Is there a FlightSim Con presentation from DTG?

[Guest]

3 hours ago, Jiri Kocman said:

Not sure about LM guy

Rob McCarthy is Prepar3D Core Lead developer (as pointed out far left of table away from the spotlight).  I'm sad I missed this, but I will hopefully have more free time next year and get in some hired development help so my work load is reduced (LM have expressed a desire to return next year).

If permitted I would have asked questions with a much different focus, more on the technical challenges around current hardware limits and rendering pipeline DX11, DX12, OpenGL, as I really didn't hear (could be audio issue and I just didn't hear) any technical questions being asked? 

It would seem unlikely 3 competitors on the same panel would divulge much information about "future plans" and/or make any commitments.

No offense intended towards the person that made this video, appreciate that it exists and was shared, hopefully another video might surface?  AirDailyX?

Cheers, Rob.

[Paraffin 2,197]

47 minutes ago, Rob Ainscough said:

It would seem unlikely 3 competitors on the same panel would divulge much information about "future plans" and/or make any commitments.

But the Laminar guys did exactly that, didn't they? Lots of new stuff in the pipeline was mentioned. Like VR, the new G1000, and (maybe) some updates to ATC.

It's just a different company culture. Austin and Ben are the lead coders, and Austin is the CEO/principal owner of the company. There is no middle man or higher-up saying "maybe we shouldn't talk about this yet." I'm sure there is still a filter so they're not promising too much, but basically everything they talked about at this conference has already been mentioned in their public developer blog. They're pretty clear about the road map, and I think users appreciate that.

DTG would benefit from making at least a small move in this direction, even if a road map for upcoming features is covered with disclaimers about timing for new features. Heck, even the game developers over at E3 are showing trailers for games that won't come out until 2018. I don't know why the flight sim developers need to keep things so closely guarded. 

[Guest]

1 hour ago, Paraffin said:

But the Laminar guys did exactly that, didn't they?

I wasn't there so I can't confirm, but in the video I didn't hear any "dates" associated to anything from anyone on that panel ... maybe they were provided else where?  Feature delivery is what matters, until that happens nothing is "real".  Why would DTG benefit?  DTG already got bitten by providing information they probably should have kept to themselves (such as release estimates other promises made).

I'll venture to guess LR mentioned VR, G1000, and ATC because those features already exist in competitor platforms.

E3 Game developers are working an entirely different audience, did you see any serious "Flight Simulators" at E3?  Did you see any games at E3 that also offer an "extensive" SDK/PDK (the primary reason Flight Simulation is sustainable) and have huge regular support of 3rd party developers that sell add-ons?

If potential customers don't know, they can't start any wild speculation (for the most part, still happens but limited).  As we know, some of the wild speculation then turns into "fact" and then users are driven away or towards based on false information, is this a good thing?  I'd venture, no it's not.  Can't tell you how many times I've had some poor individual sent down the wrong path (either accidentally or intentionally) and comes to my YouTube channel or PM's me and I have to help clear up the "alternative facts".

But like I said, my panel questions would have a different focus, I'm more interested on poly limits, bandwidth limits, forward/differed rendering in the graphics pipeline, lighting optimization, shadow optimization, what limits are they hitting on the hardware side (all the platforms are pushing the boundaries of hardware), other optimization techniques, and what might come down the hardware pike in the future that could address some of the current performance restrictions like DX12 and ability to manage GPUs (multiple) via DX12 API, etc. etc. 

Cheers, Rob. 

[torium 207]

4 hours ago, Rob Ainscough said:

No offense intended towards the person that made this video, appreciate that it exists and was shared, hopefully another video might surface?  AirDailyX?

Alternative ....?

https://www.twitch.tv/videos/150886328

https://www.twitch.tv/videos/150893642

[Longranger 148]

5 hours ago, Rob Ainscough said:

 

If permitted I would have asked questions with a much different focus, more on the technical challenges around current hardware limits and rendering pipeline DX11, DX12, OpenGL, as I really didn't hear (could be audio issue and I just didn't hear) any technical questions being asked?

Not of this format, but I am not sure if they would really talk about it in such a discussion. There would be too few people who would really understand the answer. It is obvious X-Plane will leave OpenGL. They simply don´t have the amount of control as with Vulkan or Metal. They described this explicitly in their own presentation. I haven't found recordings for such presentations by FSW or P3D.

For P3D and FSW the main problem wit DX 12 is simply Windows 10. Many computers don'switch to Windows 10.

And if we look at the Nightlighting: X-Plane now uses Deferred Rendering, this means the more light sources you have where the illuminated areas interconnect, the bigger is your speed advantage. P3D v4 256 lightsources where the reflections are rendered.

FSW currently more or less broken.

[Guest]

53 minutes ago, torium said:

Alternative ....?

Thanks, first link is broken, second link is working and is a little better, but I see a rather large professional grade camera on tripod in the left corner so I assume a higher quality video with mic audio feed will be available eventually?

20 minutes ago, Longranger said:

Deferred Rendering, this means the more light sources you have where the bright areas interconnect, the bigger is your speed advantage. P3D v4 256 lightsources where the reflections are rendered.

There is no limit to the reflective aware light sources in P3D V4, there is a Prepar3D.cfg entry one can add to increase or decrease the limit to any value.

XPlane is moving away from OpenGL - is that "official"?  Vulkan is probably easiest transition from OpenGL and still cross platform.  I'm still not sure cross platform support is a good thing in terms of extracting the most possible performance but is good from a marketing stand point.  

I believe the GPU and multiple GPUs are going to become more important for flight simulators moving forward.  With P3D V4 it is easy to trigger 100% GPU utilization as performance limits have now shifted from CPU to GPU (this is a good thing) ... although P3D V4 does support SLI and multiple GPUs they are very reliant on nVidia support under DX11.  

Agree, DX12 is bound to Windows 10 which currently IS a problem ... this means the FS developer(s) need to either implement two render paths/pipelines for DX11 and DX12 or wait years for Windows 10 to reach over 50% market share (currently around 27%).  There is also the question of making DX12 payoff in terms of performance ... there are many ways to implement under DX12, but to gain performance form DX12 will most likely require a complete reworking of the render pipeline.  Getting accelerate rendering under DX12 (prior to Pascal GPUs) was limited and the DX12 label on the those GPUs was more a "supported" not necessarily "accelerated".  From that perspective I can see why LR might find Vulkan a more attractive option not to mention the similarities it has with their existing OpenGL code base.

As far as forward/deferred ... deferred is good for dynamic lighting, but not good for AA and transparency/Alpha ... DX12 or Vulkan API would permit more control to optimize the pipeline ... but either way there are no miracles in Performance that a single GPU can do, so lets hope multiple GPUs start to take forefront in the future as I think their time has finally come to Flight Simulation with the appropriate API.

I guess I better make it for FSCon2018 so I can at least try to ask these questions along with some SDK/PDK questions.

Cheers, Rob.

 

[Longranger 148]

2 hours ago, Rob Ainscough said:

Thanks, first link is broken, second link is working and is a little better, but I see a rather large professional grade camera on tripod in the left corner so I assume a higher quality video with mic audio feed will be available eventually?

There is no limit to the reflective aware light sources in P3D V4, there is a Prepar3D.cfg entry one can add to increase or decrease the limit to any value.

XPlane is moving away from OpenGL - is that "official"?  Vulkan is probably easiest transition from OpenGL and still cross platform.  I'm still not sure cross platform support is a good thing in terms of extracting the most possible performance but is good from a marketing stand point.

You wont get it any more official.

They guarantee the OpenGL support for the rest of X-Plane 11! But the future will be Vulkan and Metal!

2 hours ago, Rob Ainscough said:

 

  

I believe the GPU and multiple GPUs are going to become more important for flight simulators moving forward.  With P3D V4 it is easy to trigger 100% GPU utilization as performance limits have now shifted from CPU to GPU (this is a good thing) ... although P3D V4 does support SLI and multiple GPUs they are very reliant on nVidia support under DX11.  

And I wouldn't bet on much more SLI support by them. And for X-Plane Ben pretty much _guarantees_ that there will be NO SLI support! Support for multiple monitors with multiple GPUs yes, but no SLI or CrossFire.

 

 

2 hours ago, Rob Ainscough said:

 

 

[rtodepart 761]

thanks for posting this videos guys    ....the question is ,,,, how much coffee did Austin had ???? before he went on the stage ...

[Paraffin 2,197]

1 hour ago, Longranger said:

And I wouldn't bet on much more SLI support by them. And for X-Plane Ben pretty much _guarantees_ that there will be NO SLI support! Support for multiple monitors with multiple GPUs yes, but no SLI or CrossFire.

Right, there is an old post from one of their dev blogs about how the way the sim works, it would actually slow down in SLI mode.

That's the way the whole industry seems to be going, between game developers and the GPU manufacturers. Recent games on Windows are optimized first for the latest fast single GPU's, with SLI and Crossfire an afterthought if it's ever supported.

This seems to be an advantage for VR too, in not having to feed frames across a buss between separate GPU cards.

I think this should take the pressure off DTG for adding support for SLI or Crossfire. We seem to be moving past that now.
 

1 hour ago, rtodepart said:

thanks for posting this videos guys    ....the question is ,,,, how much coffee did Austin had ???? before he went on the stage ...

That's normal Austin, from every video I've ever seen of the guy. 

[Guest]

I'll disagree with you there, multiple GPUs in "performance mode" (SLI/Crossfire able to leverage more than one GPU) is going to be key to improved performance just as have multiple CPU cores helped.  GPUs are more suited to shared processing because the tasks are structured and finite.  DX11 it's difficult to manage, but DX12 provides much more access to the GPUs but it's not a trivial programming task to leverage the true benefits of DX12.

Vulkan has many of the same capabilities in multi-GPU support also and could leverage that support for much better performance.

We just don't have enough GPU processing muscle for XP11 or P3DV4 ... the number of pixels to process at 4K or higher resolution is pretty staggering and to have to get it all done in 16-33ms.  I have no problem making XP11 and P3DV4 run at 2 FPS because my GPU is at 100% load while my CPU's are 60-70%.  Fast forward to 8K monitors end of this year and next year (for mass market) and I feel sorry for any single GPU trying to "perform" given the current visual capabilities of FS platforms.

I hope LR re-think their decision to support multiple GPUs to share processing loads, the move to Vulkan would be the perfect time to do it.

Cheers, Rob.

 

[Longranger 148]

15 hours ago, Paraffin said:

That's the way the whole industry seems to be going, between game developers and the GPU manufacturers. Recent games on Windows are optimized first for the latest fast single GPU's, with SLI and Crossfire an afterthought if it's ever supported.

Well, the reason is rather simple. SLI and CrossFire are a solution if the computing power of the GPU is the big limit. As every X-Plane User here knows: The most important factor of the graphics card is its amount of VRAM! The simple truth is: Once the memory limits of 32 bits were dropped, everybody used a bit more memory and suddenly the internal transfer speeds play a major role. Suddenly the CPU and GPU simply wait since they have to wait till they have all the informations that they need.

FSX and P3D didn't had these limits since everything together had to be smaller than 4 GB. For X-Plane 11 I would advice for a VRAM not smaller than 4 GB if possible bigger. If you use some sceneries it was no longer unusual if the computer used 16 GB of RAM under X-Plane 10. X-Plane 11 is better optimized for memory throughput, so its internal reserves are significantly higher while its pure speed is slightly lower. If functions can now handle 10 times the objects that they previously handled than they don't become two times as fast if they only get half as many objects..

Two cards in a SLI configuration have in fact an intertnal VRAM that is a bit smaller, than the size of the smallest card in its configuration.

As everyone should know the VRAM size of todays cards has exploded and in many cases the number of informations that have to synchronized between them. The overlays that you need for synchronisation can easily take more time, than the performance boost that you will get by using two GPUs.The real nail in the coffin are simply their much faster internal transfer busses. The external interfaces are no more than roadblocks. We shouldn't forget: the GTX 1080 is slowed down since its internal memory isn't fast enough!

As far as I see it NVidia and AMD would like nothing more, than to drop SLI and CrossFIre.

 

 

 

[ca_metal 2,015]

Let´s just wait for airdailyx´s videos, D´Andre is editing those I guess:

 

http://www.airdailyx.net/fsnewsbreaker/2017/6/13/now-processing

[Guest]

57 minutes ago, Longranger said:

Two cards in a SLI configuration have in fact an intertnal VRAM that is a bit smaller, than the size of the smallest card in its configuration.

Sorry but this isn't true, it really depends on what graphics API is used and how one uses that API.  DX12 offers several implementations that even allows the use of entirely different GPUs (mix and match AMD, nVidia, Intel onBoard, etc.).  DX12 offers IMA, EMA linked and unlinked modes ... SFR (split frame rendering) is the goal to maximize individual GPUs as each GPU is given a subset of tiles to render using their own VRAM (not duplicated).  AFR requires duplication of data (wasted VRAM) to work as each GPU renders an entire frame.

The objective is to use DX12 EMA and SFR (GPUs process tile subsets in parallel using each GPU's own VRAM making full use of all the VRAM available to each GPU) as that would produce the most dramatic increase in performance from GPU scaling and making more efficient use of VRAM across all GPUs.  I don't know if Vulkan has a similar mode of operation, perhaps someone with Vulkan experience could chime in?

Multiple GPUs and a graphics API to leverage along with code implementing EMA and SFR is really the best option moving forward ... light is and always has been the most process intensive calculation as you move forward in visuals (think Crepuscular rays and more).  Light and the shadows it creates is such a massive part of simulating a realistic environment and it's the most intensive to calculator as it affects everything being rendered.

Of course there is no technical reason why DX12 can't be implemented in Windows 7, just Microsoft's way of forcing users to Windows 10.  DX12 can be incredibly beneficial when used correctly, but getting users to move to Windows 10, that's another story and the main reason why we aren't seeing many more developers using DX12 and more specifically EMA and SFR in DX12.  BUT, with that said, I'm willing to bet if LM produced a version of P3D Vx that was DX12 implemented with EMA and SFR and people were reporting 50-150%+ performance improvements with 2 or 3 or 4 GPUs then I suspect many would move to Windows 10 (not all, but it would be a big incentive).

Cheers, Rob.

[Paraffin 2,197]

2 hours ago, Rob Ainscough said:

We just don't have enough GPU processing muscle for XP11 or P3DV4 ... the number of pixels to process at 4K or higher resolution is pretty staggering and to have to get it all done in 16-33ms.  I have no problem making XP11 and P3DV4 run at 2 FPS because my GPU is at 100% load while my CPU's are 60-70%.  Fast forward to 8K monitors end of this year and next year (for mass market) and I feel sorry for any single GPU trying to "perform" given the current visual capabilities of FS platforms.

Well, my argument there, is that it's nuts to run one of today's modern flight sims on a 4K monitor, while expecting the kind of world detail we've traditionally had at lower resolutions.

I think Ben Supnick has said something to that effect, that their sim is not intended to support the kind of 4k screens at a steady 60 fps with the sliders maxed out, like you can do with most AAA games now.

The problem as I see it, isn't that today's GPUs aren't strong enough without going to a multi-GPU solution. It's that the monitor industry has gotten way out ahead of the GPUs that are driving the monitors. And then there's VR, but VR can handle dropping the eye candy sliders because the 3D immersion is so powerful as compensation, and you can't see high-res textures at current headset resolution anyway. 

There's a reason I'm still running my flight sims on a 1920x1200 monitor. It's because I can get nice high frame rates on "just" a GTX970 card. I plan to upgrade to a 1080 (or 1080ti) later this summer, but it's mainly for the much higher VRAM that will let me bump up a few settings I have disabled now, like reflections and world object shadows. I think I'll stay at this lower resolution for a while. Being able to fly consistently at a 40 fps frame rate, with complex planes and lots of clouds, is more important to me than huge screens and high resolution.YMMV, and I know 4K and 8K monitors are popular. 

I don't know if any of this is useful to the DTG devs, but they should know that a few of us haven't jumped on the 4K bandwagon yet. 

[jabloomf1230 2,445]

4K looks perfectly fine for me with very high IQ settings, both in XP11 and P3d V4 and I can get 30-40 fps loaded up with complex aircraft, scenery, weather and add-ons, I'm not sure why 60 fps should be some type of visual nirvana.

It's just a matter of time before the GPU hardware makes these frame rate discussions obsolete, just like 64 bit sims made the word VAS archaic.