Jad Chaar
Posts: 6,481 +976
Oh yeah I understand that but still if the hardware is listed at capable it should at least run smoothly... otherwise you are wasting money on a game you think can be run but in reality is runs at like 10fps.
Project CARS Benchmarked, Tested
- Thread starter Steve
- Start date
SuperVeloce
Posts: 133 +34
I really don't know from where are you coming from with this, but it is not true. AMD drivers have almost nonexistent multithreading and bigger overhead. This you can see in all dx11-mantle benchmarks, where amd dx11 performance just crash and burn. But in this game it's evident anyways that those better cpus have power to spare (for running 290x). PhysX is off in those benchmarks
dividebyzero
Posts: 4,840 +1,271
Two points:
1. Carsten, and PCGH are pretty well respected. His content on PCGH, Beyond3D and elsewhere is top notch and provides valuable insight for many, myself included
2. You consider the 290X putting up lower framerates than the GTX 760 as "good" ?
The 290X only pulls ahead of the GTX 760 and 770 when downsampling and/or higher resolutions overwhelm the 2GB framebuffer of the Kepler cards.
20 goto 10
Posts: 6 +2
Good choice for benchmarking Steve 
Although it must be said this is yet another console title, so (shocking AMD performance aside), most aspects of performance scaling are quite easy to guess. I'm always keen to see if PC specific games differ from these patterns.
Number of opponents is a very important variable in sim-racing performance - particularly how many are on screen at any one point. Did you start at the back of the pack or the front?
Although there is validity in isolating CPU and GPU performance, it doesn't tell us perhaps the most significant information - where the crossover between CPU and GPU bottleneck is. Perhaps a couple of 'real world' situations to measure the middle ground would give us a strong clue - I,e, 960/970/980 1080/wxga/2k 4130, 4440, 4690, 4790k. Users could infer quite a lot from this. Although that is 36 tests, LOL.
Finally, I would love to know if there is a particular setting that can kill performance. On Assetto Corsa word is that post-processing cripples performance and offers hardly any visual advantage. The figures here are generally disappointing - 60fps is essential for sim-racing. Yet I've not heard many complaints from other sources regarding performance. I'm left wondering if a small adjustment in MFAA or AF could render the game playable for most users.
Cheers~
Although it must be said this is yet another console title, so (shocking AMD performance aside), most aspects of performance scaling are quite easy to guess. I'm always keen to see if PC specific games differ from these patterns.
Number of opponents is a very important variable in sim-racing performance - particularly how many are on screen at any one point. Did you start at the back of the pack or the front?
Although there is validity in isolating CPU and GPU performance, it doesn't tell us perhaps the most significant information - where the crossover between CPU and GPU bottleneck is. Perhaps a couple of 'real world' situations to measure the middle ground would give us a strong clue - I,e, 960/970/980 1080/wxga/2k 4130, 4440, 4690, 4790k. Users could infer quite a lot from this. Although that is 36 tests, LOL.
Finally, I would love to know if there is a particular setting that can kill performance. On Assetto Corsa word is that post-processing cripples performance and offers hardly any visual advantage. The figures here are generally disappointing - 60fps is essential for sim-racing. Yet I've not heard many complaints from other sources regarding performance. I'm left wondering if a small adjustment in MFAA or AF could render the game playable for most users.
Cheers~
Last edited:
The back of the grid and the results are based on an average of three runs.
Evernessince
Posts: 5,469 +6,158
Two points:
1. Carsten, and PCGH are pretty well respected. His content on PCGH, Beyond3D and elsewhere is top notch and provides valuable insight for many, myself included
2. You consider the 290X putting up lower framerates than the GTX 760 as "good" ?
The 290X only pulls ahead of the GTX 760 and 770 when downsampling and/or higher resolutions overwhelm the 2GB framebuffer of the Kepler cards.
Thanks for the tip on which websites are safe to go to. I tend to find a few websites that are good and stick with them.
If you click the buttons above their chart, you get different settings for the benchmark. What's strange is the 290x does very well at 2560 while it does so bad at 1920.
It’s not that strange, we found the same thing. Due to the fact that we believe the issue has to do with Nvidia PhysX the AMD GPU’s aren’t being limited by resolution which is why the 1080p performance is so bad. Update: I have been told by a contact at SMS that "the issue is unrelated to PhysX", what it is related to hasn't been made clear yet.
Or maybe I should say the serious performance bottleneck comes into play before the 1080p resolution.
Our 4K results make a better point of showing this…
1080p = 49fps
2160p = 35fps
That is 29% less performance at 4K despite there being 4 times (400%) as many pixels to render.
TechSpot isn’t blaming anyone, we simply stated the facts as they have come to us. I have been speaking with multiple sources at AMD and they all tell us “AMD†is working on a driver fix..
Roy Taylor said “Right now the performance is an anomaly, since clearly Radeon should perform dramatically better than it does. There is no ‘smoking barrel’ here. No conspiracy, we have been talking and working together for some time and we just need to now dig in and identify the issue. Sometimes these things happen, unfortunate when they do but between us we are confident we will fix the problem.â€
I don’t disagree but comparing indie games to a AAA title doesn’t really make sense as indies don’t push systems to the limit and take advantage of cutting edge graphics technology. When building a game that sets a new benchmark for visuals things like this can happen.
AMD are probably copping the blame because they didn’t follow the development closely enough.
You saw right through our cunning plan to fool you all.
Another note, in racing sims 30fps is quite often not enough - an input lag of even a few ms can throw you off, for example when you try to catch/prevent a skid, and 30fps max implies ~33ms frame latency. It's not about the graphics in hardcore racing sims, it's about the feel... not that we mind good graphics and all the extra jazz
Don’t disagree, we were suggesting a ‘minimum’ of 30fps be as low as gamers should go.
I wouldn’t assume to know the working relationship here, I suggest you wait until more facts are released or a new driver
I am not blaiming you Steve or Techspot of this review infact it is a fascinating review showing the bias results based on a poorly coded game for AMD card owners. It only tell us people that the game was programmed and realeased in favor of Nvidia cards with an enormous advantage I.e GTX 960 > R9 290x. Yes it is laughable if you are AMD card owner but it is also a fact that is it a bias game developed for a general consumer. You can entirely see Nvidia ads all corners of the race track, that is pretty obvious. "Tell it to the marines" hope they will believe it.
In a TV adverstisement for example the company pay a sum of money just to advertise their product(s) for every seconds it flashes on the screen. Still not obvious? "People who pretend to asleep is hard to wake up because they are just pretending that they are asleep but infact they are awake, they just want to ignore you because of their stupidity". Wake up people the truth is already on your face! Damn unfair competition. We already know that Intel paid a billion $ dollar settlement last year to AMD because of unfair competition practicies but still people are playing dumb and blind and still continue supporting the crooked company. What a shame, this only show how rotten our society.
yes it's true AMD has a bigger overhead. it seems AMD is using a custom command buffer that is similar to Dx11's Command List which nvidia added support for at a driver level. this might be legacy code that hasn't been changed yet from when there was no Command List to help with multithreading.
it might be that amd decided to not waste resources on adding support for an optional feature and wait for the newer api that will be released together with dx12.
in terms of pure multithreading, I don't remember where, but AMD had access to more threads than nvidia. nvidia does indeed have less overhead.
this is why on a few games AMD cards have lower framerate when the PC has a weak CPU (something like an i3).
But... taking all this in consideration, it doesn't change the fact that the devs are using Nvidia's proprietary code which can't be optimised for AMD and can't be turned off. Saying that it's AMD driver problem when they are using a game engine that had proper AMD support before they modified it just doesn't sit well with me.
The FPS problems when using Physx are known by all devs and gamers. (read Steve's comment from above)
I don't care what they use and how they use it, it's their game. I just don't like dishonest game devs/publishers. I always protests against them with my wallet.
Last edited:
SuperVeloce
Posts: 133 +34
Yes, I agree, this is not exactly a driver problem (as I stated in a first comment). But older nvidia cards also gets poor performance, so I bet it's not physx that is the reason. As it seems to me Nvidia corrected some issues for their newer cards, everyone else is bound to play it as the developers of the game coded it (poorly?). You can find Project cars alpha benchmarks where AMD performance was also bad
Evernessince
Posts: 5,469 +6,158
I'd also say that AMD's high bandwidth memory architecture helps here as well. Most games don't saturate the 512 bit bus on the 290/x very well but it seems practically forced at higher resolutions.
I am not sure what you are suggesting with the first part of your comment. The GTX Titan X is clearly limited at 1080p where it isn’t at 2160p, the fact that it can average 60fps at the 4K resolution is a good thing. Particularly given how amazing the game looks and how much fun the gameplay is. This is far from a terrible game.
As for the GTX 960 beating/matching the GTX 780 that’s not on the game developer.
Peter Farkas
Posts: 654 +530
Hi Steve,
I agree with the first half, but what do you mean by "that's not on the game developer"? Are you suggesting a driver problem by Nvidia?
I am, for some time now Kepler seems to be lagging further behind Maxwell and this is the worst example we have seen yet. (I would have thought the title of the last page gave that away).
It would appear as though Nvidia has stopped optimising for Kepler and is heavily focused on Maxwell.
I am not sure how you could blame this on the game developer anyway.
Filip B
Posts: 46 +9
As it seems, it's a driver problem. First, titan x I was suggesting when I get $1000 card I'd get 120+ frames, but in project cars it's pretty bottlenecked somehow, which is indeed very odd considering almost same result on 1080p and 4k (averages). Second, again drivers. It seems that nvidia is really trying to push Maxwell and doesn't care about Kepler cards (even high end 780/Ti) that is horrible and that's a few months after Maxwell release, what's going to happen in 2+ years? No support? But again that's a question for nvidia support.
Anyhow, thanks for a very nice and detailed benchmark as always.
Peter Farkas
Posts: 654 +530
I just wanted to understand as I was not aware of this situation with Kepler and Maxwell. This is not on the developer obviously.Is this just the way Nvidia is forcing customers to buy a new Maxwell based card?
If I were a 780 owner I would be pissed off big time.
Filip B
Posts: 46 +9
Hear, hear
Ahh right sorry I understand where you are coming from now. There is clearly something limiting both AMD and Nvidia GPUs. The fact that there is just a 10fps drop for 4K to 1080p on the Titan X is puzzling.
I am very eager to see what AMD comes up with as a fix.
If I spent $650 on the GTX 780 when it came out I would also be pissed.
NightAntilli
Posts: 929 +1,195
This is probably the most important post here...
Guess I won't be getting Project Cars. I will not support developers that pick sides like this.
NightAntilli
Posts: 929 +1,195
Evernessince
Posts: 5,469 +6,158
Thanks for the video. It's just sad that more people don't stand against the practices Nvidia uses. I scrolled down to the comments to find people bashing AMD. This was in no way a fanboy video touting AMD products but the behavior of Nvidia trolls is horrid.
I'm ashamed to own a 970.
Actually it affect nvidia as well, just AMD more because they weren't as strong in tessellation. So in the end nvidia was screwing their own customers, but screwing amd's more. That was ok for them
This is shameful coming from Slightly Mad Studios and Nvidia. Even I, a regular user, normal guy, would have known that developing a game using proprietary PhysX would cripple the game on competing GPU's. It's obvious. The fault lies entirely with SMS. SMS had the gall to deflect the blame off of themselves to AMD when they knowingly made a game that was by default, crippled on a subset of hardware.
Nvidia fanboys are not blameless either. Even after it was fully explained why the game runs poorly on AMD cards, they still used stereotypical insults against AMD and still place the blame on AMD. Then they act ignorant of anything told to them and pledge their support of Nvidia products EVEN MORE. Well DUH that's what they WANT YOU to do. Stupid sheep.
Then again, I don't even know why I'm shocked or surprised that this happened. A) it's an indie studio, you can't expect much ethical behavior from them and B) it's Nvidia. Downfall of gaming is already happening but people are too daft to see it.
Nvidia fanboys are not blameless either. Even after it was fully explained why the game runs poorly on AMD cards, they still used stereotypical insults against AMD and still place the blame on AMD. Then they act ignorant of anything told to them and pledge their support of Nvidia products EVEN MORE. Well DUH that's what they WANT YOU to do. Stupid sheep.
Then again, I don't even know why I'm shocked or surprised that this happened. A) it's an indie studio, you can't expect much ethical behavior from them and B) it's Nvidia. Downfall of gaming is already happening but people are too daft to see it.
Similar threads
- Locked
