Deus Ex: Mankind Divided PC GPU performance review updated with DirectX 12 results

AMD's PR hyped up DX12 to the max. Granted, ashes to date, is the only game that has really made great use of DX12, showcasing old Athlons gaining 200%+ perf boosts.

 

Also current game engines haven't really been designed with DX12 (or Vulkan) in mind.

 

No. It's just that none tests with frame variance in mind. Performance gains for actual gameplay, stutter tests, lower frame rates, is much much improved with both DX12 and Vulkan. It's not even close.

Apart from the complexities of the rendering pipeline, the defining change between DX11 and DX12/Vulkan is that the CPU can ask the GPU to "draw" items using all its cores instead of just one. That small difference is tremendous for latencies.

 

No? In what exactly?

 

Unsure how in-depth you are going to make the DirectX 12 tests but if you want to it is possible to disable async compute and check how much or if that feature is adding to the games overall performance under D3D12.

The game stores it's settings in the registry:

HKEY_CURRENT_USER\SOFTWARE\Eidos Montreal\Deus Ex: MD\Graphics

EnableDX12AsyncCompute

Just change that from 1 to 0, for me the difference was about 1 frame give or take but at 2560x1440 with everything but MSAA and TAA turned up yeah that's pretty GPU limiting so async can't really do much about that.

There's also EnableDX12StablePowerState which defaults to 0 and DisableDX12BufferPooling which defaults to 0 but I have no idea what those actually do.

For Rise of the Tomb Raider the developers (Or well Nixxes whom were responsible for the PC port for that game too.) made async toggleable via a startup parameter which might work for this game too, I haven't actually checked yet if that is available.

 

To me it feels like some people are downplaying the amount of stuff that is being rendered on the gpu in this game. Sure the animations and models do not look that good at times and some stuff here and there are rather meh. But all in all the detail in this title is immense. I don't really have a good comparison of a game that runs really good with similar amount of light sources, shadows and clutter as this one has.

 

AMD have made it for this duty and they sold it lot lower because it is more on topic with user and they still gain money on it.

starting with that the few last Fury Nano were sold at 199 Euro last month... nice for a starting price of 580 Euro.

 

You're right on thjis, but it saddens me that no matter how many times this is being brought up, people always only judge DX12 performance from charts by raw fps. Also they comment of how unoptimised games are without even playing the games themselves.

One of the biggest things I just noticed in Deus Ex is that how much more efficient DX12 is with memory. I run the game at 1080p everything maxed. Fury X literally runs out of memory, and so it stores a lot in system ram instead (GPU mem usage is at max all the time, and ram usage is between 12-14GB).

Under DX11 , when only one core can talk to gpu, and stream textures in to gpu, you get those stalled frames (heavy stutter). But under DX12 you have multiple cores feeding the gpu (also memory vise), so you don't get stalls, but instead you get a bit slower frames.

So when both paths run out of memory. DX12 suffers less (it still suffers, but game don't freeze).

 

That's because people don't generally understand how heavy those effects are. Also since they are more like additional effects that are harder to notice, people don't give them the credit they deserve. Great textures are always easy to notice. Judging pc games in general is hard, and takes a much more to analyze them than just few screenshots, performance-wise.

 

Tested with 16.8.2 , we are here for next amd driver ( 16.9.1 with more perf than 16.8.2 ) :3eyes:

 

So, nVidia fans dont like what AMD just did with this game. Just for your info, we, AMD fans, dont like what nVidia does to the majority of games. We like to see optimisations for every hardware from the developers, not just for who pays more. Project CARS its just one example where AMD is destroyed by the developers and nobody cares, its AMDs fault, they just need to be better ^-. I dont like hypocrisy.

 

pCARS was AMD's fault though. They choose not to work with slightly mad. Plus the majority of AMD's performance problems in DX11 is from their drivers not the games.

 

Alot of the time it is AMD's fault, you see that when performance fixing drivers appear at a later date.
Was PCARs not because AMD drivers lack Dx11 MT?

Doom is another example, id were getting slaughtered by AMD owners for taking Nvidia money, now they are their new fave devs and it's (wrongly) used as a benchmark for low level api's.

 

It's ok. We just won't buy it. There you go.

 

is Hilbert aware of this? he should update his article with gameplay test.

 

16.8.2 drivers used there, HH used lasted dx12 drivers from AMD 16.9.1.

That review show a different scenario. Not valid for comparison.

 

In-game benchmark is more accurate I think. This RX480 video shows DX11 vs DX12 and it's not actually clear-cut, I quote myself as I original posted this in the other Gaming section DE:MD thread. It was originally linked by "ghost" in the dsogaming comment section found here http://www.dsogaming.com/news/deus-...vailable-performs-poorly-on-nvidias-hardware/ This article was originally linked by Rich_Guy from Guru3d.

 

The same guy also posted a video of Fury X vs 1070/1080.

Have a look how your 1070 struggle to keep up with over a year old Fury card.

https://www.youtube.com/watch?v=afA9A88hP_M

 

The RX480 struggles to keep up with a 980Ti in almost every game ever.

 

Really Denial a 480 and 980ti comparison?