• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

Project CARS (crowdsourcing) racing sim by Slightly Mad Studios (fully funded 3.75M)

KainXVIII

Member
dat optimization..

http--www.gamegpu.ru-images-stories-Test_GPU-Simulator-Project_CARS_2015-test-pc_1920.jpg

http--www.gamegpu.ru-images-stories-Test_GPU-Simulator-Project_CARS_2015-test-pc_2560.jpg

http--www.gamegpu.ru-images-stories-Test_GPU-Simulator-Project_CARS_2015-test-pc_3840.jpg
 
I think that framerate for 1080p is way too low (for PCars graphics), maybe MSAA tanked it?

MSAA is the least expensive AA option outside of FXAA in the game. That being said DS2X probably runs very slightly worse yet provides a significantly better IQ. DS4X though is much more taxing than both of these, let alone DS8X.

This game truly is taxing on PCs though. In my own experience (i5-2500K @ 4.4 GHz + GTX 970 with slight OC at 1080p), I need to run the game at around medium settings and MSAA to get 45-50ish fps in the heaviest stress test pCARS can provide: 55 AI cars in heavy rain at Le Mans.

However outside of this I run a mix of ultra and high settings + DS2X AA and it does me well for most situations in dry conditions with up to around 23 AI, good enough for most situations both in career and online. Wet weather still heavily impacts my framerate as it's heavy on both CPU and GPU, so considering my CPU already is constantly being hit during races with plenty of AI, it just can't cope. Wet weather without AI generally stays above 50 fps though.

This is the first game that has ever pushed my trusty old 2500K this hard, not even GTAV pushes it this much.

I mean it looks great, but if a 5960X and a Titan X are required to get above 60 fps at 1080p with not even the most expensive AA solution used, something must be up.

The menus alone use 50% of my CPU, which has to be the most CPU heavy game menu I've ever come across haha.
 

Mascot

Member
MSAA is the least expensive AA option outside of FXAA in the game. That being said DS2X probably runs very slightly worse yet provides a significantly better IQ. DS4X though is much more taxing than both of these, let alone DS8X.

This game truly is taxing on PCs though. In my own experience (i5-2500K @ 4.4 GHz + GTX 970 with slight OC at 1080p), I need to run the game at around medium settings and MSAA to get 45-50ish fps in the heaviest stress test pCARS can provide: 55 AI cars in heavy rain at Le Mans.

However outside of this I run a mix of ultra and high settings + DS2X AA and it does me well for most situations in dry conditions with up to around 23 AI, good enough for most situations both in career and online. Wet weather still heavily impacts my framerate as it's heavy on both CPU and GPU, so considering my CPU already is constantly being hit during races with plenty of AI, it just can't cope. Wet weather without AI generally stays above 50 fps though.

This is the first game that has ever pushed my trusty old 2500K this hard, not even GTAV pushes it this much.

I mean it looks great, but if a 5960X and a Titan X are required to get above 60 fps at 1080p with not even the most expensive AA solution used, something must be up.

The menus alone use 50% of my CPU, which has to be the most CPU heavy game menu I've ever come across haha.

Makes the PS4 and Xbone performance even more impressive then, oddly enough. Why does such weaker console hardware seem to cope so admirably?
 

ShamePain

Banned
Makes the PS4 and Xbone performance even more impressive then, oddly enough. Why does such weaker console hardware seem to cope so admirably?

Because they cut a lot of corners and you only ever get 60 fps during daytime in less crowded situations. That and secret sauce of course.
 

Azzawon

Member
Makes the PS4 and Xbone performance even more impressive then, oddly enough. Why does such weaker console hardware seem to cope so admirably?

I would think it's because of the locked down hardware and memory allocation. The devs will have a set amount to play around with on the consoles and can optimise all they want to make it work, whereas a PC isn't just a gaming machine and has an endless combination of specs (to put it quickly)
 

Dilly

Banned
dat optimization..

http://gamegpu.ru/images/remote/http--www.gamegpu.ru-images-stories-Test_GPU-Simulator-Project_CARS_2015-test-pc_1920.jpg[IMG]
[IMG]http://gamegpu.ru/images/remote/http--www.gamegpu.ru-images-stories-Test_GPU-Simulator-Project_CARS_2015-test-pc_2560.jpg[IMG]
[IMG]http://gamegpu.ru/images/remote/http--www.gamegpu.ru-images-stories-Test_GPU-Simulator-Project_CARS_2015-test-pc_3840.jpg[IMG][/QUOTE]

The wonderful thing about games developed with PC as primary platform is the addition of insane graphical settings.
 
MSAA is the least expensive AA option outside of FXAA in the game. That being said DS2X probably runs very slightly worse yet provides a significantly better IQ. DS4X though is much more taxing than both of these, let alone DS8X.

This game truly is taxing on PCs though. In my own experience (i5-2500K @ 4.4 GHz + GTX 970 with slight OC at 1080p), I need to run the game at around medium settings and MSAA to get 45-50ish fps in the heaviest stress test pCARS can provide: 55 AI cars in heavy rain at Le Mans.

However outside of this I run a mix of ultra and high settings + DS2X AA and it does me well for most situations in dry conditions with up to around 23 AI, good enough for most situations both in career and online. Wet weather still heavily impacts my framerate as it's heavy on both CPU and GPU, so considering my CPU already is constantly being hit during races with plenty of AI, it just can't cope. Wet weather without AI generally stays above 50 fps though.

This is the first game that has ever pushed my trusty old 2500K this hard, not even GTAV pushes it this much.

I mean it looks great, but if a 5960X and a Titan X are required to get above 60 fps at 1080p with not even the most expensive AA solution used, something must be up.

The menus alone use 50% of my CPU, which has to be the most CPU heavy game menu I've ever come across haha.

i5 4690 and a GTX960 gives me locked 60 in every stress tress i've tried with all settings to low except car and track which are medium with MSAA and low FXAA. I've recently bought a 970 and that has allowed me to up everything across the board to high except grass (which has a known bug with nvidia which will be gone in D1P) and keep the same locked frame rate.

Personally I actually prefer the MSAA + FXAA combo to DS2X, I find the DS2X gives shimmering on poles etc.

Crazy thing is though i'll likely play mostly on PS4 as thats where my buddies will race
 
A few corrections/updates to your analysis:

- PS4 uses EQAA(MSAA) and not FXAA.
- PS4 shadow detail is between the medium to high equivalent PC setting - more precisely PC medium shadow settings are 1024x4096 with 16 bit precision and PS4 is the same resolution but with 32 bit shadow precision.
- PS4 track detail/quality is equivalent to the high PC setting.
- PS4 car detail is equivalent to high with lod scaling between medium to high for the car shadows.
- Imposter reflections are activated on both consoles in the D1P after a concerted optimisation effort.

Renderman, your insights into the game's engine and technical settings has been beyond awesome. Do you think you could perhaps do a write-up at some point about your opinions on the different versions, how the main rendering works, and what changes youw ould like to see/ would have liked to committ? I would be "eternally grateful."

It is awesome to finally have a dev (dev team in this case) post about the technical side of their game!!!
 

MaLDo

Member
It is awesome to finally have a dev (dev team in this case) post about the technical side of their game!!!

I wonder if I've striked a frail chord comparing screenshots. Something like what happened before with that performance analysis in DF. Mainly because what he says about shadow resolution and LODS doesn't correspond with any console gameplay out there. The only reasonable explanation is they have achieved a lot of progress in those last weeks and they are proud enough to correct every wrong statement in the wild.
 
I wonder if I've striked a frail chord comparing screenshots. Something like what happened before with that performance analysis in DF. Mainly because what he says about shadow resolution and LODS doesn't correspond with any console gameplay out there. The only reasonable explanation is they have achieved a lot of progress in those last weeks and they are proud enough to correct every wrong statement in the wild.

To be fair maldo, your statements aren't just any old forum comment. Your voice carries wait given your pedigree. :D

And yeah, I guess they have made some serious progress since the DF build that is not evident to see in much if any of the footage out there. I am also curious why the AA resolve for the EQAA seems so inconsistent. This game is forward rendered right? Is the MSAA coming in too early in the rendering chain (HDR breaking it?)?
 

ShamePain

Banned
Is that entirely true..?

That was true for the old analysis. I doubt few months of optimization will ensure a leap from 30s to solid 60 in stressful situations, the game already has a lot going on, on paper it's more ambitious than any FM/GT in terms of tech it's pushing.
 

ShamePain

Banned
Also pre-release PC benchmarks can be pretty inaccurate. I remember when Witcher 2 released it could barely manage 30 fps on my GTX 580, then Nvidia released new drivers and up you go I had locked 60 fps.
 

Mascot

Member
That was true for the old analysis. I doubt few months of optimization will ensure a leap from 30s to solid 60 in stressful situations, the game already has a lot going on, on paper it's more ambitious than any FM/GT in terms of tech it's pushing.

I don't think this was even true for the old analysis. There was lots of comment in the old-build VVV videos (and from memory in the DF analysis) that the game was consistently hitting 60fps in dry weather and at night, even with 40-odd car grids. The performance hits were mainly during heavy precipitation. With "leap from 30 to a solid 60" you make it sound like pCARS is a 30fps game that is trying to achieve 60fps. Isn't it fairer to say that it's a 60fps game that occasionally dips below 60 in extraordinarily loaded situations?
 

ShamePain

Banned
I don't think this was even true for the old analysis. There was lots of comment in the old-build VVV videos (and from memory in the DF analysis) that the game was consistently hitting 60fps in dry weather and at night, even with 40-odd car grids. The performance hits were mainly during heavy precipitation. With "leap from 30 to a solid 60" you make it sound like pCARS is a 30fps game that is trying to achieve 60fps. Isn't it fairer to say that it's a 60fps game that occasionally dips below 60 in extraordinarily loaded situations?

Isn't that what I was saying?
 

Helznicht

Member
i5 4690 and a GTX960 gives me locked 60 in every stress tress i've tried with all settings to low except car and track which are medium with MSAA and low FXAA. I've recently bought a 970 and that has allowed me to up everything across the board to high except grass (which has a known bug with nvidia which will be gone in D1P) and keep the same locked frame rate.

Personally I actually prefer the MSAA + FXAA combo to DS2X, I find the DS2X gives shimmering on poles etc.

Crazy thing is though i'll likely play mostly on PS4 as thats where my buddies will race

So with the developer confirming the PS4 being set to High to Med/High settings in the PS4 thread, this means PS4 = I5 & GTX 970 performance?

dat optimization..

More like:
"dat PC optimization though.."
 

Ryne

Member
dat optimization..

Guess I will need a new computer to maximize this game to get 60 FPS steady.

I know there is the incorrect obsession with maxing everything on the PC, however this is one game I set the goal of maximizing because it looks so damn great.

Gonna give it a shot with my i72600K OC'd and GTX 980 combo.
 

orava

Member
People are still surprised that a graphics heavy PC game at max settings do not run that well on their not so high end PCs?
 

SMSRenderTeam

Neo Member
To be fair maldo, your statements aren't just any old forum comment. Your voice carries wait given your pedigree. :D

And yeah, I guess they have made some serious progress since the DF build that is not evident to see in much if any of the footage out there. I am also curious why the AA resolve for the EQAA seems so inconsistent. This game is forward rendered right? Is the MSAA coming in too early in the rendering chain (HDR breaking it?)?

Project CARS uses a light pre-pass renderer.... (there is no forward rendering aside from transparencies, which is standard for this sort of renderer)

There are actually two EQAA resolves, one before Motion-Blur (which is executed in HDR) and one before tone-mapping / bloom composition. On PC with motion-blur turned off, each sample of the MSAA target is tone-mapped individually - however since console always has MB enabled per sample tone-mapping is switched off. (This is a driving game after all!)

A more exacting break-down of the shadow detail:

Shadows (CSM)

PC Low - 512x2048, 16bit, 2 spot shadows
PC Medium - 1024x4096, 16bit, 2 spot shadows
Console -1024x4096, 32bit, 4 spot shadows
PC High - 1408x5632, 32bit, 4 spot shadows
PC Ultra - 2048x8192, 32bit , 4 spot shadows + enables shadows in Environment Maps
 

MaLDo

Member
Project CARS uses a light pre-pass renderer.... (there is no forward rendering aside from transparencies, which is standard for this sort of renderer)

There are actually two EQAA resolves, one before Motion-Blur (which is executed in HDR) and one before tone-mapping / bloom composition. On PC with motion-blur turned off, each sample of the MSAA target is tone-mapped individually - however since console always has MB enabled per sample tone-mapping is switched off. (This is a driving game after all!)

A more exacting break-down of the shadow detail:

Shadows (CSM)

PC Low - 512x2048, 16bit, 2 spot shadows
PC Medium - 1024x4096, 16bit, 2 spot shadows
Console -1024x4096, 32bit, 4 spot shadows
PC High - 1408x5632, 32bit, 4 spot shadows
PC Ultra - 2048x8192, 32bit , 4 spot shadows + enables shadows in Environment Maps


Have those values changed from build on Gamersyde hands?
 

Ryne

Member
People are still surprised that a graphics heavy PC game at max settings do not run that well on their not so high end PCs?

I'm not, planned for a long time to get a new PC in order to max this game.

I just wanted to know what the spe4cs of the new PC should be in order to do so, so needed a few benchmarks to come out to see what I should be aiming for.
 
Guess I will need a new computer to maximize this game to get 60 FPS steady.

I know there is the incorrect obsession with maxing everything on the PC, however this is one game I set the goal of maximizing because it looks so damn great.

Gonna give it a shot with my i72600K OC'd and GTX 980 combo.

When you say max everything do you mean the AA settings too?

I have to be honest I don't really notice much difference when I move higher than what I have now (to be honest the only difference from my 960 settings was a decrease of object pop in the very far distance).

I play on a TV from 6ft away though so....
 
Some CPU benchmarks also from GameGPU:

http--www.gamegpu.ru-images-stories-Test_GPU-Simulator-Project_CARS_2015-test-pc_proz.jpg


The i5-2500K to i7-2600K difference is alarming. Haven't seen many games with such a difference between the two.

The weird thing is pCARS doesn't even utilize the extra logical cores from hyperthreading on the i7 all that much:

http--www.gamegpu.ru-images-stories-Test_GPU-Simulator-Project_CARS_2015-test-pc_intel.jpg


Here's some AMD core load tests as well:

http--www.gamegpu.ru-images-stories-Test_GPU-Simulator-Project_CARS_2015-test-pc_amd.jpg


Hopefully there is still some optimisaton to come from the day 1 patch and driver releases. As this is a Nvidia GameWorks title I'm expecting a game-ready driver on May 7 at least, and hopefully an AMD driver follows shortly after.

What makes me laugh is if this game had released a few months ago without the graphic optimizations that have come in recent builds, the performance would have been worse, especially for AMD users. There's a massive thread over at the WMD forums on AMD optimisation. There was a point where the 290X was getting performance equivalent of mid-range Nvidia cards. The benches show massive improvement there at least.

If the console versions can hold a steady frame rates in the Le Mans, Heavy Rain, Full AI grid stress test, then I'll seriously be impressed.
 

le-seb

Member
If the console versions can hold a steady frame rates in the Le Mans, Heavy Rain, Full AI grid stress test, then I'll seriously be impressed.
They'd probably need to offer a 30 fps lock option for this to happen.
Nobody can honestly expect SMS could improve the game's performance by a +50% margin since DF tested it.
 

Helznicht

Member
Are you serious?

i5 4690 and a GTX960 gives me locked 60 in every stress tress i've tried with all settings to low except car and track which are medium with MSAA and low FXAA. I've recently bought a 970 and that has allowed me to up everything across the board to high except grass (which has a known bug with nvidia which will be gone in D1P) and keep the same locked frame rate.

- PS4 uses EQAA(MSAA) and not FXAA.
- PS4 shadow detail is between the medium to high equivalent PC setting - more precisely PC medium shadow settings are 1024x4096 with 16 bit precision and PS4 is the same resolution but with 32 bit shadow precision.
- PS4 track detail/quality is equivalent to the high PC setting.
- PS4 car detail is equivalent to high with lod scaling between medium to high for the car shadows. .

The PS4 version is also supposed to run at 60 fps. If that's true and the above are true, the PS4 version is giving similar performance to an i5 4690 and a GTX 970. Any clarity on the subject would be welcome.
 

fresquito

Member
The PS4 version is also supposed to run at 60 fps. If that's true and the above are true, the PS4 version is giving similar performance to an i5 4690 and a GTX 970. Any clarity on the subject would be welcome.
You're building your own sand castle here.

The more demanding aspects from a graphics performance are grass, Env. Map, AA and Reflections. All of those I'd say that are either low or mid.

You can't really say the performance of the PS4 is comparable. It's really good for the hardware it's running on, yes, but it's not comparable.
 

ShamePain

Banned
The trend for this generation so far has been that a PC with a comparable GPU outclasses consoles when paired with a decent CPU(2500k and up). The consoles are simply bottlenecked by crappy CPUs and can hardly extract all of performance out of GPUs. PCars' case is probably that vendor drivers haven't been updated because it hasn't been released. Once all the components fall into place I expect PC to have a sizable framerate advantages if only because a decent CPU will handle 40+ AI entities significantly better.
 

Mascot

Member
Does anyone know FOR SURE whether you need to preorder the PS4 version only from Amazon or Zavvi in the UK to get the three preorder cars? I'm hearing mixed messages on various forums but nobody seems to know for sure. The statement on the pCARS website is open to interpretation.
 
Project CARS uses a light pre-pass renderer.... (there is no forward rendering aside from transparencies, which is standard for this sort of renderer)

There are actually two EQAA resolves, one before Motion-Blur (which is executed in HDR) and one before tone-mapping / bloom composition. On PC with motion-blur turned off, each sample of the MSAA target is tone-mapped individually - however since console always has MB enabled per sample tone-mapping is switched off. (This is a driving game after all!)

A more exacting break-down of the shadow detail:

Shadows (CSM)

PC Low - 512x2048, 16bit, 2 spot shadows
PC Medium - 1024x4096, 16bit, 2 spot shadows
Console -1024x4096, 32bit, 4 spot shadows
PC High - 1408x5632, 32bit, 4 spot shadows
PC Ultra - 2048x8192, 32bit , 4 spot shadows + enables shadows in Environment Maps
This is some awesome detail (especially about EQAA and the shadow draw). Damn. Thanks for this.

Obviously, you guys are probably pretty busy but if you have time to answer the following it would be great.
1. Do the transparencies receive only environmental lighting (sun and environment maps) as well as lighting information from point lights (such as head light beams)?
2. Similarly, do shadows cast into particles (dust for example) and or onto water surfaces/windshields?

Thanks for any answers.

BTW, If you guys ever want to implement a per-object blur... I definitely would not complain. :D
The PS4 version is also supposed to run at 60 fps. If that's true and the above are true, the PS4 version is giving similar performance to an i5 4690 and a GTX 970. Any clarity on the subject would be welcome.
That would be impossible given the two are equivalent in performance at similar settings. It would go against the vast volume of evidence from every other multiplatform game.
 

fresquito

Member
This is some awesome detail (especially about EQAA and the shadow draw). Damn. Thanks for this.

Obviously, you guys are probably pretty busy but if you have time to answer the following it would be great.
1. Do the transparencies receive only environmental lighting (sun and environment maps) as well as lighting information from point lights (such as head light beams)?
2. Similarly, do shadows cast into particles (dust for example) and or onto water surfaces/windshields?

Thanks for any answers.

BTW, If you guys ever want to implement a per-object blur... I definitely would not complain. :D

That would be impossible given the two are equivalent in performance at similar settings. It would go against the vast volume of evidence from every other multiplatform game.
1. Headlights affect particles.
2. Yes for particles, can't say about windshields and what not.

I'm talking about the PC version. I don't know about consoles, probably is the same, but can't garantee.
 

SMSRenderTeam

Neo Member
1. Headlights affect particles.
2. Yes for particles, can't say about windshields and what not.

I'm talking about the PC version. I don't know about consoles, probably is the same, but can't garantee.

Consoles are the same. Windshields and water receive shadows.

The particles receiving shadows can sometimes lead to them looking a little strange on rare occasions, since smoke/dust are rendered using a standard quad based approach, so detailed shadows on them can highlight their quad based nature! On the plus side though it does mean that particles integrate nicely into the scene in all lighting conditions.
 

Hedrush

Member
Hmm, that's odd. What region are you guys in? I don't have a PS4 myself but in the UK I'm able to purchase it from the online PS Store. Not sure whether this would preload on the console though.
Im in the UK. My pre-order has a countdown timer where the download button should be.

Edited to add that the countdown timer expires on Wednesday next week.
 

Azzawon

Member
You can preorder (from the EU stores, at least?), but games don't preload earlier than 2-3 days before release on PS4.

Im in the UK. My pre-order has a countdown timer where the download button should be.

Edited to add that the countdown timer expires on Wednesday next week.

Ahhh I see! Slightly odd practice when XB1 preload is already up, can't comment on the Steam one because I already have it as a backer! But I can imagine that could cause some problems for people with slow internet if there's a hefty day one update.
 
Top Bottom