Project C.A.R.S benchmark update (Feb 2014)

[Carfax83]

These are the latest Project Cars benchmarks, from PCgameshardware.de:

Source

GTX 770 is 50% faster than the 280x. GTX 780 is 49% faster than the R9 290x..

That's DX11 MT in action.. The Radeons are terribly CPU bound, unlike the Geforce cards.. It's only when downsampling is activated that the Radeons start catching up.

 

[ShintaiDK]

AMD lack of a proper DX driver shows.

Instead of Mantle they should just have focused on that.

 

[mikk]

Direct link doesn't work.


DX11MT wasn't enabled most likely because on default it starts with DX11. Big driver overhead is causing the bad performance on AMD.

 

[Carfax83]

AMD lack of a proper DX driver shows.

Instead of Mantle they should just have focused on that.

Actually, AMD making Mantle was good for the industry overall I think, because it got everyone focused on, and talking about Direct3D's high CPU overhead..

Eventually, Microsoft would have done these things, but I don't think they were in any rush since there was no public pressure being applied..

 

[dguy6789]

It seems odd that people would blame AMD and not the game developers when "random game" runs well on Nvidia and poorly on AMD but other games do not. Careful, your bias is showing.

 

[Carfax83]

Direct link doesn't work.

PCgameshardware.de doesn't allow deeplinking. The source link still works though.

DX11MT wasn't enabled most likely because on default it starts with DX11. Big driver overhead is causing the bad performance on AMD.

I'm assuming you can read German. Does it state they didn't enable DX11 MT?

Usually DX11 MT is auto enabled when the game detects the proper hardware and driver combination. Only when it's buggy do developers give you the option of turning it on or off via console or ini setting.

I'm sure it was on though. Just look at the benchmarks. Yes, NVidia's D3D drivers are more efficient than AMD's, but not enough to result in a 50% performance advantage.

50% is huge, thats like a generational gap.. And we know that DX11 MT has given such large increases in previous titles like Civ 5..

 

[Carfax83]

It seems odd that people would blame AMD and not the game developers when "random game" runs well on Nvidia and poorly on AMD but other games do not. Careful, your bias is showing.

There's a specific reason why the game runs poorly on AMD compared to NVidia. It's because the game supports DX11 MT, and only NVidia supports that DX11 feature.

AMD does not.. I'm sure increased driver optimization and game code will make the game run faster on AMD at launch, but NVidia is likely always going to have a large lead in this title courtesy of DX11 MT.

 

[SPBHM]

pre alpha build,
and as far as I know they are working closely with NV, not AMD.

no surprise, also look at the 2c vs 2c + ht,

this pre alpha build game is probably very poorly optimized for most hardware

 

[Carfax83]

pre alpha build,
and as far as I know they are working closely with NV, not AMD.

no surprise, also look at the 2c vs 2c + ht,

this pre alpha build game is probably very poorly optimized for most hardware

Could that result in a 50% performance advantage for NVidia though? I'm not so sure.

This game has been "pre-alpha" for years now, and people have been playing and testing it during that time. NVidia and AMD should have been optimizing their drivers for the game..

Supposedly it releases at the end of this year..

 

[SPBHM]

Could that result in a 50% performance advantage for NVidia though? I'm not so sure.

This game has been "pre-alpha" for years now, and people have been playing and testing it during that time. NVidia and AMD should have been optimizing their drivers for the game..

Supposedly it releases at the end of this year..

it was supposedly going to be released years ago, and also for PS3...

I know this game since the beginning, it started as some strange crowd funded game from the NFS Shift developers, anyway, pre alpha build is pre alpha build, I wouldn't read to much based on these results, it's pretty clear they are working closely with Nv as I said.

 

[Skurge]

I have said this before.

Radeons perform poorly with MSAA in this game. That is most likely an optimisation issue. If the Radeons were all CPU limited a 290X wouldn't be an faster than a 280X but it is. The reason the Radeons catch up when you replace MSAA with SSAA is not because the bottleneck is moved from the CPU to the GPU. It is because SSAA doesn't bog down the Radeons as much. Also, the Radeon just generelly perform poorly in this game as they have worked closely with Nvidia. It is also Pre-Alpha, but make of that what you will.

 

[VulgarDisplay]

You mean a game that's packed with nvidia exlusive tech in pre alpha that amd has not had a chance to work on their drivers for performs better on nvidia cards?

I can't believe. The benchmarks are obviously wrong.

 

[3DVagabond]

Could that result in a 50% performance advantage for NVidia though? I'm not so sure.

This game has been "pre-alpha" for years now, and people have been playing and testing it during that time. NVidia and AMD should have been optimizing their drivers for the game..

Supposedly it releases at the end of this year..

Of course it could make a 50% difference. There have been other games that nVidia has had huge advantages in until they were released and AMD has had a chance to optimize for them.

 

[Grooveriding]

Another fail thread is born.

 

[Hitman928]

This also isn't a new benchmark,

Note: This article had the most page views this week. For this reason we publish it again today

More fun quotes:

The tested Steam version (Pre-Alpha Build 635) by Project Cars is still a long way from a finished product. Therefore, this article is only available as an inventory, or to understand a development documentation, not as a test.

enjoy with caution

Many other things also do not yet work correctly: The AI turns on some routes to highly stupid, cars fall into curves, lift off suddenly or shatter into pieces.

It is noteworthy that increasing resolution shrinks the advantage of the Geforce cards and that multi sampling on AMD GPUs runs worse than on the Nvidia cards. In fact downsampling on AMD graphics cards needs less power than multi-sampling, a clear indication of a need of optimization.

I didn't see anything about enabling DC multithreading.

 

[wand3r3r]

Looks like another basically trolling attempt to find something good about mt or nv.

I can't read German but I like these charts therefore I'll post about the one advantage I think it might be showing.

Another failed thread.

 

[3DVagabond]

Looks like another basically trolling attempt to find something good about mt or nv.

I can't read German but I like these charts therefore I'll post about the one advantage I think it might be showing.

Another failed thread.

What? You don't think that the benchmark of this unreleased game proves that nVidia dominates AMD? I guess some people are just impossible to impress. /sarc

 

[ICDP]

PCARS is in pre-alpha and has been since it was announced years ago. Oh no, AMD drivers have not been optimised, for a very early pre-alpha version of a game that is not due until Q4 2014 (or 2015). Well I'm convinced, this is total proof that Nvidia do DX11 better than AMD. All those benchmarks of actual released games can be discounted completely.

Another thread that proves nothing but the inability of Carfax to be objective and honest.

 

[cytg111]

Wasnt this locked allready?

 

[Carfax83]

I have said this before.

Radeons perform poorly with MSAA in this game. That is most likely an optimisation issue. If the Radeons were all CPU limited a 290X wouldn't be an faster than a 280X but it is. The reason the Radeons catch up when you replace MSAA with SSAA is not because the bottleneck is moved from the CPU to the GPU. It is because SSAA doesn't bog down the Radeons as much. Also, the Radeon just generelly perform poorly in this game as they have worked closely with Nvidia. It is also Pre-Alpha, but make of that what you will.

Well I never knew about the MSAA issue. This game has been getting benchmarked for probably almost 2 years now on various websites, so I thought it would be in better shape....but I guess not. It supposedly ships this year too..

OK fine, I'm dropping this issue....for now. The only reason it interested me because there are so many people saying that DX11 MT is a failed technology.

 

[3DVagabond]

Well I never knew about the MSAA issue. This game has been getting benchmarked for probably almost 2 years now on various websites, so I thought it would be in better shape....but I guess not. It supposedly ships this year too..

OK fine, I'm dropping this issue....for now. The only reason it interested me because there are so many people saying that DX11 MT is a failed technology.

Microsoft is working with the hardware manufacturers for close to metal access in DX. Seems like even they don't think MT is the answer that you do.

 

[Skurge]

Well I never knew about the MSAA issue. This game has been getting benchmarked for probably almost 2 years now on various websites, so I thought it would be in better shape....but I guess not. It supposedly ships this year too..

OK fine, I'm dropping this issue....for now. The only reason it interested me because there are so many people saying that DX11 MT is a failed technology.

Like I said, I was one of the poeple who crowdfunded the game. I can do whatever tests you like. I have access to a few different graphics cards.

 

[mikk]

I'm assuming you can read German. Does it state they didn't enable DX11 MT?

I can and nowhere is said that they enabled DX11MT. DX11MT is work in progress and therefore not enabled by default. The equivalent of DX11MT works pretty well on console they are saying whilst on PC-Directx there are still some issues.

Furthermore AMD has a big driver overhead not matter if DX11MT or DX11. A render coder who analyzed it said: The main problem I can see is that on ATI it requires substantially more CPU power in DX11 to render any given scene. This is all down to shader constant setting - the cost of setting vertex and pixel shader constants is very much higher than on NVIDIA.
So at the moment to reach equivalent NVIDIA performance requires around 25% more power"



I think PS4/Xbox One shouldn't suffer from the overhead because of the low level nature in a console.

Usually DX11 MT is auto enabled when the game detects the proper hardware and driver combination.

Not in pcars. Once it works without issues it probably will replace DX11 as default though. The bad AMD performance you are seeing currently has nothing to do with DX11MT.

 

[parvadomus]

What? You don't think that the benchmark of this unreleased game proves that nVidia dominates AMD? I guess some people are just impossible to impress. /sarc

Yep, nVidia dominates AMD in pre alpha-build games, sponsored and co-developed by them, where AMD hasn't do a single bit of optimization.

PS: oh wait.. just realized it was cause of NV MT drivers, damn they are good!

 

[Bubbleawsome]

This thread has gone well.