Project Cars: CPU benchmarks (well threaded!)

[escrow4]

http://pclab.pl/art63572-29.html

AMD falls back badly once again (in rain at least, what is in that rain?), Haswell gets a few extra FPS over old SNB at stock and hexa cores excel . . . .

 

[cbn]

I don't see Hexcore doing that well in these benchmarks (though the second PC lab results are better than the first one).

The G3258 appears to do well though.

 

[Carfax83]

Here's another one from Computerbase.de:

As you can see, the game doesn't seem to like HT very much, but it does use up to six cores otherwise the 3970x would not be beating the 4770K..

Also, this game supports DX11 multithreading on NVidia GPUs, which partially explains why NVidia has such a massive lead in performance over AMD.. Make no mistake, an 86% lead is absolutely massive!

 

[TurtleCrusher]

Is "well threaded" a joke?

 

[mikk]

Also, this game supports DX11 multithreading on NVidia GPUs, which partially explains why NVidia has such a massive lead in performance over AMD.. Make no mistake, an 86% lead is absolutely massive!

On default the game is running in dx11st mode, means all tests are made with it. You have to run the game with -dx11mt in order to use DX11 MT. The gap is even bigger towards Nvidia with it. There is another test: http://www.pcgameshardware.de/Project-CARS-PC-238576/Specials/Benchmark-Test-1158026/

 

[escrow4]

Is "well threaded" a joke?

Hexa-cores are on top on all benchmarks - what are you seeing? The engine does use multiple cores and threads well.

 

[R0H1T]

Is "well threaded" a joke?

Just what I was thinking

Going from G3258 to 5820K doesn't even give you twice the performance, 3x the cores+HT & better caches wasted spectacularly D:

 

[Gikaseixas]

OP your well threaded definition is very flawed. Stock 4770k is faster than a stock 5920k

 

[Carfax83]

On default the game is running in dx11st mode, means all tests are made with it. You have to run the game with -dx11mt in order to use DX11 MT. The gap is even bigger towards Nvidia with it. There is another test: http://www.pcgameshardware.de/Project-CARS-PC-238576/Specials/Benchmark-Test-1158026/

Interesting. I hope we get to see some benchmarks with enabled vs with it turned off..

 

[TurtleCrusher]

What I find interesting is that the FX-4300 stock is faster than a 4GHZ Phenom 2 X4. With supposedly lower IPC, a module penalty for higher threaded workloads and less cache, how is it beating a 4GHZ phenom 2 quad?

Edit: also frame rate variance. Many of those scores have an average of only 3-4fps above the minimums. This means the FPS is extremely constant throughout the game, meaning even at lower framerates it'll feel consistently smooth.

Either that or how they benched it wasn't truly that demanding and indicative of real world gaming.

 

[mikk]

Interesting. I hope we get to see some benchmarks with enabled vs with it turned off..

It depends on the settings and track you choose. It can be up to 25% faster in a CPU bound test environment with lots of AI cars on Nvidia. Against 0 AI cars it isn't neccessarily faster. Just one example

Azure Circuit race start 720p noAA

DX11ST 122 fps (0 AI) /59 fps (35 AI)
DX11MT 123 fps (0 AI) /72 fps (35 AI)

(i5-4670, GTX 970 350.12, Win 7 x64)

 

[Flapdrol1337]

Hexa-cores are on top on all benchmarks - what are you seeing? The engine does use multiple cores and threads well.

But the overclocked pentium beats the overclocked 6350, very strange to see the 6350 gain so little from overclocking.

 

[BigDaveX]

What I find interesting is that the FX-4300 stock is faster than a 4GHZ Phenom 2 X4. With supposedly lower IPC, a module penalty for higher threaded workloads and less cache, how is it beating a 4GHZ phenom 2 quad?

The FX includes support for a bunch of extra instruction sets, and I suspect that enough time has passed that applications are making widespread use of them, putting the FX (and Piledriver/Steamroller APUs) above Phenom II in certain situations.

 

[beginner99]

Is "well threaded" a joke?

Exactly my thought too. And i3 beating FX eight core... not possible if it was well threaded. The game clearly needs at least 1 fast core or else it's limited.

 

[inf64]

The FX includes support for a bunch of extra instruction sets, and I suspect that enough time has passed that applications are making widespread use of them, putting the FX (and Piledriver/Steamroller APUs) above Phenom II in certain situations.

Piledriver core has roughly 10% lower IPC than 45nm Stars core. In MT workloads this drops by another 10-15% depending on the workload due to module sharing penalty (mostly because of shared FP/SIMD unit).

SR core mitigated the penalty or at least lowered it and XV core brings additional ~5% IPC improvement. We can say that XV core should be on par with 45nm Stars core. Any additional speedup that comes from newer ISA extensions is just a bonus.

As far as this test goes, it obviously is not well threaded.

 

[FlanK3r]

What I find interesting is that the FX-4300 stock is faster than a 4GHZ Phenom 2 X4. With supposedly lower IPC, a module penalty for higher threaded workloads and less cache, how is it beating a 4GHZ phenom 2 quad?

Edit: also frame rate variance. Many of those scores have an average of only 3-4fps above the minimums. This means the FPS is extremely constant throughout the game, meaning even at lower framerates it'll feel consistently smooth.

Either that or how they benched it wasn't truly that demanding and indicative of real world gaming.

Because ipc of fx is not overall lower than Phenoms...its slower only in some x86 code, but not in all variations. New games are ussually better at Fx cpus than at Phenoms 2.

 

[Erenhardt]

Here's another one from Computerbase.de:

Someone is trying hard to make us believe that amd drivers eat overclocked i7 CPU to the point that fx and nv is faster.

Your really think the driver overhead is able to make a dent on 4.4GHz haswell?

To me it seems that there is additional load on the CPU on a system with amd card. What if... bare with me here... what if there is physx engine running on CPU if there is amd card in system, and it can be of-loaded to nv gpu.
http://physxinfo.com/news/11822/sta...clude-gpu-accelerated-physx-and-apex-effects/

Since Star Citizen is based on CryEngine 3, we can expect a “PhysX SDK-less” integration, where APEX modules will work on top of CryEngine own physics system. Finally, some APEX features, such as simulated character clothing, may be available for AMD users as well, running on CPU.


Next one – Project CARS – a racing simulation game in development by Slightly Mad Studios.


The physics in the game is already utilizing PhysX SDK 3.2, and the extra efffects will consist of dynamic particles and APEX Turbulence based smoke.

Stay tuned, as we are gathering info about another PhysX technology, announced today – FLEX, an unified GPU accelerated physics solver.

 

[ShintaiDK]

There is no GPU physX in project cars. Already covered in the Project cars thread in VC&G section.

 

[Erenhardt]

There is no GPU physX in project cars. Already covered in the Project cars thread in VC&G section.

straight from horses mounth:
http://www.geforce.com/whats-new/articles/gamescom-2014-project-cars

As players got to grips with the controls and high-powered sports cars they inevitably spun, creating a cloud of NVIDIA GPU-accelerated particles that reacted realistically to the motion of the car and other forces. These effects, and others, create the most lifelike racing game seen to date, which at times can be mistaken for genuine TV footage.

So from now one I would love to see you put as much effort into spreading that information, as you did spreading the misinformation.

 

[ShintaiDK]

GPU PhysX is not implemented. I know you already desired to ignore all tests.

You do know your link is from august 2014?

Just wait for AMD to fix their driver. Roy already said they was working on it.

 

[Erenhardt]

GPU PhysX is not implemented. I know you already desired to ignore all tests.

You do know your link is from august 2014?

You have anything to support your claims?
Do you expect us to believe they backed off from GPU physx, that was showcased and is advertised on nv site?

Contrary to what seemed to happened? NV demanded to double up on physx effects after amd improved their performance?

For about a month after I personally requested a driver from AMD, there was new driver and a partial fix to the problem. Then Nvidia requested that a ton of more PhysX effects be added, GameWorks was updated, and that was that... But maybe AMD can pull a rabbit out of the hat on this one too. I certainly hope so.

Believe what you want. The facts say it clear. Project cars is using GPU accelerated physx. Which explains huge CPU usage with amd cards. Comparable to what borderlands did with CPU physx.

 

[ShintaiDK]

You have anything to support your claims?
Do you expect us to believe they backed off from GPU physx, that was showcased and is advertised on nv site?

Contrary to what seemed to happened? NV demanded to double up on physx effects after amd improved their performance?



Believe what you want. The facts say it clear. Project cars is using GPU accelerated physx. Which explains huge CPU usage with amd cards. Comparable to what borderlands did with CPU physx.

You could just read the posts from people testing it in the Project Cars thread in VC&G. Since you are very active there it shouldnt be a problem for you. Unless you already desided the outcome.

What are you going to do if a new AMD driver fixes the performance? AMD is working on it. So wait and see what they come up with. Because you are placing yourself in a very thight corner. And it will be awfully arkward for you if AMD delivers what Roy says.

https://twitter.com/amd_roy/status/596361439016685569

 

[Erenhardt]

You could just read the posts from people testing it in the Project Cars thread in VC&G. Since you are very active there it shouldnt be a problem for you. Unless you already desided the outcome.

What are you going to do if a new AMD driver fixes the performance? AMD is working on it. So wait and see what they come up with. Because you are placing yourself in a very thight corner.

I will believe what nvidia has to say on that matter, rather than forums posters.
http://www.geforce.com/whats-new/articles/gamescom-2014-project-cars

s players got to grips with the controls and high-powered sports cars they inevitably spun, creating a cloud of NVIDIA GPU-accelerated particles that reacted realistically to the motion of the car and other forces. These effects, and others, create the most lifelike racing game seen to date, which at times can be mistaken for genuine TV footage.

 

[ShintaiDK]

It may have used GPU PhysX 10 months ago in an early stage. It may also come again in the future. But as of now it doesnt use it.

 

[Erenhardt]

It may have used GPU PhysX 10 months ago in an early stage. It may also come again in the future. But as of now it doesnt use it.

I would rather see you say: I think it doesn't use it. Or you have a source from nvidia saying project cars has no gpu physx?