Jump to content
Sign in to follow this  
CFIJose

DX10 AA Performance & A GTX780

Recommended Posts

 

 


t sure did make a difference, and sorry, David 4.4-gig is not the same as 5.0-gig - newer architecture or otherwise.

 

This isn't true at all.

 

Its been shown and proven that architecture makes a huge difference. Even in single thread, Haswell at 4.4 ghz is equivalent to Sandy Bridge at 5+ ghz.

Share this post


Link to post

Sorry Glenn, we have to disagree. I've built and overclocked a number of Haswell pc's most with better boards and memory, and have found none that perform any better than my clunky old 5.2 gig (now) 2600K. 

 

 

 

It's been shown and proven

  

 

Show me. All I ever read are Tom's Hardware (or equiv) benchmarks - Halo, Bioshock, TombRaider, gamer benchmarks up the Ying-Yang, but not FSX benchmarks. I've also read of no-one, on any FS website where a Haswell will do any better than my old thing, with, say 25 on all traffic, (with 2 for airport traffic), full AG and IQ, 4096, 4.500, 8xSGSS , Bloom,  heavy REX or ASN/AS2012 or FSGRW, 3840 x 1024 x 32, the Terrain section all at max - these are my normal, daily VFR settings.

 

I may well see a more consistent framerate using a Haswell with a higher clock - say 4.6 or 4.7 - exceeding the continuous average framerate and smoothness which my old 2600k does - but all will sit around 30... 28 at some spots, bouncing around 27 to 32. Only on an unlimited frame rate will a higher number be seen, and that's always offset by a significant number of hesitations or stumbles.

 

Lastly, I also see the same number of performance-issue "Please help me"  requests, no matter what the proc. It's still all dependent upon how the whole system is set up - the hardware choices and the OS config. I'm sure Nick Neeham gets the same questions, again and again and again, whether it's C2D with an 8800 and 2 gig of ram, or the very latest all-singing, all dancing Haswell 5960X. I'm equally sure that Michael Greenblatt (now dec'd) of FS-GS fame thought the same, too.

The latest procs are borrowing their architecture from the server architectures nowadays, being forced to go wider because of the physical limitations of heat and wire size and using smp. There will come a day where games will be created to run on machines containing 8 Xeons, each containg 15 cores running at 2.8 gig.... but FSX will not run any better until it's transposed fully in to 64-bit, and using DX11 or 12. Then it will yield better performance than anything which we have today.

 

This disagreement is like comparing two small peanuts - pointless, when the real change actually comes with a 3" - 7cm Brazil nut.

 

All the Best, Glenn,

 

pj



i7 4790K@4.8GHz | 32GB RAM | EVGA RTX 3080Ti | Maximus Hero VII | 512GB 860 Pro | 512GB 850 Pro | 256GB 840 Pro | 2TB 860 QVO | 1TB 870 EVO | Seagate 3TB Cloud | EVGA 1000 GQ | Win10 Pro | EK Custom water cooling.

Share this post


Link to post

Ok, Im using the fixer with 4x AA and 4SGSS in nvidia inspector. The latter is necessary to remove jaggies, of course. In addition I'm using the tweaks BP=0 and AM=14. My GTX760 and 4,4 ghz i7 4770K can hold super stable FPS at any add-on airport with clear weather. However, it drops in really cloudy conditions. I have water effects on high because mid looks like a nintendo 64 game. 

 

Any tips for me? It seems like I need affinity mask to remove stutters, actually, but maybe I have the wrong value. 


38.jpg

Brynjar Mauseth 

Share this post


Link to post

I've been trying to get DX10 to look and perform better than DX9 for months. I get much more shimmering in DX10 than DX9 which can only be removed with higher SGSS settings. To get the same image quality in DX10 as DX9 I have to raise my SGSS so much higher that it drags the performance of DX10 below that of DX9. I've tried all the setups in the various guides presented on these forums. I have a 3770k at 4.5 Ghz and a GTX780. My monitors are 3 x 1280 x 1024 in Nividia Surround. I thought at one point that the low resolution monitors were the problem. I tried a single 1920 x 1080 and then a single 1600 x 1200 but the shimmering did not improve.

 

I have my DX9 setup per NickN's guide and it is very smooth and crisp on my machine. The only significant benefit I see from DX10 is the cockpit shadows. So I've resorted to using DX9 in bad weather and DX10 in sunny weather if I want cockpit shadows. There aren't any cockpit shadows to look at anyway when the weather is overcast. If someone has found a solution to eliminate the DX10 shimmering without the 4x or 8x SGSS I'd love to hear about it.

 

I'm about to buy P3D to see if it is any better but I'm not optimistic as I've read of the same shimmering issues in P3D.

 

Ted


3770k@4.5 ghz, Noctua C12P CPU air cooler, Asus Z77, 2 x 4gb DDR3 Corsair 2200 mhz cl 9, EVGA 1080ti, Sony 55" 900E TV 3840 x 2160, Windows 7-64, FSX, P3dv3, P3dv4

Share this post


Link to post

I consider myself picky about image quality, and I just don't have the problems other seem to have with DX10, AA and performance.  Most likely it could be that I only fly GA?  I use Steve's Fixer, 16xCSAA, 4xSSGS, 16xAF, LOD bias -1, and high quality texture filtering.  Using the latest cloud culling in the Fixer coupled with ASN's dynamic cloud radius has pretty much eliminated being hammer by heavy cloud cover, though having a gtx 780 drive a single 1920x1080 panel helps alot (like having a ski boat outboard on a rowboat).  I can see tiny amounts a jaggies in the VC if I look really carefully, and some of the tree models (but not all) shimmer a tiny bit.  The flip side is that everything looks sharp and well focused; I really hate smeary AA side effects and that I do not have (which I had with the DX9 settings I was using, the ENB mod was terrible in that regard).

 

I totally agree with Paul. We've hit a wall and there is no appreciable performance breakthrough since the 2600K when it comes to FSX.  But I do love my haswell and it definately smokes in FSX compared to the i7 860 which it replaced.


Rod O.

i7 10700k @5.0 HT on|Asus Maximus XII Hero|G.Skill 2x16GB DDR4 4000 cas 16|evga RTX 3080 Ti FTW3 Ultra|Noctua NH-D15S|Thermaltake GF1 850W PSU|WD Black SN750 M.2 1TB SSD (x2)|Plextor M9Pe .5TB NVMe PCIe x4 SSD (MSFS dedicated)IFractal Design Focus G Case

Win 10 Pro 64|HP Reverb G2 revised VR HMD|Asus 25" IPS 2K 60Hz monitor|Saitek X52 Pro & Peddles|TIR 5 (now retired)

Share this post


Link to post

Sorry Glenn, we have to disagree. I've built and overclocked a number of Haswell pc's most with better boards and memory, and have found none that perform any better than my clunky old 5.2 gig (now) 2600K. 

 

I didn't suggest that Haswell performs any better than a 2600k at 5.2 Ghz, just that you can obtain EQUAL performance at lower clock speeds with a Haswell.

 

Nick N himself agrees with this as can be seen here in his Haswell guide:

 

"You may decide you like what you see on the screen at the clock you are able to achieve without any modification of the CPU since Haswell is a FAR better chip that Sandy Bridge clock-per-clock. A Haswell at ANY clock speed, including default, is going to outrun a Sandy Bridge of equal clock!

@ 4.1 you can expect the same performance as a SB running 4500, no problem!

@ 4.5 you can expect the same performance as a SB running 5GHz, no problem!

@ 4.8 you can expect the same performance as a SB running 5.3+GHz, no problem!

It only goes up from there."
 
 
 
Architecture DOES make a difference. If you could get a Haswell upto 5.2 GHz it would blow the 2600k away at the same clock speed. 
 
Again, I said that equivalent performance can be obtained at lower clock speeds with Haswell, not better performance. Therefore, the more efficient architecture clearly makes a difference.
 
If this wasn't true, we would still be chasing 8 Ghz on Pentium 4's.
 
As for 'proc speed' in the FSX.cfg, you would have to have the exact formula for how that is calculated at FSX install to be of any meaningful use. If its purely done on Ghz 'speed' then it will not take into account the architectural differences at all. Unless you have this info, I believe that informing people that 'proc speed' is too low is a complete misnomer. It would be interesting to see how this calculates with various different architectures at comparable clock speeds. Maybe you have this info? If so I will sit corrected......

Share this post


Link to post

..and at 4.5 and above, Glenn, requires the cap to be removed so that it can be effectively cooled. That's not something for the average - even experienced guys will do.

 

"If you could get a Haswell upto 5.2 GHz it would blow the 2600k away at the same clock speed."  This still makes no change at all to the original statement that "I may well see a more consistent framerate using a Haswell with a higher clock - say 4.6 or 4.7 - exceeding the continuous average framerate and smoothness which my old 2600k does - but all will sit around 30... 28 at some spots, bouncing around 27 to 32." We are talking about FSX's usually subjective performance, as regular every day flight simmers - and in this context "blow the 2600K away" - isn't the case. It doesn't blow it away.

 

The Haswell is not the Be-All and End-All in the proc world or the FSX world, Glenn. This p*****g on other lesser or greater procs has been going on every year since the inception of the flight sim. It began when the 80-286 replaced the 8088. Then continued with the 386 DX, and then into the 486 - and it has never stopped. Throwing bucks, pounds, marks, etc., based on advertisement and "guru opinion" at a pc case - and suddenly it's a world-beater - is a fallacy. Everything depends upon context. It's what you do with the thing in a particular situation that makes it perform. I will bet that P3D running on a Haswell-based system is going to make P3D running on a fast SB system look pretty poor, because both P3D and the Haswell are "made for each other". I will drool. FSX is not made for the Haswell, the SB - nor even the the Core2Duo without SP2!!! It was also made for XP!

Nevertheless - I am merely stating that a well optimized SB 26 or 2700K will equal the Haswell when running FSX. Most everybody here is going to see their framerates bouncing around the 27 - 30/31 area, with a pretty good experience. The guys running DX10 are going to see something slightly better, with lesser issues (OOM's) and a few added features - whitecaps and smoothness. As Rod says "We've hit a wall and there is no appreciable performance breakthrough since the 2600K when it comes to FSX."

 

In an attempt to bring this obsolescent thread to a close (I hope, because I don't want to spend my retirement talking old shop) - in any IT department there will always be a healthy disagreement between techs, whether it's "the best" server, OS, HP, IBM, Dell, Oracle, database, front end, email, firewall, a/v, protocol, switch, router - you name it.  We work with what we have, and until P3D has the same functionality that FSX-DX10 displays, with it's hardware compatibility, with most every "regular" PC setup and its pretty vast array of great addons - then in my book I'm not spending beaucoup bucks on a proc that needs to be de-lidded in order to give the same FSX performance that my clunky old SB does!

 

All the Best,

 

pj



i7 4790K@4.8GHz | 32GB RAM | EVGA RTX 3080Ti | Maximus Hero VII | 512GB 860 Pro | 512GB 850 Pro | 256GB 840 Pro | 2TB 860 QVO | 1TB 870 EVO | Seagate 3TB Cloud | EVGA 1000 GQ | Win10 Pro | EK Custom water cooling.

Share this post


Link to post

Why you "high SGSSAA" guys do not give the nVidia DSR feature a chance? For the moment, I switched and see if this turns out to be "the AA option" for me. Short explanation: I also struggled with the fact that along with my 8x AA setting in the fixer only 8xSGSSAA was really looking good, 4xSGSSAA being still ok but with some jaggies. But, as for you guys, 8xSGSSAA is somehow an overkill when flying in cloudy conditions. Now, I enabled DSR for my Kepler-based GPU (GTX-780) and what to say, it seems that somehow the higher resolution alone with no AA at all seems to be less a problem for those clouds. Currently I run 3840x2160 without AA on my FullHD display and I am really happy with the IQ. Only the slightly decreased AF quality is bothering me a little bit, besides the fact that the menues are ridiculously small now. Anyway, I would recommend to give it a try. If you do not want to mod the 344.24 notebook driver to enable DSR for Kepler cards, simply wait for the next WHQL release, then it will be officially available for all nVidia cards, not only maxwell.


Greetings, Chris

Intel i5-13600K, 2x16GB 3200MHz CL14 RAM, MSI RTX 4080 Gaming X, Windows 11 Home, MSFS

Share this post


Link to post

I use Steve's Fixer, 16xCSAA, 4xSSGS, 16xAF, LOD bias -1, and high quality texture filtering. Using the latest cloud culling in the Fixer coupled with ASN's dynamic cloud radius has pretty much eliminated being hammer by heavy cloud cover, though having a gtx 780 drive a single 1920x1080 panel helps alot (like having a ski boat outboard on a rowboat). I can see tiny amounts a jaggies in the VC if I look really carefully, and some of the tree models (but not all) shimmer a tiny bit. The flip side is that everything looks sharp and well focused; I really hate smeary AA side effects and that I do not have (which I had with the DX9 settings I was using, the ENB mod was terrible in that regard).

I tried your settings last night Rod and had the same results as you described in DX10. The difference is my DX9 image quality looks just as good as DX10. On my system DX9 0x SGSS = DX10 4x SGSS and DX9 2x SGSS = DX10 8xSGSS with regards to image quality and shimmering. This extra SGSS in DX10 brings my FPS below that of DX9 and I need every FPS I can get for my AI Traffic addiction. I like flying the Citation X and it has a healthy appetite for FPS also. I didn't have time last night to experiment with the Fixer's cloud culling or ASN's dynamic cloud radius but will try it next. I appreciate the suggestions.

 

I use MyTrafficX v5.4b and even at 20% it really hits the FPS hard at large airports in major cities like KLAX. Perhaps I have a problem with my AI traffic installation that I need to investigate. The worst shimmering is on my AI Traffic that is parked at airports. In DX10 with 2x SGSS or none the parked AI Traffic shimmers like christmas tree lights. It is annoying on approaches thus I need to run DX10 at 4xSGSS minimum to reduce it.

 

I might try the new Nvidia DSR but I think I will wait until it is incorporated into an official non-beta driver.

 

Ted


3770k@4.5 ghz, Noctua C12P CPU air cooler, Asus Z77, 2 x 4gb DDR3 Corsair 2200 mhz cl 9, EVGA 1080ti, Sony 55" 900E TV 3840 x 2160, Windows 7-64, FSX, P3dv3, P3dv4

Share this post


Link to post

Why you "high SGSSAA" guys do not give the nVidia DSR feature a chance? For the moment, I switched and see if this turns out to be "the AA option" for me. Short explanation: I also struggled with the fact that along with my 8x AA setting in the fixer only 8xSGSSAA was really looking good, 4xSGSSAA being still ok but with some jaggies. But, as for you guys, 8xSGSSAA is somehow an overkill when flying in cloudy conditions. Now, I enabled DSR for my Kepler-based GPU (GTX-780) and what to say, it seems that somehow the higher resolution alone with no AA at all seems to be less a problem for those clouds. Currently I run 3840x2160 without AA on my FullHD display and I am really happy with the IQ. Only the slightly decreased AF quality is bothering me a little bit, besides the fact that the menues are ridiculously small now. Anyway, I would recommend to give it a try. If you do not want to mod the 344.24 notebook driver to enable DSR for Kepler cards, simply wait for the next WHQL release, then it will be officially available for all nVidia cards, not only maxwell.

 

Can you explain what this DSR thing is all about, very short? Is it just higher resolution? How is that going to make a change?


38.jpg

Brynjar Mauseth 

Share this post


Link to post

I didn't have time last night to experiment with the Fixer's cloud culling or ASN's dynamic cloud radius but will try it next. I appreciate the suggestions.

These are the specific settings I use.

ASN:  Cloud Options->Minimum cloud draw distance 75, Maximum cloud draw distance 145

Fixer: Clouds-> enabled these - Disabled MSAA, Cloud Clip, Discard Fogged, Missing Texture Fix

 

 

Can you explain what this DSR thing is all about, very short?

With the release of Maxwell, nvidia has rolled out (yet another) full screen AA method which promises better image quality; Kepler will be supported in later driver releases.  Read about it here: http://www.geforce.com/whats-new/articles/dynamic-super-resolution-instantly-improves-your-games-with-4k-quality-graphics


Rod O.

i7 10700k @5.0 HT on|Asus Maximus XII Hero|G.Skill 2x16GB DDR4 4000 cas 16|evga RTX 3080 Ti FTW3 Ultra|Noctua NH-D15S|Thermaltake GF1 850W PSU|WD Black SN750 M.2 1TB SSD (x2)|Plextor M9Pe .5TB NVMe PCIe x4 SSD (MSFS dedicated)IFractal Design Focus G Case

Win 10 Pro 64|HP Reverb G2 revised VR HMD|Asus 25" IPS 2K 60Hz monitor|Saitek X52 Pro & Peddles|TIR 5 (now retired)

Share this post


Link to post

For those interested - there's a good thread on DSR over at the Orbx Tips & Tricks forum here. To use it with the Kepler series, Mark Abdey has posted a link to the driver, and also to the .ini file needed to fool the install.

It does work (with a bit of patience), and will improve AA for a number of folk, I suspect. I've gone back and fore, reinstalling the driver three or four times, but I'm not sure of a noticeable improvement when set at the native size of my monitors, one being 3 x 19" at 3840 x 1024, and often switching that one out to use a 27" ASUS at 2560 x 1440 - but everyone is looking to get something different out of their sim. I like "real" water, waves, colour, whitecaps, etc., and as long as there's some fluffy white stuff hanging around I'm happy. Others are reversed in this!

 

pj



i7 4790K@4.8GHz | 32GB RAM | EVGA RTX 3080Ti | Maximus Hero VII | 512GB 860 Pro | 512GB 850 Pro | 256GB 840 Pro | 2TB 860 QVO | 1TB 870 EVO | Seagate 3TB Cloud | EVGA 1000 GQ | Win10 Pro | EK Custom water cooling.

Share this post


Link to post

Starting with a large non-AA image and reducing to a smaller size has been going on for years with these GPUs, pixels averaged down with special sampling patterns to produce the smaller AA image, make for good quality results. The difference with the new fangled DSR is for the GPU driver to spoof the higher resolution to the game, but present it in real time at the desired (lower) AA resolution. Whereas "older" AA methods start with the original resolution and then do the averaging sample pattern thing.

 

FSX and P3D build up a pile of AA images in the scene and then render it. The difference with DX9 to DX10/11 is that images backgrounds are treated differently. So with DX9 we use the NI AA settings: "Antialiasing - Mode", and "Antialiasing - Setting", but with DX10 and DX11 we use "Antialiasing - Transparency Multisampling" and, "Antialiasing - Transparency Supersampling" because of this difference with the way the backgrounds are treated by the GPU.

 

Now, with DX9 all was good and there are lots of AA settings to play with, but with DX10 and DX11 we only get a few. What's more in DX10 and DX11, is due to the way some models are built, a problem can arise whereby "Antialiasing - Transparency Multisampling" and, "Antialiasing - Transparency Supersampling = simple" cannot work on those images for pre-rendering, and so we use "Antialiasing - Transparency Supersampling = Sparse Grid Supersampling" (SGSSAA).

 

Sparse Grid Supersampling (available with FSX DX10 and P3D DX11) does what it says on the tin, takes a sparse grid pattern sample and ultimately is slightly less detailed than simple Supersampling. To keep shimmering down also use Aniso-mode and high quality texture filtering. Aniso-mode improves the look of water in the distance too.

 

On a 3960x + GTX680 the full blown DX10 and DX11 AA settings don't do too much damage at a resolution of 1920x1200x32, maybe 2 or 3 fps.

 

 

Best.


Steve Waite: Engineer at codelegend.com

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
Sign in to follow this  
  • Tom Allensworth,
    Founder of AVSIM Online


  • Flight Simulation's Premier Resource!

    AVSIM is a free service to the flight simulation community. AVSIM is staffed completely by volunteers and all funds donated to AVSIM go directly back to supporting the community. Your donation here helps to pay our bandwidth costs, emergency funding, and other general costs that crop up from time to time. Thank you for your support!

    Click here for more information and to see all donations year to date.
×
×
  • Create New...