UE4 Elemental tech demo available for download

[UTmaniac]

Hasn't the Unreal Engine had Nvidia PhysX/Gameworks in it going back to at least UT3? So no surprise AMD cards will not look so good in this demo.

I changed the demo shortcut> properties> Target to %windir%\system32\cmd.exe /q/c start ElementalDemo.exe -D3D12 -novsync -ResX=1280 -ResY=800

The fps/frametime monitor then stayed green for all but ~2 seconds (in pieces) of the demo.

 

[muziqaz]

Hasn't the Unreal Engine had Nvidia PhysX/Gameworks in it going back to at least UT3? So no surprise AMD cards will not look so good in this demo.

I changed the demo shortcut> properties> Target to %windir%\system32\cmd.exe /q/c start ElementalDemo.exe -D3D12 -novsync -ResX=1280 -ResY=800

The fps/frametime monitor then stayed green for all but ~2 seconds (in pieces) of the demo.

Yes nvidia is heavily involved with this engine. Thus al the demos on Titans and stuff. And physx folder is in there as well. But Greg tried with physx offloaded to cpu he got same fps as without the forcing, so apparently physx in this demo is not run on gpu.
Though as far as I understand physx when run on system with AMD gpu will use x87 or something like that (I might be wrong). But either way: demo is a demo, AMD wouldn't waste time to optimize drivers for some demo

 

[Mauller]

Yes nvidia is heavily involved with this engine. Thus al the demos on Titans and stuff. And physx folder is in there as well. But Greg tried with physx offloaded to cpu he got same fps as without the forcing, so apparently physx in this demo is not run on gpu.
Though as far as I understand physx when run on system with AMD gpu will use x87 or something like that (I might be wrong). But either way: demo is a demo, AMD wouldn't waste time to optimize drivers for some demo

You don't optimise a game in the drivers with DX12.

The DX12 pathway in UT4 v4.9 is only an early version. most likely developed on nvidia hardware at the moment with minimal testing on AMD hardware.

And since version 3 of physx it has been multi threaded and updated to use SSE on the cpu version. the older versions relied on x87 extensions when using cpu processing.

 

[AthlonXP1800]

I dunno mate it ran fine on dx11 I didn't actually bench it tho just saw it covering around the 80 mark most of the time. Dx12 was smooth for you?

Yes DX12 ran smoothly.

 

[UTmaniac]

Ran this yesterday while running Crossfire (borrowed 7850 out of missus' pc) and despite the "Crossfire active" logo not appearing in the top-right corner (which I had turned on), it made a massive improvement in frametime and framerate running -fullscreen 1080p.

Both GPUs were running at ~99% for most of the demo, which this Q9300 o/ced to 3GHz struggles to do in DX11.

Anyone else similar experience?

 

[UTmaniac]

Ran this demo with my new graphics card and have an odd, repeatable situation...

It runs better in DX12 mode at VSR 3200x1800 @ 60Hz than native 2560x1440 @144Hz, 1080p @144Hz is slightly better still (which is what I expected due to resolution drop).

VSR 1800p reducing the CPU bottleneck compared to 1440p?

But if that is the case, why is 1080p the best of the three resolutions?

The other thing I've noticed is that with the R9 390, the demo usually causes the AMD driver to crash, just after the "ice monster" frees itself from the ice/rocks (demo ran fine on my old R7 265).

 

[humbug]

Ran this demo with my new graphics card and have an odd, repeatable situation...

It runs better in DX12 mode at VSR 3200x1800 @ 60Hz than native 2560x1440 @144Hz, 1080p @144Hz is slightly better still (which is what I expected due to resolution drop).

VSR 1800p reducing the CPU bottleneck compared to 1440p?

But if that is the case, why is 1080p the best of the three resolutions?

The other thing I've noticed is that with the R9 390, the demo usually causes the AMD driver to crash, just after the "ice monster" frees itself from the ice/rocks (demo ran fine on my old R7 265).

If i understand what you are saying correctly, and i'm not sure i do.... then yes at you will get closer to 60 FPS at 1800P that you would 144 FPS at 1080P because the render is slanted more to the GPU than the CPU
"VSR 1800p reducing the CPU bottleneck compared to 1440p?" < indeed.

"But if that is the case, why is 1080p the best of the three resolutions?"
In what way are you saying its best?

Something you may not know, DX12 will dramatically reduce CPU bottle necks, so at 1080P you will get higher frame rates than you would in DX11.