Jump to content
Search In
  • More options...
Find results that contain...
Find results in...
byalexandr

[Unreal Engine 4] DirectX12 vs. DirectX11 Performance Comparison

Recommended Posts

Posted · Original PosterOP

I'll go ahead and post Edit #2 right here, this is probably what you wanted to see anyway:

 

I'm back with very promising results!

 

I downloaded the original version of the Elemental Tech Demo which runs on DX11, as well as the 'Improved Version' that runs on DX12. After verifying the my system was on the DirectX 12 API, I proceeded to benchmark both using FRAPS. Here are the results, I compiled them in a slightly terribly made chart:

 

post-2597-0-34758000-1439508503_thumb.jp

 

These tests were performed on the system in my signature. Your mileage may vary but I do see a noticeable improvement, although I'm sure the max FPS in the DX12 benchmark was a hiccup (probably towards at the end when the Unreal logo shows up, although it was significantly higher than the DX11 Max FPS).

 

-BACK TO THE MAIN PART-

 

So this is a continuation of another thread:

 

 

Heyyo,

So... I tried searching to see if anyone already posted this? But I can't seem to find any so... lemme know if I make a boo-boo please.  :P


http://hexus.net/tech/news/graphics/85385-dx12-unreal-engine-4-elemental-demo-download-available/

Someone ported that fancy Elemental demo of Unreal Engine 4 to work on the DirectX 12 version of Unreal Engine 4.9. I'm definitely gonna try this stuff out at home and try comparing it to the DirectX 11 version.

Here's where to download the original DirectX 11 version:
http://www.techpowerup.com/downloads/2368/unreal-engine-4-elemental-tech-demo/


Unreal Editor v4.9 also lets you run it in DirectX 12 mode:
https://forums.unrealengine.com/showthread.php?79514-Unreal-Engine-4-9-0-amp-Windows-10-How-do-i-start-up-my-project-using-direct-x12

So I'm gonna try that stuff too and see if I can get Unreal Tournament to work in DirectX 12... it'll be interesting to see!


NVIDIA also just released a newer driver today v355.60. It talks about that Ashes of Singularity which is also a DirectX 12 game. Maybe it'll help too with Unreal Engine 4.9?

http://www.geforce.com/whats-new/articles/geforce-355-60-whql-driver-released




Dreams do come true!?  ;)

 

 

 

THIS IS MY PERSONAL OVERVIEW AND HOW DX12 IS DOING FOR ME. I HAVE AVERAGE HARDWARE, COMPONENTS MOST MID RANGE BUILDS CONTAIN.

 

So how is DX12 so far? Well for me, it's probably disappointing if I'm doing this right at all. If you read above the Elemental Demo runs DX12 on Unreal version 4.9.0. Well I have done the following and you can see what happened:

 

So my results were disappointing, although I'm sure I did it wrong. After verifying in dxdiag that I was running DX12, I created a batch file that contained the following string:

 

ElementalDemo.exe -D3D12

 

Before I ran the batch file I ran the application by itself, and I assumed it was running in DX11 mode as my CPU usage was ~40% and my GPU was pinned at 100% (specs in signature). I then ran the batch file (which means it's running in DX12) and immediately noticed an FPS drop by an average of 10FPS. The CPU usage dropped to ~20% and the Elemental Demo crashed within a minute or two. I still have more testing and more figuring out to do but so far it's a failure for me.

 

So far it's a mixed bag, I have obviously gotten worse performance even though I am on the latest drivers (355.60) but a lot of others have gotten better overall FPS.

 

Edit #1 of many: I am downloading the original DX11 version of Elemental and I will try and run it normally at first then using the Unreal version 4.9.0 in DX12 mode. Hopefully the comparisons yield better results than my first attempt.

 

WHAT DO YOU SUGGEST I SHOULD DO? LEAVE A REPLY IF YOU'RE INTERESTED. I WILL CONTINUE TO WORK ON THIS POST AS I GAIN MORE EXPERIENCE AND TRIAL AND ERROR.


INTEL CORE i7 4820K | COOLER MASTER HYPER 212 EVO | GIAGBYTE X79-UP4 | CRUCIAL BALLISTIX SPORT 8GB | NVIDIA GEFORCE GTX 1070 | SAMSUNG 850 EVO 500GB | WESTERN DIGITAL CAVIAR BLUE 1TB | CORSAIR CRYSTAL SERIES 570X | EVGA BQ 650W

Link to post
Share on other sites

In this benchmark, there is a 125% fps increase. :o


Main gaming pc: Phanteks Eclipse P400S (RGB) - EVGA Nvidia GeForce GTX 1070 8GB FTW - i7 4790K - Cooler Master Hyper 212 EVO - Gigabyte Z97P-D3 - ADATA XPG v2 8GB RAM + 16GB HyperX Fury 1600Mhz  - Kingston HyperX 240GB SSD - be quiet! Pure Power 9 600Watt - Seagate 1TB HDD - Seagate 2TB HDD

 

Gaming laptop: ASUS GL552VW: i7 6700HQ - GTX960M 2GB - 8GB DDR4 2166Mhz RAM - 1TB 7200RPM HDD

Link to post
Share on other sites
Posted · Original PosterOP

Sorry about the formatting. :/


INTEL CORE i7 4820K | COOLER MASTER HYPER 212 EVO | GIAGBYTE X79-UP4 | CRUCIAL BALLISTIX SPORT 8GB | NVIDIA GEFORCE GTX 1070 | SAMSUNG 850 EVO 500GB | WESTERN DIGITAL CAVIAR BLUE 1TB | CORSAIR CRYSTAL SERIES 570X | EVGA BQ 650W

Link to post
Share on other sites
Posted · Original PosterOP

In this benchmark, there is a 125% fps increase. :o

If you read my experience you will see that I had a decrease in performance even though I assume it was running in DX12.


INTEL CORE i7 4820K | COOLER MASTER HYPER 212 EVO | GIAGBYTE X79-UP4 | CRUCIAL BALLISTIX SPORT 8GB | NVIDIA GEFORCE GTX 1070 | SAMSUNG 850 EVO 500GB | WESTERN DIGITAL CAVIAR BLUE 1TB | CORSAIR CRYSTAL SERIES 570X | EVGA BQ 650W

Link to post
Share on other sites
Posted · Original PosterOP

I think I should download the DX11 version of the Elemental Demo and then open it in 4.9.0 to run it in DX12. Shout out to @ThE_MarD


INTEL CORE i7 4820K | COOLER MASTER HYPER 212 EVO | GIAGBYTE X79-UP4 | CRUCIAL BALLISTIX SPORT 8GB | NVIDIA GEFORCE GTX 1070 | SAMSUNG 850 EVO 500GB | WESTERN DIGITAL CAVIAR BLUE 1TB | CORSAIR CRYSTAL SERIES 570X | EVGA BQ 650W

Link to post
Share on other sites
Posted · Original PosterOP

I'm back with very promising results!

 

I downloaded the original version of the Elemental Tech Demo which runs on DX11, as well as the 'Improved Version' that runs on DX12. After verifying the my system was on the DirectX 12 API, I proceeded to benchmark both using FRAPS. Here are the results, I compiled them in a slightly terribly made chart:

 

post-2597-0-34758000-1439508503_thumb.jp

 

These tests were performed on the system in my signature. Your mileage may vary but I do see a noticeable improvement, although I'm sure the max FPS in the DX12 benchmark was a hiccup (probably towards at the end when the Unreal logo shows up, although it was significantly higher than the DX11 Max FPS). I will post this in my OP as well so the chart is the main thing you see when you open this thread. Let me know what you guys think and if I should do more testing in other UE4 benchmarks. Thanks!

post-2597-0-34758000-1439508503_thumb.jp


INTEL CORE i7 4820K | COOLER MASTER HYPER 212 EVO | GIAGBYTE X79-UP4 | CRUCIAL BALLISTIX SPORT 8GB | NVIDIA GEFORCE GTX 1070 | SAMSUNG 850 EVO 500GB | WESTERN DIGITAL CAVIAR BLUE 1TB | CORSAIR CRYSTAL SERIES 570X | EVGA BQ 650W

Link to post
Share on other sites

For anyone who still thinks DX12 is going to improve performance, it won't. It'll just allow devs to have more objects on the screen, so we won't really see more fps in upcoming DX12 games.

We will be able to see performance improvement if some current CPU intensive games get DX12 update.

 

And you still have to have a powerful CPU to make use of its main feature, the CPU overhead reduction. The issue with this demo and API overhead test is that they don't use the CPU the way a game would, but rather focus on using the cores for submitting as many draw calls as possible. If you try Mantle in BF4 which does the same thing DX12 does, you won't see much improvement with 4 core CPUs, if any at all. This is because the CPU is busy with everything else going on, and is unable to make more draw calls. Low overhead API is useless in that case. 

You'll only see increase in performance compared to DX11 if you have extra CPU and GPU power, that's not used due to the fact that only 1 core can talk to the GPU and that's where the bottleneck is. So you'll get more fps only if the draw calls are the bottleneck.


i7 9700K, Gigabyte RTX 2070 Gaming OC, Gigabyte Z390 Gaming SLI, 2x8 HyperX Predator 3200 MHz

Link to post
Share on other sites
Posted · Original PosterOP

For anyone who still thinks DX12 is going to improve performance, it won't. It'll just allow devs to have more objects on the screen, so we won't really see more fps in upcoming DX12 games.

We will be able to see performance improvement if some current CPU intensive games get DX12 update.

 

And you still have to have a powerful CPU to make use of its main feature, the CPU overhead reduction. The issue with this demo and API overhead test is that they don't use the CPU the way a game would, but rather focus on using the cores for submitting as many draw calls as possible. If you try Mantle in BF4 which does the same thing DX12 does, you won't see much improvement with 4 core CPUs, if any at all. This is because the CPU is busy with everything else going on, and is unable to make more draw calls. Low overhead API is useless in that case. 

You'll only see increase in performance compared to DX11 if you have extra CPU and GPU power, that's not used due to the fact that only 1 core can talk to the GPU and that's where the bottleneck is. So you'll get more fps only if the draw calls are the bottleneck.

I know, but you see that having a more powerful CPU is actually going to make performance better because of the overhead. The only reason AMD released Mantle is to push the CPU market in the only way they [really] can, since they can't make powerful enough CPUs (or at least don't want to). This pushed the release of DX12 earlier and I hope Intel sees the potential and finally stuffs in more than four cores in their mainstream processors and actually drive the market forward since AMD isn't going to. I hope to see developers really push the limits of this new API, even though Microsoft or whoever the hell makes DirectX could have spent a little more time on it. But that probably won't happen seeing as these fangled 'new' consoles cause the graphical potential to be toned down just so it can run on those stupid things and the developers can make more money lol.


INTEL CORE i7 4820K | COOLER MASTER HYPER 212 EVO | GIAGBYTE X79-UP4 | CRUCIAL BALLISTIX SPORT 8GB | NVIDIA GEFORCE GTX 1070 | SAMSUNG 850 EVO 500GB | WESTERN DIGITAL CAVIAR BLUE 1TB | CORSAIR CRYSTAL SERIES 570X | EVGA BQ 650W

Link to post
Share on other sites

Will this mean that a hexacore CPU will be the entry-level for the gaming platform in the future?


X99 Deluxe | i7 5820K | NH-D15 | 980Ti HOF 8PACK Edition | G.Skill Ripjaws 4 16gb | Samsung 850 EVO 500Gb
Corsair Air 540 | Bose C5 |    ( ง '̀-'́) ง Hall Of Fame Owner's Club ☜(゚ヮ゚☜)    Logitech 502 | ROG Swift PG279Q

Sennheiser HD 518 |  ModMic 4.0 |  Logitech G810 Orion

 

Link to post
Share on other sites
Posted · Original PosterOP

Will this mean that a hexacore CPU will be the entry-level for the gaming platform in the future?

Possibly. Definitely not something like an FX 6300, as @Monarch stated this only really increases performance on high end components. That means something like a 5820K would last a long time, especially since it's already a high end processor.


INTEL CORE i7 4820K | COOLER MASTER HYPER 212 EVO | GIAGBYTE X79-UP4 | CRUCIAL BALLISTIX SPORT 8GB | NVIDIA GEFORCE GTX 1070 | SAMSUNG 850 EVO 500GB | WESTERN DIGITAL CAVIAR BLUE 1TB | CORSAIR CRYSTAL SERIES 570X | EVGA BQ 650W

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now


×