AMD RX Vega 64 Outperforms NVIDIA GTX 1080 Ti By Up To 23% In DX12 Forza 7

[Humbug]

1 hour ago, Vode said:

 

Damn I hope Pascal is not going to die the Kepler death.

well Volta for gamers isn't even out yet so that's not what's happening.

 

but it is pretty bad for kepler looking at the graphs in the 2nd post in this thread. Just compare former competitors; gtx 970 vs r9 390. Or gtx 960 vs r9 380. Now they are in different leagues.

[Techicolors]

Not sure whether to start a new thread or not, but the new Nvidia drivers 387.92 improved Nvidia card performance in Forza 7

https://www.overclock3d.net/news/gpu_displays/nvidia_s_geforce_387_92_driver_offers_a_15-25_performance_boost_in_forza_motorsport_7/1#.WduAk5Tv9GY.reddit

https://www.computerbase.de/2017-09/forza-7-benchmark/2/

 

Computerbase.de only tested the 1080 and 1060, but indeed gains are around 20% as claimed. however the 1080 is still 13% slower than the Vega 64 (1% slower than 56). on the other hand, the 1080 gains on the 64 at 1440p, then beating it at 4k 

[Taf the Ghost]

4 hours ago, Technicolors said:

Not sure whether to start a new thread or not, but the new Nvidia drivers 387.92 improved Nvidia card performance in Forza 7

https://www.overclock3d.net/news/gpu_displays/nvidia_s_geforce_387_92_driver_offers_a_15-25_performance_boost_in_forza_motorsport_7/1#.WduAk5Tv9GY.reddit

https://www.computerbase.de/2017-09/forza-7-benchmark/2/

 

Computerbase.de only tested the 1080 and 1060, but indeed gains are around 20% as claimed. however the 1080 is still 13% slower than the Vega 64 (1% slower than 56). on the other hand, the 1080 gains on the 64 at 1440p, then beating it at 4k 

Probably not worth a new thread. But also not unexpected. There was some oddities in the mix of the original testing, and it seemed like Nvidia cards were bottlenecking at the lower resolutions. 20% increase at 1440p is really impressive.

 

Forza's engine seems to lean heavily on Compute, which is the reason it runs so well on AMD. (And the reason for the GPU shortages.) Nvidia could adjust to the problem, but they needed to update their driver to do it. It'll be interesting if we end up with the weird reversal of normal where Nvidia has to patch to improve performance more than AMD does.

 

This is also really interesting, as it seems a lot like word came down a few years ago that Console Games needed to move to heavier Compute usage. That should filter out over time to the rest of the industry. (It's also performance that's been left on the table, but as both Nvidia & AMD are putting most of their work into Compute, it's best the Game Publishers get with the program.)

[leadeater]

2 hours ago, Taf the Ghost said:

but as both Nvidia & AMD are putting most of their work into Compute, it's best the Game Publishers get with the program

With Nvidia actively disabling compute features on GeFroce cards, wonder how quickly that will change if games start tipping the scales towards compute performance requirements.

 

With AMD doing this

[Vode]

On 10/5/2017 at 8:23 PM, Humbug said:

Vega beating the gtx 1080 happens in a number of games...

 

it's beating the 1080ti that is unexpected.

I know. My statement was badly worded. I meant Vega 64 vs. GTX 1080 Ti/Titan Xpeepee and Vega 56 vs. GTX 1080.

[Taf the Ghost]

3 minutes ago, leadeater said:

With Nvidia actively disabling compute features on GeFroce cards, wonder how quickly that will change if games start tipping the scales towards compute performance requirements.

 

With AMD doing this

There's an interesting discussion when it comes to AMD CPUs, as we've never seen any of the Cat Cores scaled up to the mainstream platforms. We don't really know how potent those cores are in the PS4/XBone because we don't have copies available to play with. They might be more potent than we realize because of their highly customized & barebone nature.

 

Which points in some interesting directions, given the convergence of Console & PC gaming.