Does Gameworks kill performance? Lets find out...as a community

D.P. · 10 Nov 2015 at 00:13

Rroff said:
Problem is for a really meaningful comparison you need to compare a feature implemented via GameWorks with the same feature implemented by someone other than nVidia (with the assumption that the developers are half competent) if an equivalent feature implemented by someone else runs significantly faster it would raise questions.

Unfortunately this is quite a hard thing to do short of finding someone with the time to build tech demos using both from the ground up.

at least someone else gets it!

Rofflay · 10 Nov 2015 at 01:33

AWPC said:
Gameworks is Nvidia's SLI marketing buy another card strategy. Obviously it will cripple performance its designed around SLI minimum even though in some games a high end GPU is enough.

Since 2011 & Batman Arkham City (which runs fully maxed out nowadays easily on most single GPU's) this seems to be Nvidia's strategy add Gameworks force highend gamers to buy another GPU from them or wait a few years until current single GPU's are fast enough.

At 4K Gameworks is nigh-unplayable on one card. The first thing i do personally is work my way up from 4K Textures max 16AF no AA and see how it goes. I would not buy another card just for Gameworks or effects when you would be suprised how good something like BF4 looks with the tard effects off. Anyone who tried the Ultra preset will obviously be the prime target audience for Gameworks.

shankly1985 · 10 Nov 2015 at 01:43

D.P. said:
Why has W3 hair-works tessellation got anything to do with nvidia, it is the developer that failed to set the correct tessellation factor.

You sure about that? Becuase last I heard GameWorks Files are build in house, no one can change them. This is why all this Black box came about..

GameWorks are files game devs can use in games but can not change how they effect things..

Please do correct me if am wrong.

D.P. · 10 Nov 2015 at 02:01

shankly1985 said:
You sure about that? Becuase last I heard GameWorks Files are build in house, no one can change them. This is why all this Black box came about..

GameWorks are files game devs can use in games but can not change how they effect things..

Please do correct me if am wrong.

You are wrong. GamesWorks are highly configurable, it all comes down to the developers preferences. It is a library like any other library, it all comes down to how you use it and the configuration you give the underlying functions.

Plus if you want you can get the source code for all of it if you ask Nvidia. I use several numerical optimization libraries. If you don't take take the computer is dead for a week!

Mei · 10 Nov 2015 at 02:11

Dicehunter said:
4790K @ 4.80GHZ + 980 Ti @ 1500/2000

Batman Arkham Knight 1440P Max Settings

All Gameworks Settings On

Min: 43
Max: 109
Avg: 83

Gameworks With Smoke Off

Min: 67
Max: 143
Avg: 101

Gameworks Off

Min: 90
Max: 168
Avg: 117

whyscotty said:
5930K @ 4.40GHZ + 980 Ti @ 1367/1774

Batman Arkham Knight 1440P Max Settings

All Gameworks Settings On

Min: 37
Max: 100
Avg: 69

Gameworks With Smoke Off

Min: 46
Max: 124
Avg: 88

Gameworks Off

Min: 56
Max: 144
Avg: 105

some pretty beasty systems there!,
i want see more ppl gaming at 4k next year, with 1440p still being a struggle i dont see it happen!

Raviolay · 10 Nov 2015 at 02:21

When I was a Nvidia user, I never turned on Physx in games as I lost too much performance. Even though I was running a GTX 690 at the time, I think if GameWorks were to be centered around performance increases. Instead of visual flair, I would jump ship back quite rapid.

The acid test for me would be a game that uses HairWorks or TressFX.......

thebennyboy · 10 Nov 2015 at 02:33

R9 290X @ 1100/1450
i7 4790K @ 4.7Ghz

-- Batman - Arkham Knight --
1080p - Max Settings

97 FPS - Enhanced Rain and Enhanced Light Shafts
104 FPS - No GameWorks
6.7% Performance Hit

-- FarCry 4 --
1080p - Max Settings

75 - FurWorks and GodRays
86 - No GameWorks
12.8% Performance Hit

No idea whether you wanted AMD scores, thought i'd contribute anyway.

james.miller · 10 Nov 2015 at 05:37

D.P. said:
You are wrong. GamesWorks are highly configurable, it all comes down to the developers preferences. It is a library like any other library, it all comes down to how you use it and the configuration you give the underlying functions.

Plus if you want you can get the source code for all of it if you ask Nvidia. I use several numerical optimization libraries. If you don't take take the computer is dead for a week!

Not only that but people found out the amount of tess and Aa when somebody found the hair works options in one of the configuration files. We were mucking about wit it as soon as we found.

Dave2150 · 10 Nov 2015 at 06:16

D.P. said:
You are wrong. GamesWorks are highly configurable, it all comes down to the developers preferences. It is a library like any other library, it all comes down to how you use it and the configuration you give the underlying functions.

Plus if you want you can get the source code for all of it if you ask Nvidia. I use several numerical optimization libraries. If you don't take take the computer is dead for a week!

What GPU do you own, D.P.? Last I read you were a console gamer who just liked talking about PC GPU's.

pgi947 · 10 Nov 2015 at 07:33

D.P. said:
Your results are completely flawed because the only way to make a fair comparison is to show the same game running with the EXACT same effects but coded in a different way using a different 3rd party library or proprietary solution.

Of course Gamesworks effects has a performance cost, by definition they are complex effects that developers may not have the ability to do themselves. You can;t get those effects for free.

I Suggest you close the thread and come back when you can make an actual comparison.

The comparison being made is between the vanilla game and then with gameworks features added. At no point have I suggested those features are free, the clues in the thread title, does enabling it hamper performance to the level of making it unplayable.

thebennyboy said:
R9 290X @ 1100/1450
i7 4790K @ 4.7Ghz

-- Batman - Arkham Knight --
1080p - Max Settings

97 FPS - Enhanced Rain and Enhanced Light Shafts
104 FPS - No GameWorks
6.7% Performance Hit

-- FarCry 4 --
1080p - Max Settings

75 - FurWorks and GodRays
86 - No GameWorks
12.8% Performance Hit

No idea whether you wanted AMD scores, thought i'd contribute anyway.

AMD results are especially welcome given the whole can they can't they optimise for gameworks debates

Will update this evening, any fallout 4 results?

Nexus18 · 10 Nov 2015 at 08:08

What about the gameworks smoke effects bennyboy, think you may have forgot to turn that on

chroniclard · 10 Nov 2015 at 08:33

TNA said:
Not a fair comparison really.

Crysis actually looked perceivable MUCH better than anything else in 2007 though, you could see why you needed the extra power. These effects make little difference for the fps drop you get.

Imo, its exactly the same. These effects are complex additions, that while they made not add a huge amount(and often are implemented poorly by the developers) they take much more processing power. Just like Crysis added loads.

In a couple of years GPU's will be churning out higher frame rates with all the effects on.

thebennyboy · 10 Nov 2015 at 10:49

Nexus18 said:
What about the gameworks smoke effects bennyboy, think you may have forgot to turn that on

In Batman? The option was greyed out. Not allowed to use it.

Avenged7Fold · 10 Nov 2015 at 11:04

shankly1985 said:
What I would like to just say nice idea on the thread, but would it not be better to say compare the feature against another. So lets say Godrays vs standard performance plus a screenshot of the two?

Just an idea

+1

Screenshot will show the image quality difference and fps show the performance hit. With both you can see what you are trading for and make a judgement on it. I dont know how good recording software but even a 10 second clip would be nice, although a bit harder to control and replicate the same conditions.

Problem with a thread like this (which i think is a great idea) is that some of the people are set on a witch hunt and others are out to ardently defend, where as it should be looked at subjectively. eg. This is the performance drop, this is what it looks like with it on and with it off.

Personal judgements on why it is can be reserved for after that and in reality, reasons for the performance hit matter little compared to what the performance hit is and what you get for it.

pgi947 · 10 Nov 2015 at 11:39

Avenged7Fold said:
+1

Screenshot will show the image quality difference and fps show the performance hit. With both you can see what you are trading for and make a judgement on it. I dont know how good recording software but even a 10 second clip would be nice, although a bit harder to control and replicate the same conditions.

Problem with a thread like this (which i think is a great idea) is that some of the people are set on a witch hunt and others are out to ardently defend, where as it should be looked at subjectively. eg. This is the performance drop, this is what it looks like with it on and with it off.

Personal judgements on why it is can be reserved for after that and in reality, reasons for the performance hit matter little compared to what the performance hit is and what you get for it.

I'm going to update op tonight with what games have what gameworks available, I can also throw in the nvidia on/off videos for what I can find and link the performance guides that have those slider images to show it on/off?

I'm sat firmly on the fence with gameworks, some of it looks nice in demos but doesn't add anything to gameplay

Nexus18 · 10 Nov 2015 at 12:00

thebennyboy said:
In Batman? The option was greyed out. Not allowed to use it.

Oh that's interesting! I thought gameworks could be used by all....

The smoke is what absolutely kills FPS.

razorpakk · 10 Nov 2015 at 12:19

Why are you bothering with such a broken game like Arkham Knight?
They are extra effects, of course the performance is impacted.

I'm confused by humanity

.

opethdisciple · 10 Nov 2015 at 12:20

Obviously turning GW off improves performance.

Buy why stop there? Just turn every down to the lowest settings and see frames skyrocket. :rolleyes:

pgi947 · 10 Nov 2015 at 13:22

Again, no one is doubting they impact performance, of course they do, what we want to see is to what level performance is impacted. Is the impact the same for both vendors, does one suffer with certain things more then the other, does the impact drag frame rates to an unplayable level etc...

D.P. · 10 Nov 2015 at 13:52

pgi947 said:
Again, no one is doubting they impact performance, of course they do, what we want to see is to what level performance is impacted. Is the impact the same for both vendors, does one suffer with certain things more then the other, does the impact drag frame rates to an unplayable level etc...

That is also irrelevant until you show a competing technology sow the same visual quality. Then you could make all sorts of interesting observations, but otherwise it is pointless.

Things like "unplayable FPS" is entirely down to the developer and what settings they use. GamesWorks features are extremely well optimized and if a developer doesn't create the correct configuration profiles for different hardware then it is not Nvidia's fault.

And of course there will be differences between vendors, AMD have woeful tessellation for example, while Fiji cards should have better pixel pushing capabilities than Maxwell. So performance will all depend on the actual effect, what the develop configures, everything else that is being rendered on the screen, drivers and lastly the GamesWorks code.