A first glimpse at Prepar3d v2.3

[Guest]

 

 

How did you determine that Rob?

 

I use Adobe PP CC 2014 provide detail video analysis (frame by frame) ... the original video wasn't exactly 60 fps (this other was) ... if you use frame blending or smoothing be sure to turn it off.  But it's a difficult task to demonstrate stutter free ... so many variables at work including what is being used at the "client" end to view the video ... video players often have settings along with video playback settings in the GPU driver.  For example Windows Media player has an options for "Video smoothing" ... "drop frames to keep audio in sync" ... and GPU driver settings, etc.

 

QuickTime will also present video differently.  YouTube streams can also skew what is actually happening.  Heck, I've recorded video using ShadowPlay and it showed more stutters than I actually experienced while flying.

 

And then finally we have the capture process, Shadow Play, Fraps, DXTory, etc. etc ... they all have their own spin on how to capture the source when recording and even during the recording process frames can get dropped producing a stutter that was never there.

 

But just to be clear, I agree that if you have a 60Hz monitor you want to hit a solid 60 fps (as in that's lowest for entire flight session) ... however, where I want to extend that beyond those "Vsync types of stutters" is the 30 fps you locked at in your video is what I would consider "stutter free" because it presented a consistent pattern of 1 frame every 2/60ths of a second ... you were able to have that consistency because your hardware/settings are able to achieve a solid 60 fps so no issue with being able to "keep up".  Going back to my original point about 1/60 - 60/60 and the consistency of frames.  

 

View port is also very important in the demonstration of the inconsistency in frames ... if you record video flying straight and level at 500 Kts looking out front view the stutters aren't as obvious because there is less actual object motion (physical pixel space motion) vs. when you look to the side you see more stutters because the relative motion of those objects have to traverse more physical pixel space.

 

To demonstrate that point, you could make another video using the same settings to get your solid 60 fps and fly the world at 500 Kts at AGL 500 ... forward view may seem very fluid, but turn to a side view and you will clearly see that a tree or building is NOT moving 1 pixel per frame draw, it's moving 10 or 20 or 30 pixels per frame draw.  Also speed relative ... travel at 40 kts will tend mitigate stutters vs. travelling at 500 Kts.

 

That's why I get comments from people saying "I see more stutters at the sides" ... the stutters are actually unchanged ... they are the same stutters from my 1/60 - 60/60 sample ... BUT, because those buildings/objects are now on the side they have to cover more physical pixel space so a stutter is much more noticeable in that situation because if it would normally cover 10 pixels (at 30 fps) before next frame whereas a stutter is making that 20 pixel or even 30 pixel difference before one sees the next frame.  That's why when looking at the side (assuming straight and level flight) the stutters are more apparent.

 

To me, the "real" stutter solution is trying to figure out how to provide even distribution of frames over a 1 second.  Finding out what's the cause of not have this even distribution of frames?  I believe LM are looking at this, hence the "long frames" reference.  Stutters is by no means a P3D exclusive, many other titles exhibit the same issues from DX9, DX10, DX11.  In many cases stutters are often solved by GPU driver updates ... and yes we're back to getting a profile from nVidia.

 

IMHO, I believe the only way stutters are going to be significantly reduced is with help from nVidia.

 

 

So for me i stay on nVidia with that being said by AMD

 

Yes, nVidia is probably a safer option for P3D right now ... too bad, because I was taking a serious look at the AMD R9 295X2.  Over the years I go back and forth between AMD and nVidia.  

 

Cheers, Rob.

[stryder011 2]

+1

++1

[sascha 314]

Just curious, why FS9? If visuals are not important ... FS8 would be a better option. I'm not sure I'd equate someone's level of "seriousness" about a simulator with "eye candy"

 

Indeed - I wouldn't equate "simming seriousness" with visuals. I was just listing some examples of serious simmers who do without ORBX products for various reasons, because the post that got me riled implied that because most serious simmers would use ORBX, anyone not wanting to use it would not be a serious simmer.

 

FS9 offers a good balance between visual experience and the ability to run multiple monitor setups with lots of peripherals so that serious va pilots can fly smooth approaches into complicated airports, which they can't with FSX, or FS8 I assume.

 

ORBX doesn't come into the equation for them at all, which was the point I was trying to make.

 

 

 

Anyway, point being, just because "eye candy" is not important to someone, that doesn't qualify that person as more or less serious about flight simulation.

 

That was actually what I was trying to say in so many words, but I didn't manage as elegantly as you!

 

Cheers,

Sascha

[peppy197 165]

How did you determine that Rob?

 

This is a very cogent point! We are not talking FSX anymore.

The accepted wisdom for a long time has been the Nvidia is the go to card for FSX.

But P3D V2 is not FSX. Is there any informed opinions about which of the 2 better option

for P3DV2.X? It is a different graphics engine. Right?

 

I remember discussing with a serious geek whether in general Geforce or Radeon was superior.

Although Nvidia in benchmarks generally seemed to have the upper hand, he swore blind that the

game play experience with AMD card was consistently, what for it! "Smoother" than it was with

Nvidia cards. And just in case anyone didn't get that he meant and told me that in general

gaming it was not unusual to see micro stutters with Nvidia card even at 60FPS where the equivalent

AMD card delivered perfect smoothness despite performing less well in benchmarks!

 

I use an R9 290 which with its 4G of VRAM allows me to get to 3.6G usage of VRAM almost every flight. Now if I had a 3G NVIDIA card doing the same flight, there would be a lot more visible instances of wait/loads/etc... causing even more stutters than can be caused by an insufficient NB or IMC frequency match

 

al

[niyoko 7]

The stutters didn't happen in the video but I saw a lot of shimmering and a very light vehicle without any gauges. 

 

My main concern about P3D v2.x is that up until now there hasn't really been a native HEAVY addon like a PMDG T7 was for FSX....Everyone is flying around with a default plane during these tests....In FSX I can get 70 frames with no stutters with the default plane.

 

Do these test have premature written on them even a little? 

 

Majestic Software's Dash 8 Q400 Pilot & Pro versions are a fairly systems complex add-on for FSX and Prepar3D. Granted the systems in a Q400 are not on the scale of a 737 or 777, but the Q400 has the same quality of model and texturing that you get in a PMDG product. The Q400 also visually simulates icing on the plane's surfaces and fogging of the cockpit glass, seen in A2A planes.

 

In Prepar3D is just as good if not better if you include how it runs with Prepar3D's HDR, shadowing, and FXAA. The weather radar and terrain avoidance mapping works great as well.

[denali 74]

To me, the "real" stutter solution is trying to figure out how to provide even distribution of frames over a 1 second.

 

 

That's is what is needed to not "offend" the sense of continuity.  It won't seem fluid at 30, but will won't be disturbing.  The eye is keen to "change", even if it is tempo.

 

 

IMHO, I believe the only way stutters are going to be significantly reduced is with help from nVidia.

  This is probably true, in that they are the only one's who can provide the "slack" which game developers can use to clean up their timing.  If NVidia causes any bad/long timing/slices/frames, that's not something a gamedev is going to be able to clean up, and it may make the difference in the frame getting in.  Even if they are consistent, if they are not tight, it could throw off what gamedevs expect, or are hoping for.

 

I believe NVidia has a hardware tool for developers.  It attaches to a monitor port and gives metrics on the frame timings. "long frames" are one of the things it will measure, which is the core "cause" of microstutters.  It came out last May or June IIRC.

 

Hardware reviewers are Finally beginning to look at metrics such as long frames and microstutters over FPS.  This is one area that NVidia has shown a distinct lead in over the past several years; whereas AMD has been winning in the FPS category, NVidia is finally proven to be much smoother with the new ways of measuring GPU quality.  Hard core gamers have always suspected this, able to be more satisfied with a 30fps NVidia over a 60fps AMD driven display.  This is unlikely to change anytime in the near future; something to do with Titan architecture, developed and continually funded by the demands of NVidias supercomputing R&D half (.... as of last year when I researched this thoroughly ... you snooze you lose)

[gandy 18]

I was watching a replay of the ADX Orbx Dev interview on twitch and while they was flying around PNG it was about 99.9% stutter free, that included a 3 hour flight while the interview happened. That was a really smooth flight.

 

I was getting some stutters in P3Dv2 right up until my dimm_A1 and dimm_B1 slots failed on my motherboard so it was good for me to find the cause of my stuttering. The only downside is i now need a new motherboard and i have to buy compatible ram to go with it as well so when i get them i reckon i will be 99% stutter free

[Guest]

 

 

I was getting some stutters in P3Dv2 right up until my dimm_A1 and dimm_B1 slots failed on my motherboard so it was good for me to find the cause of my stuttering.

 

Bad RAM will not cause stutters, it will however cause lockups, BSOD, reboots.  There is no such thing as the RAM "sorta working" it's either working or it's not working, there is no gray area.

 

I've heard a few people claim that overclocking causes stutters, frankly I have NO supporting evidence to validate that claim (and from my knowledge of hardware it just doesn't make sense).  I do have evidence to support that overclocking has NO impact on stutters at all.  For my P3D display settings, overclocking from 3.9Ghz to 4.8Ghz produces 1 fps difference with no impact to stutters.  GPU overclocking will produce about 3 fps difference.  I've also tested different RAM timings and BCLK going from 1333Mhz (34GB/s, 933us copy time, latency 22.7ns) to 2133Mhz (51GB/s, 627us copy time, latency 18.1ns) and they produce no difference in fps.

 

Cheers, Rob.

[DocBird 79]

 

I've heard a few people claim that overclocking causes stutters, frankly I have NO supporting evidence to validate that claim (and from my knowledge of hardware it just doesn't make sense).  I do have evidence to support that overclocking has NO impact on stutters at all.  For my P3D display settings, overclocking from 3.9Ghz to 4.8Ghz produces 1 fps difference with no impact to stutters.  GPU overclocking will produce about 3 fps difference.  I've also tested different RAM timings and BCLK going from 1333Mhz (34GB/s, 933us copy time, latency 22.7ns) to 2133Mhz (51GB/s, 627us copy time, latency 18.1ns) and they produce no difference in fps.

 

Cheers, Rob.

 

Wow, Rob! You take this really serious! Thanks for that!

 

I do hope that you will get paid for all your hard work taking a scientific approach on this software (either in cash or rewarded otherwise ^_^)

[gandy 18]

Bad RAM will not cause stutters, it will however cause lockups, BSOD, reboots.  There is no such thing as the RAM "sorta working" it's either working or it's not working, there is no gray area.

 

I've heard a few people claim that overclocking causes stutters, frankly I have NO supporting evidence to validate that claim (and from my knowledge of hardware it just doesn't make sense).  I do have evidence to support that overclocking has NO impact on stutters at all.  For my P3D display settings, overclocking from 3.9Ghz to 4.8Ghz produces 1 fps difference with no impact to stutters.  GPU overclocking will produce about 3 fps difference.  I've also tested different RAM timings and BCLK going from 1333Mhz (34GB/s, 933us copy time, latency 22.7ns) to 2133Mhz (51GB/s, 627us copy time, latency 18.1ns) and they produce no difference in fps.

 

Cheers, Rob.

 

 

Hi Rob, i didnt have bad ram but faulty ram slots on the motherboard ( i have read its a know issue with the mobo i have ), but i have heard and read that using ram that is not compatible with your motherboard may cause stutters but its rare and having badly set up ram can cause stutters but this is with fsx not p3d.

 

I have not heard that overclocking can cause stutters but i guess if you have a really bad overclock its possible just as certain older sandforce controllers built into ssd drives are know to cause stutters out side of the sim but then you would see the stutters in other games and apps.

 

To me i do feel the main cause of stutters is a badly configured pc and sim ( with the addons we use and the settings we choose ) and as i see p3d as a memory sensitive app as i use to have a pretty stutter free sim with 2.2 ( i would say 96 to 98% stutter free and the other stutters could be addon related ) until my dimm slots started to fail then i started to see stutters. To me there is to many outside things that can effect the sim in a bad way which is why i dont like to blame any one thing for them but rather a group of possibility's that might cause them.

[Guest]

 

 

I do hope that you will get paid for all your hard work taking a scientific approach on this software (either in cash or rewarded otherwise )

 

I just wanna see folks happy with the product ... I'm fortunate to be an end user beta as most beta folks are 3rd party content providers ... but I think it's a good perspective for LM to have as 3rd party can often be focused on their specific issue and not the entire package as a whole.

 

 

 

To me i do feel the main cause of stutters is a badly configured pc and sim

 

That can be part of the problem but that's mostly around Graphics settings and dialing them back to reach a point where one is happy with the fluidity of the simulation, for some it has to be 60 fps for others 20 fps.  

 

But as far as video's you see on YouTube you need to be careful of how they were produced to obtain "stutter" free fluidity ... a common technique used is to turn up unrealistic graphics settings that actually run at 15 fps, record the flight (flight recorder or some other tool), playback the flight and set the simulation rate to "Slowest" (1/4 speed) and record that.  Then in one's video editing software (Adobe PP CC for example) you increase the speed by 400% and be sure to "retain pitch" to keep the audio working correctly and this will produce VERY fluid stutter free videos.

 

If I have time I will demonstrate this technique showing you "how it really flew" vs. "how it was recorded to fly" ... I personally don't have the time to do this with my videos nor do I want to fps "fake" what my experience was really like.  I'm not saying that was the case in the Orbx video you mentioned, but it's something folks should be aware off when watching a fully detailed scenery/airport with very fluid motion and everything maxed out.  I'm not here to judge the usage of the technique ... it's sometimes used as a sales tool, it's sometimes used as bragging rights ... just need to be aware all is not what it seems

 

Cheers, Rob.

[gandy 18]

Yeah the one i was watching was a recorded live stream on twitch so i dont think the flight was faked with the playback at a low speed to do that magic as the two of the dev's was giving him some guidance on his flight.

 

If anyone wants to look at the smoothness of the stream and p3d v2.2, its a 3 hour recording and software to broadcast to twitch server was on the same pc

 

http://www.twitch.tv/adxlive/b/544818794

 

It was a great stream btw.

[denali 74]

 I do have evidence to support that overclocking has NO impact on stutters at all.  For my P3D display settings, overclocking from 3.9Ghz to 4.8Ghz produces 1 fps difference with no impact to stutters.  GPU overclocking will produce about 3 fps difference.  I've also tested different RAM timings and BCLK going from 1333Mhz (34GB/s, 933us copy time, latency 22.7ns) to 2133Mhz (51GB/s, 627us copy time, latency 18.1ns) and they produce no difference in fps.

 

All of these things point to some other place in the system being the bottleneck.  If a 10% increase in CPU, or a 10% increase in GPU, or a 20% increase in RAM speeds produces NOTHING, then what makes the difference.  Where is the bottleneck?  This says that getting a faster CPU/GPU/RAM is not worth in any sense of the budget.  That we're being all ripped off paying for that extra 20% over last years models.

 

Or maybe the difference does come in when it is 12%, or 13%.  If you've ever watched a bottleneck of anything, sand, water, oil, a process in an assembly line, what you can see is that there is a threshold, once passed, that things begin to flow properly again. 

 

It used to be said, in all my 2 decades of FS lurking, that in FSX 4.5ghz is the sweet spot.  That it didn't matter what you were running, if you could get OC'd to 4.5 your troubles went away.  I never understood the structure of that, but In my experience it does feel logical.   That if you're holding up work at the CPU, meaning it doesn't flow enough to the GPU, the GPU doesn't get all the work it could push out, and the CPU being a bottleneck actually causes a bottleneck in the GPU.  Again, the GPU doesn't get all the parts it needs to process.  Please don't deride me; I'm only trying to establish a though/frame which to work from.  Because if what you say is not true, then we're being ripped off every iteration of hardware, things will never get better.  

 

I don't see things as a 1 for 1.  Let me digress.  In genetics today they know now that it's not a single switch that turns on or off a trait, but a network or matrix of switches, and each switch in the matrix can be part of another matrix trait.  When dealing with flow and systems it's often the same thing.  It's not one line, it's several lines front to back affecting each other.  It's a matrix of balances.  So hardware speeds are only a few nodes in the matrix.  To see this you put them ALL up on the board at the same time, and if what you are measuring isn't explaining something, doesn't account for net changes, you're not measuring enough places, not finding where the buck stops.

 

Could it be bandwidth issues, and how much you're trying to jam through those at a time?  Threads size, some unequal?  You can't just put everything in a giant hopper, e.g. 6GB worth of RAM, or VRAM, and expect it all to pump through when your pathways are not wide open, but an array of pathways only a few bits wide, they themselves having a tempo independent of RAM, VRAM, CPU and GPU.  This used to be called multiplexing.  And today every piece of hardware still has these arrays of tiny little straws.  

 

It's NVidia's responsibility to manage the tolls and the bridge, and a gamedev's responsibility to understand the what those comprise of.

[Guest]

 

 

All of these things point to some other place in the system being the bottleneck.  If a 10% increase in CPU, or a 10% increase in GPU, or a 20% increase in RAM speeds produces NOTHING, then what makes the difference.  Where is the bottleneck?  This says that getting a faster CPU/GPU/RAM is not worth in any sense of the budget.  That we're being all ripped off paying for that extra 20% over last years models.

 

Your answer is in my original context ...  but you'll need some reference ... going from 4.3 Ghz to 4.8 Ghz is about a 10% increase in CPU ... but when one is running at 15 fps that's about 1.5 fps gain ... if running at 30 fps, then 3 fps gain ... the problem is I need 2X performance increase at 15 fps to get back to 30 fps and "fluidity".

 

The bottleneck is pretty obvious ... if 10% CPU increase only nets 1 fps (ref 15 fps) and 10% GPU increase nets 3 fps (ref 15 fps) ... that tells you where the bottleneck is, in the GPU.  So where one will gain significant performance improvements will be on the GPU side and that means Multi-GPU and SLi.

 

If you look at some of my other videos on my Channel I show bus load, and frame buffer load which are very low.  Bottleneck is the GPU ... or more specifically how the nVidia drivers are working with P3D DX11 code.  More GPU is needed IMHO -- hence the desire for SLi -- but, as with Xplane 10.x getting multi-GPU support seems to be a much bigger challenge for LR and LM.  Which is a little ironic given that flight simulations will always demand more processing power than 3D type restricted (non-world) type games which for the most part do support Multi-GPUs.

 

The computer is a sum of it's parts, the software that operates on it defines how to manage those parts ... since P3D is software sits on top of an OS and DX11 is someone else's API, and nVidia drivers are always in a state of change, it can be a difficult to bring it all together.

 

Cheers, Rob.

[denali 74]

the software that operates on it defines how to manage those parts

  The hardware defines how to manage those parts.   The software is the means to manage those parts.  From Microcom, to Motorolla, it's the hardware that defines the software.   Anythings else is fantasy.   All it is, all it ever was, all it will ever be, is electrons (and photons) flying around along little paths, sometimes crammed into them, more often now than in the past, because the youngins don't remember what it was like to have to count and they're living the the clouds their ancestors created.    Please don't deride me; I think differently.  I'm not a fan of clouds and fog.  Trying to stay grounded, so to speak. 

 

 

 

improvements will be on the GPU side and that means Multi-GPU and SLi.

 I'm so looking forward to this.  Did I mention I am running at 3/4 of what you are?  5760?

 

Your efforts and passion are greatly appreciate Rob. Thank You