AMD Radeon Vega 64 Outperforms NVIDIA GTX 1080 Ti in Forza Motorsport 7, DX 12

[JalleR]

*Cough* ... console port ...*cough*

Ding ding ding we have a winner...

 

[efikkan]

The test wasn't based on the in-game benchmark, so as to avoid specifically-optimized scenarios.

Anything but a reproducible benchmark is completely worthless.

Drivers don't optimize games in real time, in fact, driver optimizations are mainly limited to profiles with driver parameters, with rare cases of tweaked shader programs or special code paths in the driver. The driver never knows the internal state of the game.

This will probably be the case when more than 90% of all games made are dx12 only.

No, this is another case of a console game ported to PC. There is nothing inherent in Direct3D 12 that benefits AMD hardware.

 

[Nate1492]

Anything but a reproducible benchmark is completely worthless.

Drivers don't optimize games in real time, in fact, driver optimizations are mainly limited to profiles with driver parameters, with rare cases of tweaked shader programs or special code paths in the driver. The driver never knows the internal state of the game.


No, this is another case of a console game ported to PC. There is nothing inherent in Direct3D 12 that benefits AMD hardware.

The thing is, there are other 'console ports' that don't follow this pattern.

Mind you, it's not really a port anymore, since they share the same x86 processor, it is more of a 'tweak the game differently'.

 

[EarthDog]

^1080Ti just sayin'
Unlikely, in terms of sheer HP the 1080Ti is still better.

That's a HUUUUUGE assumption.

Didn't they have to adjust a config file on the beta to get it to run past the 60 fps limit or something?

Then, oddly enough, as the res increases, typically AMD cards with HBM.HBM2 do better, however, the less bandwidth NVIDIA cards are catching up... I don't get that. And as such, part of the reason I'm not sold on this becoming a norm.

As I said in the other thread, so far, it is an anomolous result.

I don't think the surprise is that Vega does so well: it's that RX 580 does so poorly at high resolutions. Forza Motorsport 7 is a game made for Xbox One X at 4K and, by extension Polaris but the RX 580 performs well under the GTX 1060 it should easily trump.


I think what's happening is that AMD's memory subsystem rears its head again. When bandwidth isn't a problem, GCN does fantastic. As bandwidth demands increase, GCN takes a bigger blow than Pascal does. Vega manages to stay on top only because of HBM2. In games that are more demanding, HBM2 is not enough to satisfy Vega either.

This is actually quite easily explained by using 8xAA. That's something that's not going to result in a lot of cache misses, hiding AMD's problem and giving a misleading good impression.

I think 8xAA at 4K explains it.

 

[niko084]

Forza is a neutral title.

But is it? It's a console port from the Xbox One.
https://www.techpowerup.com/gpudb/2086/xbox-one-gpu ---Made by AMD.

Personally this makes zero difference to me as I'll be playing on my Xbox.

 

[Nergal]

DX12 games are growing in numbers, making this more relevant day by day.

 

[bug]

DX12 games are growing in numbers, making this more relevant day by day.

I don't know about "day by day". Starting from zero, of course the usage can only go up. But DX12 is 26 months old and we still don't have 26 DX12-only titles. Just imagine if you bought a "futureproof" video card for Christmas 2015

 

[Th3pwn3r]

Forza is a neutral title.

As I see some 1080 owners just popped up and started crying.

Who is crying? For one it's just ONE game performing well . Two, 1080ti owners can probably afford another video card, it's not a big deal. I just got a 1080ti, have a 1080 and can return the 1080ti whenever I want to change cards or upgrade to Vega64

All that being said I'd hate to have to swap cards for whenever I want to play a DX12 game though.

 

[EarthDog]

DX12 games are growing in numbers, making this more relevant day by day.

True... .but.. how long has it been out and how many titles are DX12? Not as much as we would all like to see.

Also, this doesn't happen in other DX12 titles... so, is it really DX12 in the first place? We hope, but are we sure?

 

[HisDivineOrder]

DX12 games are growing in numbers, making this more relevant day by day.

Looks more like other factors, less about DX12. Unless you have other DX12 results that mirror these results?

 

[eddman]

Mind you, it's not really a port anymore, since they share the same x86 processor, it is more of a 'tweak the game differently'.

It doesn't matter if they are all x86. Intel and AMD CPUs have different architectures, and optimizing for one does not mean it'd run just as well on the other.

Same applies to GPUs. Consoles use AMD GPUs, so when porting to PC, developers have to optimize for nvidia too, specially with DX12 where the driver plays a much smaller role compared to DX11, although some don't bother to.

DX12 games are growing in numbers, making this more relevant day by day.

We had 9 non-store DX12 games in 2016 and a grand total of 3 in 2017 so far. That's a decrease. Maybe 2018 would be better?

We could also count store games but that's a slippery slope. Store games can only be DX12 so developers have no choice even if they didn't want to use it. Look at quantum break. It's a DX12 game on store but when it came to steam it became DX11.

We had 5 store only DX12 games in 2016 and 2 in 2017 so far, all published by MS.

 

[wiak]

Is this the way to use DX12 the right way scenario or just an implementation where AMD is favored in programming ?

Is the engine just anti-nvidia ?

Will other Games follow where Polaris and Vega can shine?

same way project cars is anti amd

 

[lightofhonor]

My guess is this is due to the overhead DX12 games have on Nvidia hardware, similar to how DX11 games are often limited by AMD driver overhead. 1080ti is probably not pegging 100% utilization, similar to this:

With CPU's as fast as they are now I don't think this will crop up too often, but just one example of hardware beating fast software.

 

[Krzych]

Another one to the collection of games where Vega is competitive, it now contains two games.

 

[efikkan]

I don't know about "day by day". Starting from zero, of course the usage can only go up. But DX12 is 26 months old and we still don't have 26 DX12-only titles. Just imagine if you bought a "futureproof" video card for Christmas 2015

Those who bought "future proof" AMD cards back then are still waiting for them to take off, basing their anecdotes on games which are outliers rather than the norm, and concluding that every game can perform like that on this (superior) hardware.
Meantime, in the real world the competition looks better than ever.

 

[Th3pwn3r]

Those who bought "future proof" AMD cards back then are still waiting for them to take off, basing their anecdotes on games which are outliers rather than the norm, and concluding that every game can perform like that on this (superior) hardware.
Meantime, in the real world the competition looks better than ever.

I don't know man, there are some people who are really, really into Forza that would buy the card just for this game.

 

[sinnedone]

It doesn't matter if they are all x86. Intel and AMD CPUs have different architectures, and optimizing for one does not mean it'd run just as well on the other.

Same applies to GPUs. Consoles use AMD GPUs, so when porting to PC, developers have to optimize for nvidia too, specially with DX12 where the driver plays a much smaller role compared to DX11, although some don't bother to.


We had 9 non-store DX12 games in 2016 and a grand total of 3 in 2017 so far. That's a decrease. Maybe 2018 would be better?

We could also count store games but that's a slippery slope. Store games can only be DX12 so developers have no choice even if they didn't want to use it. Look at quantum break. It's a DX12 game on store but when it came to steam it became DX11.

We had 5 store only DX12 games in 2016 and 2 in 2017 so far, all published by MS.

Correct me if I'm wrong but I thought I read somewhere that in order for DX12 to be correctly utilized the game engine needed to be built on DX12 and not just tacked into an existing engine.

 

[efikkan]

Correct me if I'm wrong but I thought I read somewhere that in order for DX12 to be correctly utilized the game engine needed to be built on DX12 and not just tacked into an existing engine.

Correct. More or less all games currently uses an abstraction layer to make Direct3D 12 act like Direct3D 11.

 

[MxPhenom 216]

Is this the way to use DX12 the right way scenario or just an implementation where AMD is favored in programming ?

Is the engine just anti-nvidia ?

Will other Games follow where Polaris and Vega can shine?

Considering it was made with Xbox One in mind, id say it just favors AMD hardware.

 

[Xzibit]

Considering it was made with Xbox One in mind, id say it just favors AMD hardware.

Doesnt explain FH3 or F6

This one just swings way back the other way

 

[Camm]

IMO, the texture assets are insane in Forza, and this is just a prime example of how fast can you feed the card being the primary driver.

1080P\1440P - HBM2's lower latency is feeding the GPU faster, even though bandwidth similar to GDDR5X.

4K - Vega VRAM buffer exceeded and caching from system, textures still fully in VRAM for 1080 Ti. Lower FPS also slows the amount of fill requests.

 

[meirb111]

it also outperformer in tetris and pacman 3d .

 

[ratirt]

I think there's more things to consider here than HMB2 memory or the fact that DX12 always favors AMD or it's a plugin or other from a console. (BTW consoles use same stuff now as PC's do). I think it's the matter of coding and programmers must code the games to utilize the performance of the cards and what they're capable off plus the resources the card has itself. What i'm saying is there's no 1 straight answer to this. You have to see the bigger picture and put all that matters into one. Saying it's HBM2 makes it faster or anything other is crap and not true. There's more important things in this equation that matter here but I guess it would be hard to list them all since there's so many of them.

 

[anubis44]

It's unlikely that MS would favor AMD over NVIDIA "just because". It is known that AMD in general works better in DX12. NVIDIA has a way higher market share. Unless MS wants to create a more equal balance by doing so in favor of AMD. Then again, MS games aren't sold in numbers as high as other games so... not relly sure.

Actually, it's a common misperception that nVidia has a 'higher market share'. Yes, on just the PC platform, nVidia has higher market share, but people forget that the PS/4 and XBOX consoles are really just x86 PCs with some custom hardware tweaks. If you look at it THAT way, then you can see that AMD Radeons likely power >60% of the x86 gaming market. nVidia is actually slowly getting squeezed out of the x86 gaming market.

Here's a link from 2016 showing AMD with 56% of the x86+GPU market--I'm projecting that it's higher now, since consoles generally outsell PC graphics cards (if anyone has anything more recent, please post): https://www.pcgamesn.com/amd/57-per-cent-gamers-on-radeon

Game developers know this, and they are in fact migrating their coding over to Radeon-optimized code paths. And before anybody jumps on this and says 'but nVidia customers spend more money on hardware, well, it's important to understand that to a game developper, some millionaire with two GTX1080 Ti's is only worth the same $60 as a 12 year old with an XBox One. They're both spending $60 on the latest World of Battlefield 7 (yes, made up game) title, so in the end, the developer is looking at which GPU is in the most target platforms.

If I'm Bethesda, for example, I realize that over 60% of the x86 machines that have sufficiently powerful GPUs (XBox, Playstation, PC) that I'd like to put my game on are powered by AMD Radeons. It's only logical that I'm therefore going to build the game engine to run very smoothly on Radeons. All Radeons since the HD7000-series in 2011 have hardware schedulers, which makes them capable of efficiently scheduling/feeding the compute/rendering pipelines from multiple CPU cores right in the GPU hardware. This makes them DX12 optimized, whereas nVidia GPUs, even up to today's 1000-series, still don't have this built-in hardware.

The extra hardware does increase the Radeon's power draw somewhat, and nVidia has made much of how their GPUs are more 'power efficient', but in reality, they're simply missing additional hardware schedulers, which if included in their design, would probably put them on an even level with AMD's power consumption. It's a bit like saying my car is slightly more fuel efficient than yours because I removed the back seats. Sure, you will use a little less gas, but you can't carry any passengers in the back seat, so it's a dubious 'advantage' you're pushing there.

Those who bought "future proof" AMD cards back then are still waiting for them to take off, basing their anecdotes on games which are outliers rather than the norm, and concluding that every game can perform like that on this (superior) hardware.
Meantime, in the real world the competition looks better than ever.

Well, I bought a Radeon R9 290 for $259 Canadian back in December of 2013, which has played every game I like extremely well, and it might just run upcoming fully optimized DX12 titles like Forza-7 as well as or better than an $800 1080 Ti in the 99% frame times, so I'd hardly call that a fail. In fact, when the custom-cooled Vegas hit, I'm probably going to upgrade to one of those in my own rig, and put the R9 290 in my HTPC, because it'll probably still be pumping out great performance for a couple more years to come.

 

[EarthDog]

and it might just run upcoming fully optimized DX12 titles like Forza-7 as well as or better than an $800 1080 Ti in the 99% frame times, so I'd hardly call that a fail

I wouldn't hold my breath on that....