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

[Raevenlord]

In an interesting turn of events, AMD's latest flagship videocard, RX Vega 64, has seen a gaming performance analysis from fellow publication computerbase.de, which brought about some interesting - and somewhat inspiring findings. In their test system, which was comprised of a 4.3 GHz Intel Core i7-6850K (6 cores), paired with 16 GB of DDR4-3000 memory in quad-channel mode, and Crimson Relive 17.9.3 / GeForce 385.69 drivers, the publication found that the Vega 64 was outperforming the GTX 1080 Ti by upwards of 23%, and that percentage increases to 32% when compared to NVIDIA's GTX 1080. The test wasn't based on the in-game benchmark, so as to avoid specifically-optimized scenarios.





8x MSAA was used in all configurations, since "the game isn't all that demanding". Being it demanding or not, the fact is that AMD's solutions are one-upping their NVIDIA counterparts in almost every price-bracket in the 1920 x 1080 and 2560 x 1440 resultions, and not only by average framerates, but by minimum framerates as well. This really does seem to be a scenario where AMD's DX 12 dominance over NVIDIA comes to play - where in CPU-limited scenarios, AMD's implementation of DX 12 allows their graphics cards to improve substantially. So much so, in fact, that even AMD's RX 580 graphics card delivers higher minimum frame-rates than NVIDIA's almighty GTX 1080 Ti. AMD's lead over NVIDIA declines somewhat on 2560 x 1440, and even further at 4K (3840 x 2160). In 4K, however, we still see AMD's RX Vega 56 equaling NVIDIA's GTX 1080. Computerbase.de contacted NVIDIA, who told them they were seeing correct performance for the green team's graphics cards, so this doesn't seem to be just an unoptimized fluke. However, these results are tremendously different from typical gaming workloads on these graphics cards, as you can see from the below TPU graph, taken from our Vega 64 review.



View at TechPowerUp Main Site

 

[_Flare]

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?

 

[dicktracy]

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?

Find out on the next episode of dragon ball z

 

[RejZoR]

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.

 

[R0H1T]

Find out on the next episode of dragon ball z

DBZ Super Saiyan GodZ

 

[Prima.Vera]

Nice. One cherry-picked game.
Let's have it for at least 50 games and then draw a median line with the performance.

 

[dj-electric]

Apparently we need a news piece every time vega 64 beats gtx 1080 at something

 

[ensabrenoir]

Find out on the next episode of dragon ball z

.....the comment just made this thread relevant.....

 

[Bytales]

Nice. One cherry-picked game.
Let's have it for at least 50 games and then draw a median line with the performance.

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

 

[R0H1T]

Apparently we need a news piece every time vega 64 beats gtx 1080 at something

^1080Ti just sayin'

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

Unlikely, in terms of sheer HP the 1080Ti is still better.

 

[I No]

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

 

[londiste]

I wonder what the technical reason behind this is, especially considering the rather quick falloff as the resolution increases. Simply geometry/rasterization vs shaders and changes in bottlenecks?

Edit:
Looking at the scaling from 1080p to 1440p and from 1440p to 2160p, I would say 1080Ti (and to a lesser degree, rest of the nVidia lineup up to 2160p) is somehow CPU-limited. nVidia drivers having larger overhead, async scheduling in drivers or something?

In GPU limited circumstances, I would expect the performance difference from going to larger resolution to be 20+% (maybe 15-20% for 1080p>1440p primarily because this usually tends to be CPU-limited anyway). Here is what the 1080p>1440p and 1440p>2160p performance hit looks like based on these graphs:

1080ti - 7.6% and 11.7%
vega64 - 16.0% and 27.2%
gtx1080 - 8.6% and 20.5%
vega56 - 12.0% and 24.1%
gtx1070 - 6.4% and 22.0%

 

[evernessince]

Apparently we need a news piece every time vega 64 beats gtx 1080 at something

Actually it's Vega 56 beating the 1080 Ti. But you know, actually looking at the pictures or reading must be hard.

 

[Imsochobo]

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?

Polaris doesn't shine here and is positioned accordingly in the charts vs 1060 (it's right place, nothing out of the ordinary)
When I mean shine here, shine more than usual in DX12\Vulkan.

Thus Vega is in a right place I guess.

 

[Imsochobo]

I wonder what the technical reason behind this is, especially considering the rather quick falloff as the resolution increases. Simply geometry/rasterization vs shaders and changes in bottlenecks?

Edit:
Looking at the scaling from 1080p to 1440p and from 1440p to 2160p, I would say 1080Ti (and to a lesser degree, rest of the nVidia lineup up to 2160p) is somehow CPU-limited. nVidia drivers having larger overhead, async scheduling in drivers or something?

In GPU limited circumstances, I would expect the performance difference from going to larger resolution to be 20+% (maybe 15-20% for 1080p>1440p primarily because this usually tends to be CPU-limited anyway). Here is what the 1080p>1440p and 1440p>2160p performance hit looks like based on these graphs:

1080ti - 7.6% and 11.7%
vega64 - 16.0% and 27.2%
gtx1080 - 8.6% and 20.5%
vega56 - 12.0% and 24.1%
gtx1070 - 6.4% and 22.0%

Nvidia have always had less overhead so it seems weird.
DX12\vulkan is still a bit of a wildcard though

 

[darksf]

Well it is console port and on consoles they use AMD graphics so it looks normal they will need some optimisations for nVidia now probably.

 

[bug]

Nice. One cherry-picked game.
Let's have it for at least 50 games and then draw a median line with the performance.

Still good to know if you're into these games. Plus, it's an oddity, us enthusiasts like to know everything about these

 

[Parn]

I think the reason behind this is a combination of "AMD being generally better in DX12" and "it's a console port i.e. optimised for AMD".

 

[B-Real]

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?

Forza is a neutral title.

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

 

[FeelinFroggy]

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

Will either of these cards be relevant then? Other than Ashes, what other games are exclusively dx12?

 

[londiste]

Other than Ashes, what other games are exclusively dx12?

Anything Microsoft does/publishes - Gears of War 4/Ultimate, Forza series, Killer Instinct. Everybody else seems to have DX11 for backup.
In anything not exclusively DX12, nVidia users are practically always better off relying on DX11 path.

 

[bug]

Will either of these cards be relevant then? Other than Ashes, what other games are exclusively dx12?

That's actually a good question. When I buy a video card, I look at how well it performs in current games (I play titles up to 2-3 years after they're launched anyway). Futureproofing means nothing to me, because in 2-3 years, I know there will be another card out there for me to buy.

 

[FordGT90Concept]

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.

 

[eddman]

These just seem off. 1080 Ti is usually about 20-25% faster than GTX 1080 at 1440, yet here it's just ~9%?!

IMO pascal is under performing here. It could be a driver issue, or game's optimization, or both. It might also be possible that the FPS meter software they're using is causing a drop. It's not unheard of.

It's also not a DX12 thing, because 1080 Ti is faster than vega in other DX12 games and usually by quite a margin.

 

[FordGT90Concept]

I think 8xAA at 4K explains it.