Unreal Engine 4 - Infiltrator Released for Free!

[Silverforce11]

Comparison benchmark DX11 vs DX12 for 980Ti & Fury X.

https://www.youtube.com/watch?v=llzhKw6-s5A

Note both GPU perform worse in DX12 mode.

Clearly, UE4 is un-optimized, not ready for prime time DX12 usage. No game from it so far has produced amazing visual for the performance, with Ark running horribly on low/medium even on uber setups. This explains why Ark delayed their DX12 patch.

 

[AtenRa]

Comparison benchmark DX11 vs DX12 for 980Ti & Fury X.

https://www.youtube.com/watch?v=llzhKw6-s5A

Note both GPU perform worse in DX12 mode.

Clearly, UE4 is un-optimized, not ready for prime time DX12 usage. No game from it so far has produced amazing visual for the performance, with Ark running horribly on low/medium even on uber setups. This explains why Ark delayed their DX12 patch.

nice thanks,

Took a few pics from the video,

Have a look at the DX-11 the CPU/GPU usage and the fps.
Also worth of notice, the Fury X even when it has lower CPU usage, the CPU temp is always higher than the CPU with the GTX-890Ti.

And at the end of the benchmark, with the the Drone, the GPU usage on the Fury X goes down to 4%.

 

[PhonakV30]

hmm , I see fury X uses at least 600Mb lower than 980TI.looks like 980TI has fixed video memory usage in this demo.
980 Ti = 2585 to 2624Mb
Fury X = 1780 to 2091MB

 

[AtenRa]

hmm , I see fury X uses at least 600Mb lower than 980TI.looks like 980TI has fixed video memory usage in this demo.
980 Ti = 2585 to 2624Mb
Fury X = 1780 to 2091MB

Thats probable because of the larger 6GB of ram, the game just utilize more than 2GB without even need it.

 

[Head1985]

i cant enable console and uncap fps.How enable console?

Edit need switch keyboard to ENG...


system
i7 6700 4.5Ghz
16GB DD4 3060Mhz cl15 1T
GTX970 1500/8000
win10

Fraps log
Avg: 67.715 - Min: 39 - Max: 101

CPU 4.8Ghz and max fps 200
Avg: 69.481 - Min: 41 - Max: 163

 

[zlatan]

Comparison benchmark DX11 vs DX12 for 980Ti & Fury X.

https://www.youtube.com/watch?v=llzhKw6-s5A

Note both GPU perform worse in DX12 mode.

Clearly, UE4 is un-optimized, not ready for prime time DX12 usage. No game from it so far has produced amazing visual for the performance, with Ark running horribly on low/medium even on uber setups. This explains why Ark delayed their DX12 patch.

The main engine version clearly has some problems with D3D12, but there will be an UE4 based D3D12 benchmark in the near future from Microsoft. That will be a good start. I think Microsoft will write back all the optimization to the main engine, so the other guys will get a boost also.

 

[Head1985]

I was about to start a thread about that, but AtenRa beat me to it.

I made a dev build for Infiltrator, which you can uncap with t.maxfps XXX (XXX your prefered max framerate) and then benchmark if you like.

It is 1GB, extracts to 3GBs in its own folder and needs winrar 5.x to decompress correctly.

The included runme.bat file executes the following command.
WindowsNoEditor\InfiltratorDemo.exe -bUseVSync=False -ResX=1920 -ResY=1080 -FullscreenMode=1 -Fullscreen

You can edit it to what you like of course but it would be nice if people could share their benchmark results. Don't forget to uncap it with t.maxfps.

Here are my benches (spicy desktop wallpapers warning).

Infiltrator Demo uncapped 1920x1080 GTX 970 @1.5Ghz Core i5@4.8Ghz - 63fps

Infiltrator Demo uncapped 1920x1080 7950 @1.1Ghz CORE i7-860@4Ghz - 38fps

Infiltrator Demo 1920X1080 5850 @950Mhz Q9550 @4GHz - 17fps

A few notes

- It is obviously quite lighter than the kite demo

- The end scene with the drone is quite cpu limited. People with skylakes and good gpus are kindly requested to post an msi afterburner cpu/gpu/framerate graph, like I do at the end of my videos.

- Was happy to see the ancient 5850 produce something better than a slideshow. The other systems took 2GBs of vram, while my 5850 only had 1GB. This also resulted the overall RAM usage go up by 1GB compared to the other systems. Typical behavior of this system.

If you are interested in cpu limited scenes look here.In some areas there is 60% more Fps with skylake 4.8Ghz.
https://youtu.be/4-qu3P1tKOg?t=147

I recorded it with skylake at 4.8Ghz.
Note:its recorded with shadowplay so in gpu limited situations there is 5-8% fps frop.Cpu limited situations are same.

 

[railven]

I should download it and give it a shot. Got turned off by the texture streaming on the angel statue in one of psolord's videos.

I'd figure by now that wouldn't be an issue anymore, but there it is.

 

[psolord]

If you are interested in cpu limited scenes look here.In some areas there is 60% more Fps with skylake 4.8Ghz.
https://youtu.be/4-qu3P1tKOg?t=147

I recorded it with skylake at 4.8Ghz.
Note:its recorded with shadowplay so in gpu limited situations there is 5-8% fps frop.Cpu limited situations are same.

Great job, thanks.

Here is a somewhat synced side by side link, from your 6700k video vs my 2500k.

Indeed the 6700k seems to do a lot better in this scene.

 

[psolord]

I should download it and give it a shot. Got turned off by the texture streaming on the angel statue in one of psolord's videos.

I'd figure by now that wouldn't be an issue anymore, but there it is.

Your 980Ti with its 6GBs may not show this if it does more precashing.

All my systems did this though and so did Head's. It must be a feature.

 

[tential]

In my unscientific experience, I found DSR to have the higher performance hit but provide the best image quality. It's brute forced AA in the purest sense. Render four times the pixels, grid them to fit the new resolution, then sample every set of four into one pixel with a user adjustable algorithm, and display the image.

If I can run 4.0x DSR and keep frames above 50FPS I'm pretty happy with it usually. I don't do run and gun games anymore so IQ tends to take a priority.

I don't ever use AA. Downsampling? YES! AA? Nope.... not even a chance. IQ takes priority here as well for me.

 

[Pottuvoi]

I don't ever use AA. Downsampling? YES! AA? Nope.... not even a chance. IQ takes priority here as well for me.

Downsampling is also known as Super Sample Anti Aliasing.

Anyway, if you use high enough downsampling you should try to use some post AA with it.
This gives some gradients and improves the quality. (Ingame postAA happens before downsampling, so DS fixes some of the problems with postAA.)

If you dislike wide resolve in TAA used in UE4 downsampling helps as resolve blur gets thinner.
TAA in UE4 is really good to fix specular aliasing and such, it helps the image quality even with high downsampling.

 

[antihelten]

Comparison benchmark DX11 vs DX12 for 980Ti & Fury X.

https://www.youtube.com/watch?v=llzhKw6-s5A

Note both GPU perform worse in DX12 mode.

Clearly, UE4 is un-optimized, not ready for prime time DX12 usage. No game from it so far has produced amazing visual for the performance, with Ark running horribly on low/medium even on uber setups. This explains why Ark delayed their DX12 patch.

Is it just me, or are the various instances of motion blur significantly heavier on AMD?:

Spoiler

meh, probably just an issue with the videos not being properly synced.

 

[soccerballtux]

blur is fps dependent

 

[antihelten]

blur is fps dependent

doh, of course

 

[N7SpectreElite]

Demo looks very good for being created back in 2013!