NVIDIA Game Works, good or bad?

[CarstenS]

People seem to be sold on repeating the story about wireframe-visible water over and over again, I was just making a suggestion to re-check if this is the same with Crysis 3 (the game specifically, not Cryengine 3), since if this would be very hurtful towards Radeon cards, IMHLO, it would not be present there.

Since Crysis 3 still is around, people might want to check their facts against that. Nothing more did I suggest.

I know that in the original Far Cry, there was water overlay as well, but then - it was set on a comparatively small island, so it'd make sense that you don't remove water completely.

 

[Razor1]

ah ok, err I could probably do it later, have a copy of crysis 2 and 3 laying around here in my stack of old dvd's lol, and today's cry engine 3 (today's cry engine 3 does have a setting to turn off water entirely for a level btw but didn't see anything for the FOV saw that last night) tonight maybe.

 

[RobertR1]

Hardocp did a good article on witcher 3 in which they discuss Hairworks a fair amount. The performance between nv and ati is similar with hairworks. Infact, the FuryX has less of a performance hit than 980Ti with AA + hairworks. It's a good read and I highly recommend checking it out. Still, my OC'd 980ti is the clear winner here and the Furyx has no real OC overhead and are priced similar. http://www.hardocp.com/article/2015...hunt_gameplay_performance_review#.VdzKPvZVjcg

HBAO+ is still limited to nvidia cards but that fine.

Quite a few years ago when TWIMTBP was being called out for making games run better on nvidia hardware through develop influence, I stated that AMD should do a better job of doing similar with the GITG program. They just never learned how to market to consumers AND developers and got left behind in both areas.

Gameworks isn't the reason why ATI/AMD is having a hard time. Having fallen behind for years now without a hit product on their hands is the core issue here. With their R&D shrinking, they'll need to rely more on luck than anything going forward.

 

[Razor1]

finally got around to playing Crysis 3 again, i don't see tessellation working at all in this game, the e_tessellation command isn't doing anything in the game, did anyone come across this?

reading up on Crysis 3's version of the engine, looks like tessellation was indeed disabled at the launch version of the engine, there seems to be 3 cVar for tessellation

r_WaterTessellationHW
e_Tessellation
e_TessellationMaxDistance
r_TessellationTriangleSize

Will test this out tonight or tomorrow to see if these will activate if I put them into the system.cfg file.

 

[Ethatron]

Ocean water tesselates on high spec and other geometry tesselates when you use materials with displacement mapping enabled (and high spec).

 

[Razor1]

tried that, do I have to update Crysis 3? cause I can't seem to get it turned on with the new install.

 

[madyasiwi]

I don't have the game, but does Crysis 3 has the kind of "open sea" (i.e. extended to horizon) like it's predecessors have? All Crysis 3 scene shots I can find on the web only shows "contained" body of water in a more constrained level design. In which case, may not required to use similar water rendering techniques (camera aligned grid mesh) and thus could make comparison pointless.

 

[lanek]

Hardocp did a good article on witcher 3 in which they discuss Hairworks a fair amount. The performance between nv and ati is similar with hairworks. Infact, the FuryX has less of a performance hit than 980Ti with AA + hairworks. It's a good read and I highly recommend checking it out. Still, my OC'd 980ti is the clear winner here and the Furyx has no real OC overhead and are priced similar. http://www.hardocp.com/article/2015...hunt_gameplay_performance_review#.VdzKPvZVjcg

.

Just for make a little precision, this is the 8 or 9th patch who have been released since the release of the game, there is a reason why every review out before have just disable hairwork..

I dont know if they have decrease the tesselation level, or optimize it differently, it will be interessant to be able to compare the level of tesselation effectiveness or know what optimisation have been done.

I will be glad to see other reviewers of the initial release conduct their own test.

 

[Deleted member 13524]

Actually, the recent Catalyst versions have a dedicated Witcher 3 profile that AFAIK overrides the maximum tesselation factor to 32 or 64x, if you chose "AMD Optimized" in the driver's tesselation options.

The reviewer didn't mention what option he chose for the review, so there's a chance he's not actually making an apples-to-apples comparison.

 

[Razor1]

I don't have the game, but does Crysis 3 has the kind of "open sea" (i.e. extended to horizon) like it's predecessors have? All Crysis 3 scene shots I can find on the web only shows "contained" body of water in a more constrained level design. In which case, may not required to use similar water rendering techniques (camera aligned grid mesh) and thus could make comparison pointless.

Well Cry Engine 3 still has the "open sea", so that hasn't changed it can be turned off on a per level basis so that has been changed from Cry Engine 2. I'm not sure when that happened (which version of the engine)

 

[Novum]

It's used in the first level of Crysis 3

 

[Davros]

Another game that has the "open sea" thing is Lock On modern Air Combat

 

[lanek]

Another game that has the "open sea" thing is Lock On modern Air Combat

Lock On dont really exist anymore the UBisoft version anyway , it is DCS Lock-on 2 ..... even if it is still the light version of other DCS Simulator.

This said, Lock dont use displacement mapping and tesselation on water ( im a long long time user of DCS A10C simulator, DCS KA-50 Blackhawk and other recent stuff ) http://www.digitalcombatsimulator.com/en/

 

[rapso]

I heard that the water under the ground was only shown in wireframe mode.

hmm
http://docs.cryengine.com/display/SDKDOC2/Debug+Views#DebugViews-Wireframe

yes, that's the case, water is using occlusion queries to figure whether it should be rendered. in wireframe there is no solid terrain to occlude it, thus it is tagged as visible.

the cost of water is not the waterplane, but the reflection. everything in your fov is rendered twice. that's why the query is crucial, otherwise you run half framerate.

 

[lanek]

yes, that's the case, water is using occlusion queries to figure whether it should be rendered. in wireframe there is no solid terrain to occlude it, thus it is tagged as visible.

the cost of water is not the waterplane, but the reflection. everything in your fov is rendered twice. that's why the query is crucial, otherwise you run half framerate.

Dont the problem of this particular iteration of Cryisis iteration was that tesselation and displacement mapping was still computed.. putting the geometry of the AMD gpu#s to to knee with no apparrent reason ( as the other tesselation part related ? )..

personally, when this story have been out.. i have not much have take care of this, its when you look the series of it and what have happend on this period it start to be funny.

All cryengine games run excellently well on AMD gpu's .... but this one was a catastroph..

 

[rapso]

Dont the problem of this particular iteration of Cryisis iteration was that tesselation and displacement mapping was still computed.. putting the geometry of the AMD gpu#s to to knee with no apparrent reason ( as the other tesselation part related ? )..

to what computation are you referring by "still computed" if water is tagged as invisible and not rendered at all?

the proper way to validate the water culling in C2DX11 is to keep everything solid (not wireframe!) and to make fast motions between a point where you see water and where it is covered. Occlusion queries are one frame delayed (they report whether to water might have been visible the _last_ frame), hence if it wasn't visible last frame, it is not rendered the current frame, thus you'll see it pop in the next frame. (might work if you are rapidly rotating also).

 

[megavoxel]

to what computation are you referring by "still computed" if water is tagged as invisible and not rendered at all?

the proper way to validate the water culling in C2DX11 is to keep everything solid (not wireframe!) and to make fast motions between a point where you see water and where it is covered. Occlusion queries are one frame delayed (they report whether to water might have been visible the _last_ frame), hence if it wasn't visible last frame, it is not rendered the current frame, thus you'll see it pop in the next frame. (might work if you are rapidly rotating also).

I think he is referring to this article.
http://techreport.com/review/21404/crysis-2-tessellation-too-much-of-a-good-thing/2
http://techreport.com/review/21404/crysis-2-tessellation-too-much-of-a-good-thing/3
IIRC you were the senior graphics architect of Crytek? Would you mind making some clarification on this famous rumor?The author of this article is using a graphics debugger to test the game.