. But then they run 20 AI planes on the airports, 10 more from Vatsim, have car traffic like its the worlds end (one time in a stream i saw like 500 cars driving with like 200 km/h on the street near airport, fully unrealistic but whatever...).

Its basically very simple, you dont have to go into configs, dont have to spend tweaking, you just have to make priorities. What is more important? Constant fear, hours lost just for a few more houses and a view that lets you see the houses 1km earlier?

Or a sim that runs fluid, very fluid even on lower systems, with all the known addons, never worry about  a ctd and just a few houses less and maybe the AI traffic to like 20% only? 

I know what i would choose. But to each his own i guess.

I member the day that the three Concords flew over my workplace in Hayes uk going to land at Heathrow, they looked Brill and I was PO that I had left my camera at home ,sorry nothing to do with VAS ,but  just had to post

Peter

 

7 hours ago, Schlotterknie said:

You can´t say that in general as the recent drivers work properly on other systems like they do on mine. And I´m one of those who stuck with 376.33 for half a year for well known reasons.

I´ve now just completed a flight from Dubai (FlyTampa) to Vienna (FlyTampa) with a 777-300 (PMDG) using 100% AI traffic, LOD radius 6.5, and maximum scenery and autogen density.
This is how my VAS has been handled.

This graph was recorded from entering the virtual plane until leaving it at the gate with 1GB free VAS after a flight of more than 6 hours.

I honestly never had a sim working that smooth and I´m on latest GeForce 382.05. As I wrote before I have HT switched off, disabled the shadercache, disabled multithreading of the nvidia-driver, disabled optimize_parts ans so on... I already wrote most of the settings I use in previous posts.

You´re complaining that P3D doesn´t release VAS well enough. Well... I suppose it does...  Let me explain my guess...
I think that the P3D-engine simply doesn´t unload objects which anyway most likely have to be loaded again later on a flight and that behavior might be for performance reasons. Why unloading and reloading objects from the harddisk (especially autogen!) if there is anyway plenty of memory available? A constantly unloading and reloading of exactly the same objects would not be efficient and definetly at the cost of performance if there anyway was enough memory available.
After I left Dubai across the persian gulf about 300MB where released from memory. That´s a least not too bad and considering that FlyTampa´s Dubai is mainly based on customized scenery objects (which are very unlikely to be loaded again on the same flight) instead of repetitive autogen (which you have plenty of in the area of London btw.) that could be exactly the according amount of memory released after departure.

On the other side if the sim is starting to run closer to the risk of an oom my experience is that P3D 3.4 starts releasing VAS much more aggressively at this point. It still does that far away from perfect but it seems to me that the system only unloads common objects like autogen, ground textures, vegetation or whatsoever when an imminent lack of memory starts forcing it to do so.

All I actually wanted to say... That insufficient release of VAS you´re complaining about (although you still have plenty of VAS left in your scenarios) is eventually not a bug... It might be a feature ;-)

I should have said it works for me. I wasn't saying everyone should use those. I'm staggered you can have so much free VAS remaining with maximum autogen and scenery and 100% Ai. What Ai package are you using and do you limit the Ai number via FSUIPC?

I've searched Rob Ainsclough's excellent guide for hyperthreading and there's no mention if it. Curious. All 8 of my virtual cores are working so why would I change that?

My earlier flights did have some VAS remaining but on a later flight when using 27L at EGLL I flew over central London and that killed it for me. I've dialled back some settings that were clearly too high so I'll see how I get on now.

3 hours ago, Ray Proudfoot said:

I've searched Rob Ainscough's excellent guide for hyperthreading and there's no mention if it. Curious. All 8 of my virtual cores are working so why would I change that?

Hi Ray,

Are you seeing fluid performance (not talking frame rate here)? If so then stick with what you are doing because, in the end, that's all that really matters.

Whether or not to use Hyperthreading depends on what you want from your CPU. If the CPU is a 4-cored model and you are not intending to overclock then HT=ON is probably the way to go as you will see best performance. If you overclock then you will require a decent cooling solution, preferably water, as there is significant generation of heat. If you have a high end multiple cored (6-8) CPU, intend to overclock and don't mind the slight performance penalty then turning off HT can be beneficial. I have overclocked my 5960X by 41% to 4.4GHz and, while monitoring with the Core Temp utility, max temps on all 8 cores never exceed 60C. In P3D temps are usually significantly less. Aerofly FS2 and X-Plane 11 seem perfectly happy with HT=OFF as do all my installed games. Frame rates are far higher than I will ever need for the forseeable future so is not a major consideration. Fluidity of performance free of stuttering is the key.

Rob has run exhaustive tests with Prepar3D using 6 and 8-cored CPUs and, in the 8-core case, like me, has found HT=OFF works best. I believe he has now turned it back on, but in conjunction with a P3D AffinityMask setting of 21845 (01,01,01,01,01,01,01,01) so is still 'restricting' P3D to 8 physical cores. His results are all tabulated here (scroll to the bottom of the page):

http://www.robainscough.com/P3D_Performance_Analysis.html

AffinityMask

Whether or not you intend to use an AffinityMask setting with Hyperthreading=ON will depend on current performance and the decision to run various .exe addons alongside P3D. The objective is to jobschedule these addons in such a way that they avoid the main P3D thread which is already exercising one of the CPU cores quite heavily. On an 8-cored CPU with HT=OFF this seems to be less of an issue. Not sure why this is the case, but my experience is that being able to use 8 physical cores (8C, HT=OFF) rather than 8 logical cores (4C, HT=ON), as on a 4-cored CPU, appears to have its advantages in this respect. I have further complicated the picture by throwing an illogical AffinityMask setting of 21760 into the mix which, somewhat intriguingly, appears to randomise the jobscheduling core allocation of the P3D main thread each time the sim is run. However, that needn't concern you.

If you are running with HT=ON and intend to run various .exe Addons alongside the sim then you might consider using an AffinityMask. This will help to reduce any stuttering which might occur while there is contention for the main thread. On 4-cored CPUs the two values which have proven to be effective are 85 and 116. However, these are not set in stone and other values can also be tried. Again, if you would like to learn more then read Steve Waite's excellent article:

http://www.codelegend.com/forums/viewtopic.php?f=2&t=532

Regards,

Mike

Hi Mike,

I have a 4770K overclocked professionally to 4.4Ghz with Corsair water cooling. HT is set to 255. I'm assuming HT is on in the Bios but I've never checked. I keep away from things I don't fully understand especially the Bios.

Performance is smooth and FPS are great.  I see no need to tweak this aspect of my computer.

I'm still installing scenery and need to fly more routes so I can judge if settings are okay for hungry areas. Once everything has settled down I can take some time out to tweak a few things but by then the 64-bit version of P3D will be out and all this can be put in the dustbin. 

Why is LM locking vas threads?