Benchmark analysis: Battlefield V Open Beta: PC

Interesting the Vega performance seems lacking at 1080p & 1440p, yet take a jump at 4K.

Thank you HH!

 

Not really surprising. In one of dev. team talk they said something weird. That, they have bad performance from using many instantiated objects. And instead they try to combine as much geometry under one instance as possible.

That should not be problem, but it looks like their solution is choking AMD's driver or cards somehow and preventing them from fully utilizing brute part of architecture. That bottleneck is likely "removed" by higher resolution as there is more brute force work required than there is load on bottlenecking serial operation.

May be anything, but my guess would be geometry discard as they make those objects complex.

 

I wonder what it is about Frostbite that makes DX12 such a mess. You'd think the team that requested/helped AMD build an entire low level API would be the best at DX12 and yet they are somehow one of the worst.

I guess we'll see when the final build ships and numbers updated but the performance/stuttering mentioned doesn't look any better than their previous DX12 implementations.

 

I don't know about DX12 with the Frostbyte engine... I tend to wonder the same as you @Denial , especially as they are actively working together with Nvidia to implement RTX, which is DX12. But to what use if the DX12 adaption itself is lacking? Maybe they should first fix their DX12 "container" itself before thinking about RTX performance numbers, since those are crippled by, well being DX12 on Nvidia and Frostbyte in the first place.

 

1080ti looking ok at 4k I am assuming that's stock not overclocked (only need 60fps)

 

Well, Hilbert is usually not benchmarking with overclocked HW when comparing different CPUs/GPUs. Only doing OC potential for particular HW in review of that HW. So reader sees what he can expect.

That's why his 4K result matches that linked one. It is no longer bottlenecked by other things than brute GPU power.

 

DX12 vs DX11, 4.2Ghz X99 processor vs 5Ghz Z370 processor.

Not to mention a quick google search shows multiple sites with results +- 15% of one another, on top of Hilbert literally saying the performance is all over the place in the game, on top of multiple user reports complaining about how the performance is a mixed bag..

 

lol, dx12 fails again...
Where is promised boost?

BTW this RAM usage is kinda weird too, AMDs eats a lot less of it.

 

Both cards are AMD.

Failure of DX12 is on the developer, not the API.

 

Pretty typical - Vega's overall performance isn't that impressive but its performance proportionately scales up better than Nvidia's hardware.

DX12 and Vulkan reduce CPU and PCI-e overhead. If the GPUs are bottlenecked, there isn't going to be [much of] an improvement.

 

Silly question, do DX12 and Raytracing go hand in hand ?

 

In case you get curious... all the answers you seek & more lol

"DirectX Raytracing Tutorials - Github code repository"
"DirectXTech Forums - Getting started with DirectX Raytracing "

 

Thanks for benchmarks. The game itself sucks really bad though.

 

No, it's just you...

 

I think by now it's safe to conclude that doing DX12 right is hard, and that everyone is still learning ins and outs.

 

I don't think MS makes it any easier as well.

 

Or it's getting hard to fully utilize all those GCN cores - something we already saw with Fury