[GameGPU] Dying Light - Horrible game engine CPU optimizations

[RussianSensation]

This game fun and running like butter on GTX 980 with everything max and draw distance to 70%.

Sorry i have to say this i tired this game on R9 290X Lighting Edition and i am not even 30fps out and no matter what setting i try even on low i am getting 30fps but if i use GTX 980 than i get average of 75fps at max settings.

Your data contradicts every professional review so far. In scenes where 980 flies at 70+ (indoor), the 290X also flies. Techspot tested this at 100% draw distance, even higher than your 70%. In outdoor scenes on the ground, the performance is absolutely brutal, with Total Biscuit and GameGPU showing 40-50 fps on 980 SLI. PCgameshardware also shows low 40s in a 980. Not 1 person in this thread with a 970/780Ti is even approaching your data. Furthermore, even NV's own testing with Max settings and 70% draw distance shows just 42-43 fps on a 980!
http://international.download.nvidi...view-distance-performance-overclocked-cpu.png

So in summary 3 professional reviews and NV's own testing all show peformance WAY lower than yours. It's literally impossible to believe your results, unless you cherry-picked some section where the 980 is averaging 75. Also, you expect us to believe that a 980 is 2.5x faster on average than a 290X Lightning? ... Right

 

[iiiankiii]

So far, all Gameworks titles run like crap. Yo, Gameworks FTW!!!

 

[RussianSensation]

So far, all Gameworks titles run like crap. Yo, Gameworks FTW!!!

In this case I genuinely believe it is not GW, but the game engine. CPU multi-threading issues with 1 core pegged at 98-100%, draw distance at 100% dropping performance in half with minimal difference to graphics, 3.3-4GB VRAM usage at 1080p at max draw distance, and major GPU under-utilization in SLI/CF modes -- hard to believe NV's GW's code could have broken so many things at once as these seem to run to the core of Chrome Engine 6. Having said that, for the end user the title still has GW's branding and the game does exhibit these major issues. You are right though that so far not 1 GW title is well optimized, and that is what we remember, not if it's the developer's or NV's fault.

What's interesting is even NV-centric regions like Russian players on GameGPU who love NV cards over AMD's are hating on GW's in the comments. Soon we will see even more GW titles to be able to draw more conclusions. IMO, TW3 will be a big title. Right now it seems GW targets 100% console ports, and just adds sprinkled graphical effects the developer chooses towards the end of development phase. Thus far this seems to have resulted in a poorly optimized console game that GW makes a "PC-enhanced" console version.

 

[desprado]

Your data contradicts every peofessional review so far. In scenes where 980 flies at 70+ (indoor), the 290X also flies. Techspot tested this at 100% draw distance, even higher than your 70%. In outdoor scenes on the ground, the performance is absolutely brutal, with Total Biscuit and GameGPU showing 40-50 fps on 980 SLI. PCgameshardware also shows low 40s in a 980. Not 1 person in this thread with a 970/780Ti is even approaching your data. Furthermore, even NV's own testing with Max settings and 70% draw distance shows just 42-43 fps on a 980!
http://international.download.nvidi...view-distance-performance-overclocked-cpu.png

So in summary 3 professional reviews and NV's own testing all show peformance WAY lower than yours. It's literally impossible to believe your results, unless you cherry-picked some section where the 980 is averaging 75. Also, you expect us to believe that a 980 is 2.5x faster on average than a 290X Lightning? ... Right

Techspot really?

U can check any benchmark of Dying Light which shows that GTX 780 is beating all AMD cards except Techspot which is no surprise.Techspot dont even show any Screenshot or video like most of benchmark site shows so it is kinds fake desperate attempt.

About My results i dont care what people say or think because i just posting what i have experienced.

 

[desprado]

In this case I genuinely believe it is not GW, but the game engine. CPU multi-threading issues with 1 core pegged at 98-100%, draw distance at 100% dropping performance in half with minimal difference to graphics, 3.3-4GB VRAM usage at 1080p at max draw distance, and major GPU under-utilization in SLI/CF modes -- hard to believe NV's GW's code could have broken so many things at once as these seem to run to the core of Chrome Engine 6. Having said that, for the end user the title still has GW's branding and the game does exhibit these major issues. You are right though that so far not 1 GW title is well optimized, and that is what we remember, not if it's the developer's or NV's fault.

What's interesting is even NV-centric regions like Russian players on GameGPU who love NV cards over AMD's are hating on GW's in the comments. Soon we will see even more GW titles to be able to draw more conclusions. IMO, TW3 will be a big title. Right now it seems GW targets 100% console ports, and just adds sprinkled graphical effects the developer chooses towards the end of development phase. Thus far this seems to have resulted in a poorly optimized console game that GW makes a "PC-enhanced" console version.

Bro seriously dont post.

There has been a patch of Dying light that has solved CPU problems in this game.All CPU core are at 50% usage.

Do some research before posting and dont miss guide people or post wrong info.

 

[Erenhardt]

In this case I genuinely believe it is not GW, but the game engine. CPU multi-threading issues with 1 core pegged at 98-100%, draw distance at 100% dropping performance in half with minimal difference to graphics, 3.3-4GB VRAM usage at 1080p at max draw distance, and major GPU under-utilization in SLI/CF modes -- hard to believe NV's GW's code could have broken so many things at once as these seem to run to the core of Chrome Engine 6. Having said that, for the end user the title still has GW's branding and the game does exhibit these major issues. You are right though that so far not 1 GW title is well optimized, and that is what we remember, not if it's the developer's or NV's fault.

What's interesting is even NV-centric regions like Russian players on GameGPU who love NV cards over AMD's are hating on GW's in the comments. Soon we will see even more GW titles to be able to draw more conclusions. IMO, TW3 will be a big title. Right now it seems GW targets 100% console ports, and just adds sprinkled graphical effects the developer chooses towards the end of development phase. Thus far this seems to have resulted in a poorly optimized console game that GW makes a "PC-enhanced" console version.

It probably is GW running on CPU with amd cards.
Other than that, amd do not have driver for that game, and will probably take a long time to have one thanks to anti-gamer practices from nv partnership program.

Alos, nice optimizations:
"Lets disable some options by default to make it run better".
That is not how you optimize games!:thumbsdown:

 

[RussianSensation]

Bro seriously dont post.

There has been a patch of Dying light that has solved CPU problems in this game.All CPU core are at 50% usage.

Do some research before posting and dont miss guide people or post wrong info.

I am not misguiding anyone. My post actually defends GW and the patch resolving some CPU threading issues shows GW is unlikely to blame.

You keep comparing TechSpot's review which used the latest patch vs. earlier reviews that didn't. As far as you dismissing TechSpot's results overall and site as amateurish, it's one of the oldest sites and has provided unbiased coverage for years. Something that can't be said for TR, PCPersp, or Hardware Canucks. Also, their review of a 960 showed poor frame times and noted the poor price/performance of the card. I would believe TechSpot over those other sites any day since the owner is from New Zealand and doesn't get any special treatment that NV provides to the North American reviewers. More so, he actually uses after-market 290/290X cards in his review and doesn't make moronic statements how all R9 290s run hot and loud, if you ever paid attention. All of these are clear signs TechSpot still has integrity.

Sorry, but even if we disregard AMD's performance, your claim that a 980 with max settings and 70% draw distance averages 75 fps+ in the game is BS.

 

[parvadomus]

It will be fun when AMD release a driver for this and the 780ti gets killed by 290s like every latest game.

 

[SPBHM]

(they also have numbers for 100%, but considering the performance drop it makes no sense using it)

http://pclab.pl/art61159-9.html

they also have numbers for medium, low and 1440P

http://pclab.pl/art61159-6.html

 

[Silverforce11]

The only difference Techspot does to other sites is they don't run with GameWork's HBAO+

Could this be the major performance killer for AMD cards?

http://www.techspot.com/review/956-dying-light-benchmarks/
"There are no quality presets in Dying Light so you must set everything up. For testing we maxed out every setting with the exception of Nvidia HBAO+ to try and keep the results consistant across AMD and Nvidia cards. That said, we did leave 'Nvidia Depth of Field' enabled."

Also, it's strange to see a developer list those features as "Nvidia", its definitely not the norm to see that.

 

[DownTheSky]

Could it be that higher driver overhead is killing performance on AMD cards?

Also any of you guys tried OCing the system RAM? Because both ~50% GPU utilization and ~50% CPU utilization in a game, it's just weird. The bottleneck has to be somewhere.

 

[sontin]

I guess Techspot is using an indoor scene which is mostly limitated by the GPU power.

PCGH and pclab are using an outdoor scene which needs more cpu power and reflects the quality of the driver overhead.

 

[Erenhardt]

Are there any dust, leafs, hair, cloth or other GW effects that run on CPU when amd GPU is present in this games?

 

[Enigmoid]

Your data contradicts every professional review so far. In scenes where 980 flies at 70+ (indoor), the 290X also flies. Techspot tested this at 100% draw distance, even higher than your 70%. In outdoor scenes on the ground, the performance is absolutely brutal, with Total Biscuit and GameGPU showing 40-50 fps on 980 SLI. PCgameshardware also shows low 40s in a 980. Not 1 person in this thread with a 970/780Ti is even approaching your data. Furthermore, even NV's own testing with Max settings and 70% draw distance shows just 42-43 fps on a 980!
http://international.download.nvidi...view-distance-performance-overclocked-cpu.png

So in summary 3 professional reviews and NV's own testing all show peformance WAY lower than yours. It's literally impossible to believe your results, unless you cherry-picked some section where the 980 is averaging 75. Also, you expect us to believe that a 980 is 2.5x faster on average than a 290X Lightning? ... Right

People please read the techspot review.

Post-test notes:
After further investigation we have learned a few things about Dying Lights performance with the latest v1.2 patch. The performance we found indoors is accurate, however many reviews show performance to be around half what we saw. The issue seemed to be the draw distance, which doesn’t have an impact inside, even when looking out over the landscape as we did.
Our test does include a section where the player walks to the edge of the tower and looks out over the landscape, here the frame rates actually increase significantly which was unexpected.
As it turns out the draw distance isn’t an issue here either as finer details are not displayed. It isn’t until you hit ground level and walk outside that the frame rates take a huge hit with the draw distance set to 100%. Backing the draw distance off to 50% restores performance and gamers will see the numbers we showed in this article.
It is believed that the draw distance performance at 100% can be cured via a patch and if this is the case then the performance we showed inside will be much the same outside as well. For now although we tested with 100% draw distance the results should be treated as though the distance was set to 50%.

 

[SPBHM]

another source claiming consoles are bellow 0%

"The particular setting responsible for most of the performance woes faced in the PC version is the view distance slider. Increasing this to its maximum value brings the frame-rate on a Core i5-3570K/GTX 780 PC well under 60fps. By dropping to 35 per cent or less, we find that it is possible to attain a stable 60fps on our system, but going beyond this - particularly over 50 per cent - sees drops performance dramatically. We examined the view distance slider at various intervals to determine how it stacks up on consoles and the results are interesting - the Xbox One and PS4 setting appears to fall below the PC version's lowest possible setting. It's just a hair lower, of course, but it is lower nonetheless, which may help to explain the high CPU load we see on PC."

http://www.eurogamer.net/articles/digitalfoundry-2015-dying-light-face-off

interesting observation:

"Strange as it may seem, the higher view distance settings actually have a somewhat negative impact on image quality. The low detail trees used at a distance while using lower settings are mostly free of aliasing due to their simplistic shapes. However, on the maximum setting, full detail models are visible into the distance, introducing a lot of unsightly shimmering."

 

[Qwertilot]

They even seem to find it playing OK at 1080 with an i3/750ti
(With sensibly modest settings of course! Actually intriguing to see if anything will overwhelm a 750ti before the end of this console generation.).

Have to say that using huge chunks of vram to stamp on pop in seems a commendably sane choice compared to using it for massive, very subtly better, textures.

 

[desprado]

another source claiming consoles are bellow 0%

"The particular setting responsible for most of the performance woes faced in the PC version is the view distance slider. Increasing this to its maximum value brings the frame-rate on a Core i5-3570K/GTX 780 PC well under 60fps. By dropping to 35 per cent or less, we find that it is possible to attain a stable 60fps on our system, but going beyond this - particularly over 50 per cent - sees drops performance dramatically. We examined the view distance slider at various intervals to determine how it stacks up on consoles and the results are interesting - the Xbox One and PS4 setting appears to fall below the PC version's lowest possible setting. It's just a hair lower, of course, but it is lower nonetheless, which may help to explain the high CPU load we see on PC."

http://www.eurogamer.net/articles/digitalfoundry-2015-dying-light-face-off

interesting observation:

"Strange as it may seem, the higher view distance settings actually have a somewhat negative impact on image quality. The low detail trees used at a distance while using lower settings are mostly free of aliasing due to their simplistic shapes. However, on the maximum setting, full detail models are visible into the distance, introducing a lot of unsightly shimmering."

So i was kind of right that this game meant to be played on PC.

 

[Pneumothorax]

Does AMD have a history of fixing this kind of performance disparity and potentially enabling CF for games like this? Just concerned as I already got rid of my 3 970's and have 290x on their way.

 

[.vodka]

Don't base what your 290x can do, or can't, on an outlier of a game which is less optimized than Unity. That is an achievement all by itself.

It'll eventually get patched and will probably become playable on a wide range of hardware. It'll also be $5 by then.

 

[5150Joker]

Does AMD have a history of fixing this kind of performance disparity and potentially enabling CF for games like this? Just concerned as I already got rid of my 3 970's and have 290x on their way.

I wouldn't count on it. You'll be missing the green team once the anger subsides and you start getting more acquainted with AMD drivers and support. I stubbornly suffered through three consecutive generations worth before permanently giving up on them. Btw this game runs great on my nvidia based pc, I bet most of the haters have AMD hardware.

 

[desprado]

nvm.

 

[desprado]

Don't base what your 290x can do, or can't, on an outlier of a game which is less optimized than Unity. That is an achievement all by itself.

It'll eventually get patched and will probably become playable on a wide range of hardware. It'll also be $5 by then.

it is playable on all Nvidia cards but i dont know about AMD.

As DF said that GTX 750 ti and core i3 is managing to get some what equal or better performance than PS4 in this game.

 

[.vodka]

I wouldn't count on it. You'll be missing the green team once the anger subsides and you start getting more acquainted with AMD drivers and support. I stubbornly suffered through three consecutive generations worth before permanently giving up on them. Btw this game runs great on my nvidia based pc, I bet most of the haters have AMD hardware.

May I ask what three generations worth of poor AMD support have you suffered from? Please detail your experience so we can all learn from it.

it is playable on all Nvidia cards but i dont know about AMD.

As DF said that GTX 750 ti and core i3 is managing to get some what equal or better performance than PS4 in this game.

Which boggles the mind, since the game was developed for PS4 and XB1. Should I remind everyone both consoles have GCN hardware inside?

For the PC port, one of the additions has been Gameworks, and you see a 290x hanging with a GTX 670 on one review made by a German site a few pages back, for example. Go figure where the problem is.

Give it a few patches to sort out gameworks' effect on AMD's cards and it'll be fine.

edit: All of this apart from the already weird results this game puts out. OP has all the release day information, just look at these CPU graphs. It's insane to even take this game seriously with that kind of results, confirmed by other sites... at least techspot's results with 50% draw distance paints a much more balanced view.

edit2: what happened to vendor agnostic games, like a few years back when you could enjoy a good game with either ATI/AMD or NV? TWIMTBP/Gameworks and GE aren't doing the gaming industry any good like this.

 

[desprado]

May I ask what three generations worth of poor AMD support have you suffered from? Please detail your experience so we can all learn from it.



Which boggles the mind, since the game was developed for PS4 and XB1. Should I remind everyone both consoles have GCN hardware inside?

For the PC port, one of the additions has been Gameworks, and you see a 290x hanging with a GTX 670 on one review made by a German site a few pages back, for example. Go figure where the problem is.

Give it a few patches to sort out gameworks' effect on AMD's cards and it'll be fine.

edit: All of this apart from the already weird results this game puts out. OP has all the release day information, just look at these CPU graphs. It's insane to even take this game seriously with that kind of results, confirmed by other sites... at least techspot's results with 50% draw distance paints a much more balanced view.

Which Patch are u talking about?

Assassins Creed 4 and unity have still poor performance on AMD and sames goes for Watch dogs, Lord of fallen and MGS V which has been out for 3 months.

 

[NickelPlate]

What a pig of a game. No room in my library for shat like this.