Z buffer accuracy

[plethaus]

Forget it, it's no big deal. The important thing is you're happy with what you're seeing :)

[buzz456]

I think this might be like serious audiophiles. They hear things us peasants can't hear.

[Bananarama]

The problem has to do with coplanar faces on the model - how much space the author allowed in between and increasing that space to take into account LODs.

[buzz456]

The problem has to do with coplanar faces on the model - how much space the author allowed in between and increasing that space to take into account LODs.

Too deep for me.

[plethaus]

I think this might be like serious audiophiles. They hear things us peasants can't hear.

Yeah, it's basically the same thing.

[g_nash]

The problem has to do with coplanar faces on the model - how much space the author allowed in between and increasing that space to take into account LODs.

True, though as mrennie points out the new camera zoom feature does play a now part in how coplanar faces are rendered.

[Bananarama]

True, though as mrennie points out the new camera zoom feature does play a now part in how coplanar faces are rendered.

It should have little effect if the model is set up correctly. The zoom feature only changes the field of view - you're not actually getting closer to objects. Thus z-buffer problems will be more noticeable at high zoom yet always be present regardless.

[Ericmopar]

I'd like to really see what's going on, with these "maxed out" systems.
I have a GTX 780 with a overclocked i5 3570k at 4gig and can't run maxed out with my SSAA at 2x2.
I do however try to maintain at least 60fps so I maintain V-Sync. I have bloom and other things turned off, and most everything at High or Medium.
I can say that I've had a GTX 560, 660 and 780, with the only appreciable difference being that I can run 2x2 SSAA with the GTX 780. That helps alot with flickers and moir.

[buzz456]

I have far from a "maxed" out system. I suspect that mother boards and various other components factor into this. I have never understood the thing about fps since you can go way slower than 60 without seeing any flickering. It's stuttering that used to drive me nuts until I got the video card that I have now. Additionally there is a major difference between a one gig memory on the card and a two gig one. BTW my card is a GTX650 ti boost for whatever that adds to the discussion.

[artimrj]

I have a 6 core Phenom at 2.7 Ghrz. Thats 6 real cores. I also have 8 gig of ram, that was superfast 2 years ago. Nothing is overclocked, I have yet to try that as I never saw a need. So really nothing special about my system except the GTX760 which is only a month or so old in my system.

[Ericmopar]

All I know, is that I've run those graphics cards in two to three different systems, with entirely different setups for mother boards, CPUs etc.
Plenty of power at the PSU.
No background processes getting in the way. I never use more than 1 core, plus about 40% on the second, core.

I think something fishy is going on. According to gaming sites, I'm getting the performance that I should be and others get.

I get bad screen tearing below 60FPS (my monitor's refresh rate is 60hz). I don't know if that's my cheap Dell monitor or not.
I also hate aliasing. Therefor I set things up to run over 60 FPS, so it will maintain V-Sync, even on a cluttered route like Donner, and I also run 2X2 SSAA.

When I turned on 3X3 SSAA the Roseville freeroam was running at 30 -35 fps and suffering from control lag, and huge low fps screen tearing.

I have to admit, when I'm on an older route like Cajon Pass, I can run higher settings plus the 2X2 SSAA without going over 75% GPU usage, and it usually hovers around 40 - 50% when out of yards and cities.

[peterhayes]

Phil

system ram cant be that significant... can it?

System Ram is only significant if it does not match the processor and if you have less than 8GB running TS2014 on a 64-bit OS and your RAM at 1600 matches a Sandy Bridge 2500K, if correctly set in the BIOS - XMP setting - without depending on the mobo it may default to 1300 (665 x 2).

With your rig ie a 40" LED TV even a 4GB video card would be struggling to render the pixels needed with full SSAA on in TS2014.

The sad fact is that we cannot improve the z-buffer, shadow thingies etc issue using NVidia CP or Nvidia Inspector. I can eliminate 99% of jaggies on OHW and Moire in the switches/points/turnouts with TSX off but am only partially successful with TSX ON.

What nobody can explain is why we get most of the jaggies and moire in the center of the monitor - they do not seem to appear in the periphery of the monitor ie top, bottom and/or sides.

When I ran 3 x 3 SSAA I saw a huge increase in frame dwell times even when I ran a 2700K on two cores @ 5.2GHz (fps was poor and with stutters) so I find with TSX on 2 x 2 is good and gives a smooth performance even at 30 fps. Remember FRAPS and Bandicam report FPS before the frame is displayed and they do not record frames that are produced and not displayed ie lost. So 60 fps could mean 90% of frames being 60 fps but 10% could be at 10 - 20fps (or over 60fps) and that could show as a stutter. I avoid vsync if I can as there can also be a vsync stutter (Largely reduced using adaptive vsync), and its only needed if the tearing is bad usually if you have a low entry monitor/video card combo. I have one true 120Hz monitor and that removes most of the tearing issues without setting vsync.
Lets hope UE 4 is better than the present UE2/3!!
pH

[peterhayes]

Phil

How accurate is afterburner? i dont use either fraps or bandicam, but have noticed that afterburner and the in game fps meter are only 1 off at times.

They are all accurate to a point but they sample the fps early in the piece as I say before they are displayed, and they do not show lost frames. Newer methods use frame dwell times where you are looking to ensure that 99% of any frame dwell time is less than 20 milliseconds and with no FDT >50 milliseconds where you are likely to see stutters.
FDT measuring is quite complicated and may not give any better indication than say fps. NVidia have also introduced a new tool (http://www.geforce.com/hardware/technology/fcat) to help with a newer method of measuring a gpu efficacy but its complicated and expensive - this is the gist of fps and its issues http://techreport.com/review/21516/insi ... nchmarking.

Watching fps as a measure of what you see on the screen is poor - it only indicates really poor performance (eg <14 fps in TS2014) and you can still get stutters @60fps. Hence drop the fps monitoring and if the display is smooth (the odd random stutter happens in any game) and only use it if you have slide show or a high recorded fps with stuttering.
Regards
PeterH

[plethaus]

I'd like to really see what's going on, with these "maxed out" systems.
I have a GTX 780 with a overclocked i5 3570k at 4gig and can't run maxed out with my SSAA at 2x2.
I do however try to maintain at least 60fps so I maintain V-Sync. I have bloom and other things turned off, and most everything at High or Medium.
I can say that I've had a GTX 560, 660 and 780, with the only appreciable difference being that I can run 2x2 SSAA with the GTX 780. That helps alot with flickers and moir.

I've been experimenting with the options for getting TS2014 to stay at 60fps/vsync on for a fluid experience. If I run all settings at High and DoF/flares/bloom/foliage on, and ONLY FXAA I can get 60fps on most routes, most of the time. Stevens Pass is one where you really have to turn down settings (WAY down) to get close to 60fps. 2x2 SSAA starts to cause noticeable drops, but I really can't live with just FXAA. Way too many jaggies.

My 780 Ti can't run 3x3 even at 30fps without dipping down. Certain things like turning on headlights or cab lights causes a huge drop in FPS with 3x3 (but it definitely looks really good.)

I've kind of settled on running 1/2 Refresh vsync and capping TS2014 at 30fps. It gives a surprisingly smooth experience, and I can run 2x2 SSAA and max all other settings and it never drops down once (except in Stevens Pass in the dense scenery areas)

BTW I also have a 3570k running at 4.6ghz and 8gb 2400 ram.

[plethaus]

It's not that bad, it's just noticeably more demanding than most other routes.