yeah 9fps on 4k for a fury x seems really buggy

Really? It takes 4k resolution for the GTX1080Ti to beat the Vega 64? and at 1080p the RX580 has better minimum fps than a GTX1080Ti? This is some BS.

I mean the Ti is only running 25% faster than a 1070 in FHD, there is clearly some kind of CPU issue. Vega is performing nicely though compared to the rest of AMD's stack, so it seems like the game is well optimized for it. The article states there is a CPU comparison benchmark but I don't see it?

Denial:

I mean the Ti is only running 25% faster than a 1070 in FHD, there is clearly some kind of CPU issue. Vega is performing nicely though compared to the rest of AMD's stack, so it seems like the game is well optimized for it. The article states there is a CPU comparison benchmark but I don't see it?

Knowing what I know about Quantum Break, I've seen enough.

zero_cool:

not bs this.. Software vs. hardware schedulers in DX12. AMD can actually send commands to the GPU hardware schedulers on any thread in DX12 where they were limited to only the main driver thread in DX11. Starting to see the benefits of the hardware schedulers AMD uses. here you go

FM7 is the only DX12 game to exhibit these kind of performance numbers for the GTX1080Ti vs every other AMD card including Vega. Not even Hitman, Deus Ex: Mankind Divided or AoTS does what this game is doing. Not even Doom using Vulkan. Now suddenly with FM7 we see this huge discrepancy in performance, especially at 1080p and 1440p? Come on, surely even you can see something isn't right.

Loophole35:

Just testing the game out. Core0 bug is still there. FPS locked at 60. I'm calling all kind of BS on this "article".

Had it occurred to you that maybe you have vsync on? Or perhaps you don't actually have access to the latest build? Unless whoever formed the charts just completely outright lied about them, clearly, they have a version of the game that goes beyond 60FPS. Even at 4K, a 1070 will go beyond 60FPS. That sounds pretty damn good to me. Not sure what you're getting all huffy about. Does it really hurt your pride that much that a Vega outperforms a 1080Ti? To my recollection, there are currently tests available where that can happen; it's not un-heard of. It's just a bit unusual in gaming. Don't forget, the Fury X did abysmally in 4K

Definitely something happening ... looks like either a game bug or drivers not as Game Ready as AMD's were for Vega.

Have Nvidia for this game game ready drivers? My gtx 980 ti wont even start the game?

Loophole35:

No v-sync Frame rates are capped in engine and not unlocked Turn 10 stated so on their forum. Or they don't have a special version and are lying. It's not unheard of. I don't care if Vega out performs 1080ti the point is that 1080ti only being 20% faster than 1070 is a problem.

End of Quote Copy from the article you should have been reading before posting: Our German colleagues from computerbase.de have found a workaround, the full version can be unlocked in the cap.

Quote "Basically what the guys did was use an Ultra HD monitor, got rid of the cap and VSYNC, however in order to not restrict themselves, they re- sampled with the in-game up/down sampling options to match Full HD and Ultra HD resolutions (emulating them). The methodology they used certainly doesn’t win the equivalent of a beauty contest, but for some quick raw numbers it does get the job done towards a reasonable enough proximity." https://s26.postimg.org/8xt9px9b9/there_you_have_it.png

+schmidtbag This isn't about pride, this is about the results being suspect, considering all the other results we have comparing these cards against each other, especially in other DX12 games. The Fury X 4K numbers is a seperate issue entirely and has no bearing on what's being said.

There's definitely a problem as some noticed 80% underutilization on their 1070's.

Loophole35:

Just testing the game out. Core0 bug is still there. FPS locked at 60. I'm calling all kind of BS on this "article".

On my Desktop with Titan X (Pascal) and laptop GTX 1080, my framerates do exceed the monitor's refresh rate, no matter if at native resolution or DSR. I have Performance Target set to Unlocked in Forza Motorsport 7 (full version Ver. 1.106.2356.2, not the demo, so perhaps the Demo is based on an older build). I'm on Windows 10 1703 with the latest updates and using the 387.78 Nvidia drivers (not the game ready ones). The only problems I have is that the framerate tanks during some pre-race menus and victory celebrations (there seems to be a 30 FPS limit in place for some of those). Then again, I'm not using the latest "Game Ready" drivers and have no driver profile for the game, so perhaps those would solve these issues (Update: Tried the 385.69 Game Ready driver, made a little bit of a difference in menus if any at all, so I went back to 387.78 after yanking the updated profiles from the Game Ready driver on the way). I'm not sure about the Core0 bug, I can get several Cores running at close to 100%, depending on the resolution and traffic, but it's mostly less. Core 0 isn't locked at 100% either. The GPU usage is high, too. The game feels smooth (especially when enabling GSync on my laptop, but also without GSync on the desktop although it's slightly worse), framerates are pretty consistent and I haven't observed any stuttering except when the framerates tank in some of the menus. Except for the menus, the game performs roughly the same as Forza Horizon 3 (latest version) on both systems for me (using maxed out custom settings except Anti-Aliasing set to 4x at UHD resolution).

Denial:

I mean the Ti is only running 25% faster than a 1070 in FHD, there is clearly some kind of CPU issue. Vega is performing nicely though compared to the rest of AMD's stack, so it seems like the game is well optimized for it. The article states there is a CPU comparison benchmark but I don't see it?

I wouldn't read too much into this early benchmark. There are so many issues with the PC version of this game and this benchmark is a proof of that. It took them nearly a year to fix Forza Horizon 3 so hopefully it won't take as long with FM7.

Loophole35:

How are you playing on windows 7? Latest drivers are 385.69. You have a driver from 5 months in the future?

Heh, the 7 was a Freudian slip, playing on Windows 10 1703, obviously. Sorry Forza Motorsport 7, Windows 10. Fixed the post. 😛 The driver is the Nvidia 387.78 driver from the latest Windows 10 1709 Insider Preview Windows Update. The driver works just fine in 1703 (actually, best driver I have used on 1703 so far). You can download it in a nice installer package courtesy of @Cyris here on the forums. Good to see the game perform so nicely on AMD hardware, by the way. Looks like there is still a lot of untapped potential under the hood of the Vega chips, if games are optimized properly for them. I expect Nvidia performance will go up a bit eventually, but still, it's good to see AMD perform strongly.

Loophole35:

More playing around with the game. I don't know what I did but framerates unlocked for me. The Core0 bug is really maxing 2 threads out one main thread on core 1 and one HT on core 2. They both pretty much stay at 100%. Not sure it this is driver or game. My bet is game. Could you monitor what core 1 and core 4 are doing on your end also what is your GPU usage?

Seems drivers issue. TWW2 at DX11 with GTX1080ti and 6800K at 4.3 (with uncore at 3.75) is using core 6 and 5 at 100% and 75-100% respectively, while cores 1,2,3,4 are between 25-60% usage. The HT threads are at 0-3%. At DX12 the core 6 still is at 100%, core 5 hovering between 40-50, other cores between 20-30% and fps tank by 35% compared to DX11. That at 2560x1440, with everything on Ultra, and 2x AA.

Wooow...Performance is par for AMD...Seeing first time ever...i think driver problem for Nvidia

sunnyp_343:

Wooow...Performance is par for AMD...Seeing first time ever...i think driver problem for Nvidia

Its Dx12, maxwell being.. ass and pascal slightly less ass - expected, could 've told you that before the game/demo was out.

I think its a driver or game problem . RX 580 and a fury faster then a GTX 1080 Ti. And the performance between some GPU's are so close with so much difference in CU's and cores. There is someting wrong. Also i can not play the game.

zero_cool:

not bs this.. Software vs. hardware schedulers in DX12. AMD can actually send commands to the GPU hardware schedulers on any thread in DX12 where they were limited to only the main driver thread in DX11. Starting to see the benefits of the hardware schedulers AMD uses. here you go

This only explains AMD's higher than normal performance, it removes the bottleneck specific to their architecture. On the Nvidia side, the cards are clearly not being utilized at 100%. A 1080Ti is not 25% faster then a 1070 - it's significantly more. And that limitation isn't something in Nvidia's hardware. The hardware might require a separate render path for Nvidia, maybe even one that's potentially is harder to manage than AMD's, but it's not impossible because other developers have done it and Nvidia does it in every DX11 game. To me this is an example of AMD's console presence paying dividends across their product lines. What sucks is in the past Nvidia could do something to fix the implementation for the game dev - with DX12 it's all on them.