[computerbase] Project CARS benchmarks

[Erenhardt]

Umm yeah that AF difference was looked into and it was nvidia that caught instead.

https://forum.beyond3d.com/threads/...af-filtering-broken.49106/page-9#post-1468349

Keep reading and there wasn't exactly any hardware issue.

As for IQ, I've often read that it's the opposite, AMD cards have more vibrant IQ by default and is remarked on those who change cards. It is a separate thing from the HDMI full range RGB issue.

http://pokde.net/pokde-blog/nvidia-gtx-titan-vs-amd-r9-290x-image-quality-comparison/

^^ This

 

[SirPauly]

Umm yeah that AF difference was looked into and it was nvidia that caught instead.

https://forum.beyond3d.com/threads/...af-filtering-broken.49106/page-9#post-1468349

Keep reading and there wasn't exactly any hardware issue.

As for IQ, I've often read that it's the opposite, AMD cards have more vibrant IQ by default and is remarked on those who change cards. It is a separate thing from the HDMI full range RGB issue.

http://pokde.net/pokde-blog/nvidia-gtx-titan-vs-amd-r9-290x-image-quality-comparison/

AMD had an issue with texture mapping units with texture filtering and high frequency, higher complex textures with the 5xx series.

 

[master_shake_]

http://www.techspot.com/review/1000-project-cars-benchmarks/page6.html

techspot review of Project CARS. Finally a website has the guts to calls out Nvidia on the performance of Kepler in recent titles and says Nvidia has forgotten Kepler owners. At 1080p with rain effects GTX 960 beats GTX 780. Ouch !!!

AMD is facing a different problem. They have slacked off on the game ready drivers for latest titles which were more regular last year and their last WHQL driver is more than 5 months old. This is bad. I hope AMD picks the slack up and has game ready drivers for titles like Witcher 3 and Batman Arkham Knight with good CF performance scaling and smoothness.

no surprise.

look at techpowerups performance summaries.

kepler owners are left out in the cold when it comes to optimizations.

hell even the 7970ghz is damn near a 780 in performance.

tell me that was supposed to happen.

http://www.techpowerup.com/reviews/Gigabyte/GTX_960_OC/27.html

 

[Azix]

this is interesting

http://www.gamedev.net/topic/666419-what-are-your-opinions-on-dx12vulkanmantle/#entry5215019

looks like things should be much better when it comes to games working right. I have a feeling nvidia doesn't like it

 

[Vaporizer]

no surprise.

look at techpowerups performance summaries.

kepler owners are left out in the cold when it comes to optimizations.

hell even the 7970ghz is damn near a 780 in performance.

tell me that was supposed to happen.

http://www.techpowerup.com/reviews/Gigabyte/GTX_960_OC/27.html

Wait for the bombing of Maxwell when Pascal releases...

 

[jj109]

Since all of this pCARS controversy was based on cageymaru's "quotes" from the developer, have anyone of you long-winded folks actually checked to see if the guy wasn't just lying?

https://www.reddit.com/r/pcgaming/c...eworks_project_cars_and_why_we_should/crc71vw

It would be pretty funny if this was a 10/10 troll job.

 

[railven]

Since all of this pCARS controversy was based on cageymaru's "quotes" from the developer, have anyone of you long-winded folks actually checked to see if the guy wasn't just lying?

https://www.reddit.com/r/pcgaming/c...eworks_project_cars_and_why_we_should/crc71vw

It would be pretty funny if this was a 10/10 troll job.

Pro NV guy used his posts, snipping parts out, to support his claim. I think on this forum, at least, I was the one that read the two posts that were quoted and made note of the information cut out.

I personally can't verify what is said in 3rd person as I don't have access to those forums. Until other information from those forums is brought out, what can I (or anyone else) really do? [EDIT: ANd that Reddit post was made 21 hrs ago, so that information is relatively new.] I also made sure to state based on those posts, PhysX wasn't running on GPU, it was just alluded to. I asked if anyone could rune some benches. And some people thankfully did. Which is how a community should work. Not witch hunts.

I'd hate to think it was a troll, but hey, do people really go on the internet and lie?

 

[psolord]

Has Physx been debunked as the performance culprit yet?

I have a 970 and a 570 and was thinking to maybe do two benchmarks, with just the 970 and then with the 570 installed and set as the physx card.

I wonder if that would shed some light in all this.

 

[railven]

Has Physx been debunked as the performance culprit yet?

I have a 970 and a 570 and was thinking to maybe do two benchmarks, with just the 970 and then with the 570 installed and set as the physx card.

I wonder if that would shed some light in all this.

That horse is buried. Some good sleuthing from users here put it to bed, at least for me. AMD got some work to do on their driver CPU optimization overheard, whatever you wanna call it.

 

[psolord]

Ok cool.

Indeed I did a new test on my 7950@1Ghz stock clock, using the default settings, but changed MSAA with SMAA and saw that the the gpu load was around 60% and the framerate was around 40fps or less.

The game could easily hit 60fps with these settings, if it was not cpu limited.

This was on my i7860@4Ghz. I am this close to running the same benchmark on my 2500k@Ghz just to see how it fairs.

Since Witcher 3 is coming and I will do some hardware moving around for testing, I will try to remember to test this as well.

 

[Erenhardt]

Has Physx been debunked as the performance culprit yet?

I have a 970 and a 570 and was thinking to maybe do two benchmarks, with just the 970 and then with the 570 installed and set as the physx card.

I wonder if that would shed some light in all this.

Yep. Developer saying it that it is calculating physx on CPU with amd card setup, implying that with nv card, it is accelerated by nv gpu, is not a proof.


Some have feeling in the guts, that the problem is only amd driver overhead that is killing 4,5GHz Haswell i7 to the point it only renders half of the frames compared to nv setup. That in itself is a proof that can't be ignored.

 

[Goatsecks]

That horse is buried. Some good sleuthing from users here put it to bed, at least for me. AMD got some work to do on their driver CPU optimization overheard, whatever you wanna call it.

Feels like a lot of the users on this forum are ignoring most of the evidence. I suspect this thread will continue on in the same vein.

I think the mods have done a piss poor job over the last few week by allowing the forum to get worked up into such a furore over nothing but speculation (in several threads, not just this one). Sorry mods.

 

[96Firebird]

I think the mods have done a piss poor job over the last few week by allowing the forum to get worked up into such a furore over nothing but speculation (in several threads, not just this one). Sorry mods.

What are the mods to do? Everyone has an opinion, no matter how wild it may be. They are free to post that opinion, just as I am able to laugh at them.

 

[Flapdrol1337]

Yep. Developer saying it that it is calculating physx on CPU with amd card setup, implying that with nv card, it is accelerated by nv gpu, is not a proof.

Only it wasn't a developer, but some guy saying he was quoting a developer.

 

[psolord]

Ok I will do and post 970+570 and 7950/2500/860 tests, just out of curiosity and for whoever may be concerned.

 

[Azix]

Only it wasn't a developer, but some guy saying he was quoting a developer.

Think I did see a developer on the steam community forum say something similar.

The idea that its driver overhead doesn't seem right. If its driver overhead trying to handle the code in this particular game, maybe. If its driver overhead in general, AMDs drivers can't cripple high end processors and don't in any other games. At the very minimum there is something in the game that is part of the problem

 

[Eymar]

I find that "Detailed Grass" on ultra and "Particle Density" on ultra effect graphic smoothness (ie. framerates maybe high (>60FPS), but apparent visual stuttering and hitches are seen even with G-SYNC so likely game engine problem and not graphic engine). Setting those two settings on high (everything else highest settings, MSAA +FXAA med) removed VSYNC type stuttering for me. I'm running 5930k@4.4, Titan-X SLI + 680 GTX as dedicated Physx, surround 1440P GSync and getting 80-90FPS in most tracks. Setting weather to Rain drops min FPS to around 60FPS.

 

[96Firebird]

Don't believe everything you read on the internet.

http://www.dsogaming.com/news/sligh...nt-about-project-cars-on-amds-graphics-cards/

 

[xthetenth]

Oh good, now that we have independent confirmation from the developers that the developers aren't.... waiiiiit a second.

 

[96Firebird]

I'm not sure anything can be done to convince those who do not believe words straight from the developer...

 

[AnandThenMan]

I'm not sure anything can be done to convince those who do not believe words straight from the developer...

Well the dev says Nvidia is not a sponsor yet there are Nvidia logos plastered all over the game. So what are people supposed to think? Are there any AMD logos in the game? Serious question I have not played it.

 

[SirPauly]

Don't believe everything you read on the internet.

http://www.dsogaming.com/news/sligh...nt-about-project-cars-on-amds-graphics-cards/

Its that poster at hardocp that started the train wreck.

 

[xthetenth]

I'm not sure anything can be done to convince those who do not believe words straight from the developer...

Words from an actually independent source that can confirm/deny things with evidence?

 

[Erenhardt]

I like how they didn't say We don't use gpu accelerated physx, but rather physx on cpu with amd gpu takes only additional 10% of all physics engine computation time.

I call this a BS:

The MADNESS engine runs PhysX at only 50Hz

The MADNESS engine uses PhysX for collision detection and dynamic objects, which is a small part of the overall physics systems

• The physics threading does not interact with the rendering, it is a push system sending updated positional information to the render bridge at 600Hz

Why would it send the same data over and over again at steady frequency that is not tied to rendering frequency?
Also, they render objects in different places than the collision detection is calculated when fps is not exactly 50. That would end in a bad clipping and getting stuck in things.

Also, dynamic objects? Do they mean particles? Like in NFS shift that worked abysmal?

Project CARS does not use NVIDIA specific particle technology - the system we use is a modified version of the same technology we used on the Need for Speed : Shift and Shift Unleashed games, and was entirely developed in-house. The reason the performance drops when there are a lot of particles on screen is simply because processing a large number of particles is very expensive.

A few leafs seems not a large number, or is it? How many leafs does a leaf have?

The MADNESS engine runs PhysX at only 50Hz

So if lets say I have 100 FPS in this game, but physx is calculated at 50 times per second. How is my frame rendered. Does every other frame have no physx rendered - that would probably cause flickering.

Lets say I have Radeon card (doesn't really matter which as there is little to none scaling) and run this game at 25 FPS. My physx rendering is still at 50hz. Which means my every other frame is blank with only physx effects - I didn't noticed that. Maybe I should take a printscreen and have a closer look.

Dig deeply and greedily.

 

[monstercameron]

I like how they didn't say We don't use gpu accelerated physx, but rather physx on cpu with amd gpu takes only additional 10% of all physics engine computation time.

I call this a BS:



Why would it send the same data over and over again at steady frequency that is not tied to rendering frequency?
Also, they render objects in different places than the collision detection is calculated when fps is not exactly 50. That would end in a bad clipping and getting stuck in things.

Also, dynamic objects? Do they mean particles? Like in NFS shift that worked abysmal?

A few leafs seems not a large number, or is it? How many leafs does a leaf have?


So if lets say I have 100 FPS in this game, but physx is calculated at 50 times per second. How is my frame rendered. Does every other frame have no physx rendered - that would probably cause flickering.

Lets say I have Radeon card (doesn't really matter which as there is little to none scaling) and run this game at 25 FPS. My physx rendering is still at 50hz. Which means my every other frame is blank with only physx effects - I didn't noticed that. Maybe I should take a printscreen and have a closer look.

Dig deeply and greedily.

I am sure it is more complex than you make it seem.