[GameGPU] Dying Light - Horrible game engine CPU optimizations

[zlatan]

The irony of the whole nvidia Gameworks program was the claim it made things easier for developers and was more efficient. Gameworks has turned out to be one of the worst things to happen to PC gaming in recent history.

Did you really believed that a new closed-source black box thing with unpredictable results will help the PC, especially an era when all the devs asking for more control?

 

[psolord]

I made a somewhat long gameplay video (22mins), recorded with an external recorder at 1080p/60fps, so the recording itself would not consume system resources at all. The OSD shows how the system gets strained at various parts.

Dying Light - PC Gameplay on gtx 970 @1.5ghz - 1080p 60fps

Once more, we witness cpu limits on a new AAA title.

I also made this video that illustrates how draw distance affects cpu performance and what the rest of the settings have to say about all this.

Dying Light - Graphics settings impact on performance - 1080p 60fps

All that on a GTX 970 boosting at 1.5Ghz and a core i5 2500k@4.8Ghz.

I will do proper benchmarking today on my other systems as well.

 

[n0x1ous]

Any win 8.1 vs win 7 major differences?

 

[monstercameron]

Seems crazy the game is using this much vram inside a building. I could understand if it was outside with the draw distance and what not... I still put Metro LL up there, as some of best looking textures(and game) with very little use of video memory. I mean running the game 1440p with 4XSSAA - barely uses 2GB of Vram.

not hard to accomplish when all the textures look the same.

 

[Spjut]

PCgameshardware.de also have CPU and GPU benchmarks up now
http://www.pcgameshardware.de/Dying-Light-PC-257307/Specials/Technik-Test-Benchmarks-1149149/

 

[Leyawiin]

Yup. See post #13 or Texture Quality section in the NV Guide:
http://www.geforce.com/whats-new/gu...performance-guide#dying-light-texture-quality

This game is crazy. Poor VRAM/texture caching optimizations, poor CPU multi-threading, poor quality AA, 2X the performance hit with max draw distance, and some users are reporting system crashes with 6GB of system memory or less.

And they griped about Watch Dogs. Sounds like this one makes it look extremely well optimized.

 

[.vodka]

PCgameshardware.de also have CPU and GPU benchmarks up now
http://www.pcgameshardware.de/Dying-Light-PC-257307/Specials/Technik-Test-Benchmarks-1149149/

4790k, 1080p + AA, maxed out, a 290x on par with a GTX 670, 29 FPS average, 26 FPS minimum.

Hey, let's all contribute a penny to building a 4790k/290x rig with an endless supply of LN2. Maybe if we get that 4790k to 6 ghz and that 290x to 1400 mhz, then maybe, and that's a huge maybe, we can break the 30 FPS minimum mark on AMD's cards.

Gameworks, a lovely thing.

 

[wand3r3r]

The irony of the whole nvidia Gameworks program was the claim it made things easier for developers and was more efficient. Gameworks has turned out to be one of the worst things to happen to PC gaming in recent history. Horribly unoptimized games with graphical bugs and loads of performance issues regardless of whether you are running nvidia's own cards or AMD's.

Witcher 3 may be next on the hit list of games to arrive with craptastic performance thanks to Gameworks.

Based on launch reviews and people griping about these latest GameWorks games I haven't bought any of them yet. I would actually be interested in a couple but am not willing to support a. buggy games and b. full retail price for buggy games.

 

[Erenhardt]

been fooling with ProcessLasso and fx6300
So the game runs better tied to 4 threads than when it can use all 6.
+50 CPU resources gives a negative of 3% improvement.
Also, the supposed "hack" with killing core 0 for a second to have load balancing across multiple cores gives no performance improvement. The core load is more spread with no cores hitting 100%, but the fps is the same.

 

[Ranulf]

Totalbiscuit covering view distance and SLI. He starts on about sli comparisons (980s I think) at the 14min mark or so:

https://www.youtube.com/watch?x-yt-...re=player_detailpage&x-yt-ts=1422411861#t=850

 

[n0x1ous]

Totalbiscuit covering view distance and SLI. He starts on about sli comparisons (980s I think) at the 14min mark or so:

https://www.youtube.com/watch?x-yt-...re=player_detailpage&x-yt-ts=1422411861#t=850

I wish he would have VRAM usage in his overlay. I know he had Titans for a long time - did not know he went to 980's

 

[Erenhardt]

I wish he would have VRAM usage in his overlay. I know he had Titans for a long time - did not know he went to 980's

They most likely sent him titans as well as 980. Just like bunch more mid-range cards for give-aways.

 

[Headfoot]

Yep, was going to get Watchdogs but after the performance fiasco there at launch I passed on it. Might pick it up for $5 of steam sale. I was going to get Dying Light too because it looked pretty cool but a single render thread in 2015 and completely crippled AMD performance? Going to skip this one too. I also was looking forward to AC: Unity but the performance issues and bugs there dissuaded me as well. 3 games now that lost my full price sale due to launch-day bugs and performance problems. I'll probably get them once they're $5

 

[cmdrdredd]

Yeah a lot of developers phone in their PC code, or copy over too much from the PS4/XB1 and do as little as possible to get it up to par.

 

[StrangerGuy]

So can we call Minesweeper in Windows 10 the greatest gaming achievement of all time because it can bog down an OCed Haswell-E and quad-SLI GTX 980s?

This industry is shoulder-deep into the shitter.

 

[RussianSensation]

Op, can you be more precise on the VRAM details. The graph you have shown does indeed show the game surpassing 3gb of Vram. But when I look at other charts provided the gtx 690 is doing quite well regrading VRAM. As is other gk104 cards vs Amds 3gb cards. When the dust settles is the 3gb going to matter that much on single GPU's. I'd really like to see some benches a few moths down the road. My take is there probably won't be enough difference between the two camps to seriously sway one way or the other regarding VRAM, as cards from both camps is going to lack performance without fine tuning settings.

Read some other posts in the thread. Per NV's reviewer's guide, with High textures and far draw distance, VRAM usage can range 3.3-4GB. That means 2GB cards will have more pop-in and 280X 3GB cards will be worse than 4GB 290. The texture pre-caching system doesn't mean that 2GB cards will run at 10-15 fps, which is good since it manages things dynamically. I will link any review that looks at how VRAM is affected. Waiting for HardOCP's analysis of the game.

If you are talking about 2GB vs. 3GB for Dying Light specifically, it matters less since the game is so heavily NV-favoured. If you are talking about 2GB vs. 3Gb in general, that's a completely different story. You pretty much need 3GB for 1080p for AC Unity, SoM, FC4, Titanfall, Wolfenstein NWO, etc. I wouldn't buy a $200 2GB card for 1080p in 2015, and professionals agree.
http://m.hardocp.com/article/2015/01/22/msi_geforce_gtx_960_gaming_video_card_review/10#.VMmRZUZXfCQ

 

[Head1985]

I made a somewhat long gameplay video (22mins), recorded with an external recorder at 1080p/60fps, so the recording itself would not consume system resources at all. The OSD shows how the system gets strained at various parts.

Dying Light - PC Gameplay on gtx 970 @1.5ghz - 1080p 60fps

Once more, we witness cpu limits on a new AAA title.

I also made this video that illustrates how draw distance affects cpu performance and what the rest of the settings have to say about all this.

Dying Light - Graphics settings impact on performance - 1080p 60fps

All that on a GTX 970 boosting at 1.5Ghz and a core i5 2500k@4.8Ghz.

I will do proper benchmarking today on my other systems as well.

Thanks for video.It looks like view distance really killing Fps.
BTW i remember you from like 4 years back in 5850 crossfire time.I used to watch your channel but after that you stop making new videos.
I was also using 5850 crossfire back then a now i have i5 2500K 5Ghz and GTX970 and you also have i5 2500K@4.8Ghz and GTX970:awe:

its good that you are back.

 

[Erenhardt]

If you are talking about 2GB vs. 3GB for Dying Light specifically, it matters less since the game is so heavily NV-favoured. If you are talking about 2GB vs. 3Gb in general, that's a completely different story. You pretty much need 3GB for 1080p for AC Unity, SoM, FC4, Titanfall, Wolfenstein NWO, etc. I wouldn't buy a $200 2GB card for 1080p in 2015, and professionals agree.
http://m.hardocp.com/article/2015/01/22/msi_geforce_gtx_960_gaming_video_card_review/10#.VMmRZUZXfCQ

Oh man, you don't even know.

I played through FC4 on 7870 2GB. Settings ultra -aa -fur and textures on high. It was running high 40 which I usually fine. But man it stuttered to the point I wanted to quit. I blamed the game for stutter.

I sold my 7870 before 960 came out, because I though I will upgrade. So after release I decided to look for something that is an actual upgrade. Had 290 or 280 on eye. Decided that upgrade to 3gb 280 for <$30 is well worth it.

At the same settings as 7870, the 280 uses 2.5GB of VRAM in FC4.
Now games run smooth again.

 

[TheELF]

5930K @ 4.4ghz showing 98% CPU usage on Core 1.

If you get 98% usage on two gpus with one render thread than whats the problem? (with the one thread approach)

Even if all the cores could be made to run at 100% the gpu(s) are already maxed out and cant give you anything more.

This game is just way too graphics demanding...
You can go to
C: \Users\username\Documents\DyingLight\out\settings
and edit the video.scr file in a simple text editor to lower the settings some more,like textures to low and even lower resolutions

 

[Magic Carpet]

Oh man, you don't even know.

I played through FC4 on 7870 2GB. Settings ultra -aa -fur and textures on high. It was running high 40 which I usually fine. But man it stuttered to the point I wanted to quit. I blamed the game for stutter.

I sold my 7870 before 960 came out, because I though I will upgrade. So after release I decided to look for something that is an actual upgrade. Had 290 or 280 on eye. Decided that upgrade to 3gb 280 for <$30 is well worth it.

At the same settings as 7870, the 280 uses 2.5GB of VRAM in FC4.
Now games run smooth again.

I have noticed similar behavior in Far Cry 4 with my 270. But still very much playable 45-60 FPS Vsync On (everything on max except God Rays / Fur / Shadows). However, with my 670 2GB I've experienced 0 issues in that game with nvidia settings at 1680x1050. Apparently Geforce is more optimized for this game. Of course, if you look at some misleading reviews, 270x is on par with 680 in this game, LMAO. I question every benchmark now. I have spent more than 100 hours playing it. 670 > 270, any time. Did not run into any vram bottlenecks at my settings in FC4. Well, I only had "smaa" and 1050p, though.

There are only two cards, I would consider upgrading to, R290 and GTX980. But I am not in a rush.

 

[Erenhardt]

At least on consoles it runs better than AC:U.
http://www.eurogamer.net/articles/digitalfoundry-2015-dying-light-performance-analysis

Strange that it is heavily CPU bottlenecked on PC, but runs better on 1.6GHz (ps4) jaguar than on 1,75 one (xbone)

Wonder how PC would handle console settings. I hope they update their test.

 

[SPBHM]

At least on consoles it runs better than AC:U.
http://www.eurogamer.net/articles/digitalfoundry-2015-dying-light-performance-analysis

Strange that it is heavily CPU bottlenecked on PC, but runs better on 1.6GHz (ps4) jaguar than on 1,75 one (xbone)

Wonder how PC would handle console settings. I hope they update their test.

well, it's the typical lower overhead for the CPU, and:

"The conclusion is that, as per usual, if you have the machine to back it up, Dying Light looks better on PC. It’s not only the resolution flexibility, HBAO and better anisotropic filtering, but the locked View Distance setting on PS4 is actually equivalent to 0% on PC, or quite possibly even less. That’s not to say the game looks bad on PS4, but the draw distance can be far from desired."

http://www.hardcoregamer.com/2015/01/28/dying-light-face-off-pc-vs-ps4/131644/

 

[KaRLiToS]

I have noticed similar behavior in Far Cry 4 with my 270. But still very much playable 45-60 FPS Vsync On (everything on max except God Rays / Fur / Shadows). However, with my 670 2GB I've experienced 0 issues in that game with nvidia settings at 1680x1050. Apparently Geforce is more optimized for this game. Of course, if you look at some misleading reviews, 270x is on par with 680 in this game, LMAO. I question every benchmark now. I have spent more than 100 hours playing it. 670 > 270, any time. Did not run into any vram bottlenecks at my settings in FC4. Well, I only had "smaa" and 1050p, though.

There are only two cards, I would consider upgrading to, R290 and GTX980. But I am not in a rush.

Of course it runs better on Nvidia, it is a gameworks title which in my opinion is killing the gaming industry.

 

[psolord]

Thanks for video.It looks like view distance really killing Fps.
BTW i remember you from like 4 years back in 5850 crossfire time.I used to watch your channel but after that you stop making new videos.
I was also using 5850 crossfire back then a now i have i5 2500K 5Ghz and GTX970 and you also have i5 2500K@4.8Ghz and GTX970:awe:

its good that you are back.

Ah yeah, good times. Our preferences seem to match my friend, although I am sure there will me thousands out there with a good 2500k and a good 970. Pretty nice couple those too!

I had a family misfortune and didn't feel like playing around with video making, benchmarks and the lot, but I am better now, so I'm back!

My 970 and 7950 benchmarks have finished uploading, but I will post all together when the 5850 and 570 ones have finished as well.

 

[Magic Carpet]

Of course it runs better on Nvidia, it is a gameworks title which in my opinion is killing the gaming industry.

I know but in some reviews (that obviously weren't benched with maximum settings), Kepler GPUs perform significantly worse than GCN v1.1.

I have both and played with both and when I changed from GTX 670 to 270 at the same settings I got a frame drop to as low as 7 fps as soon as I got in-game. Game was patched to 1.7, latest drivers. Just saying. I had to reduce settings in order to play smoothly.

Every review should be taken with a mega-grain of salt. Including this one in the OP.