DX12 comparison

Dx 12 shortcut doesn't seem to work
I made a shortcut to the dx12 Civ6.exe, but when I use it the game launches in dx 11 mode. I confirmed it by running Win 10's resource monitor which shows the regular Civ 6 exe running. When I launch it from Steam the resource monitor shows the dx 12 exe. Maybe just some weirdness with my computer, but strange nontheless.

I have the same problem, you would think a shortcut to the dx12 exe would work but it doesn't it seems

CS
 
I have a feeling that DX12 is causing me some very strange long turn times near the start of a game for some reason. Didn't notice this with DX11 on my first game which went on far longer than this one with no problems. Problem appeared to be barbarians AI taking forever to work out. I actually thought it crashed a couple of times.
 
DX 12 is slower on my machine. And it was supposed to be faster.
 
Hi guys! I signed up just to participate in this thread. I might hang around for the rest of the forum though.

Wow, is my FX6300 really that much slower?

Yes, yes it is.

Dx 12 shortcut doesn't seem to work
I made a shortcut to the dx12 Civ6.exe, but when I use it the game launches in dx 11 mode. I confirmed it by running Win 10's resource monitor which shows the regular Civ 6 exe running. When I launch it from Steam the resource monitor shows the dx 12 exe. Maybe just some weirdness with my computer, but strange nontheless.

This is a holdover from Beyond Earth. I contacted Dan Baker about this a while ago and it's something they're aware of, but they never got around to fixing it because they expect everyone to use the Steam menu that pops up when you launch the game. What's funny is that the Beyond Earth benchmark allowed me to call "-mantle" in a command line, but the game wouldn't launch with that argument even when it's put into Steam. It is an option, and it exists in the documentation that I can dig out from the .exe file, but it never works. Steam also directly called the mantle executable, but I couldn't do the same via command line.

I hope they don't abandon Civ VI's benchmark like they did for Beyond Earth.
 
Whether you see improvement in DX12 will depend on your specific GPU and CPU. Here are the benchmarks I ran. (with updated AMD and nvidia drivers)

Specs are:
CPU: X5670 @ 4.4Ghz 6-core w/ HT (hyperthreading)
GPU: GTX 1070
1600x900 (GPU never exceeded 70% utilization)

I ran the benchmarks in 4-core and 6-core mode, with hyperthreading on and off for comparison:

DX11 Graphics Benchmark:
4C/4HT: 15.1ms (66 fps)
4C/8HT: 12.8ms (78 fps)
6C/6HT: 12.7ms (79 fps)
6C/12HT: 12.8ms (78 fps)

DX12 Graphics Benchmark:
4C/4HT: 14.3ms (70 fps)
4C/8HT: 12.6ms (79 fps)
6C/6HT: 12.1ms (83 fps)
6C/12HT: 11.8ms (85 fps)

I ran the game at low-res to measure the *CPU* impact of switching to DX12. My results show that if you are *CPU-limited*, DX12 may be an improvement.

If you're running a modern CPU at 4GHz or more, you probably won't see improvement. If you're GPU-limited, you probably won't see improvement.

If you're running Win7, there's no point in upgrading IMHO. The improvements in framerate were minor, and would only apply to certain people.

EDIT: For those who think it's BS that a game would run slower in DX12, keep in mind the following:

* The DX11 drivers are *very* mature
* DX12, at best, will only show minor GPU performance improvements, unless you're on AMD
* Even on AMD, significant speedup occurs *only if the game takes advantage of Async Compute*
* Companies have to write *different code* for DX12 than they do for DX11

The last one is the key. First-pass implementations of a new version of DX usually aren't faster. DX12 is especially challenging to implement, and even harder to retrofit if your engine wasn't designed for it.

The bad news is, clearly the engine wasn't designed with DX12 in mind. But that's understandable, given only 50% of gamers run Windows 10, and almost 100% of gamers have a DX-11 compatible setup.

The good news is, over time Civ6 will probably run as fast or better in DX12 no matter what setup you have. If you're lucky, they'll even optimize the game for it, and if/when they do, the game will run noticeably better than DX11 on certain hardware.

EDIT 2: Forgot to include my AI results...

AI benchmark
4C/4HT: 18.3 seconds/turn
4C/8HT: 19.1 seconds/turn
6C/6HT: 18.2 seconds/turn
6C/12HT: 19.5 seconds/turn DX11 (17.6 in strategic view)

Bear in mind that the AI results right now are sort of meaningless. The benchmark are still influenced by graphics, which is silly, and "Quick Combat" and "Quick Movement" are just sped up, not instantaneous, so there's a minimum turn time right now no matter how fast your machine is. :(
 
Last edited:
I'm running a GTX 980M and I lose about 1.5 fps using Dx12, nothing big or noticeable in a game, but it does look better.

CS
 
there's deff a difference the resource icons are different and the game looks sharper. pretty much the same performance i was getting b4

evga 980 ti
4ghz
16gb ddr 1600 ram
win 10 64x
running at 2k 165hz
 
Here's my benchmark results on my PC. CPU is an i7-2600K @ 4.6GHz and GPU is GTX 1080 @ 1925 GPU / 1350 Mem nVidia Driver version 375.95 All graphics options set to the highest possible, running at 2560 x 1600 res

DX11:
Avg. Frame Time: 12.848ms
99th Percentile: 19.112ms
Average Turn Time: 18.98 secs

DX12:
Avg. Frame Time: 14.558ms
99th Percentile: 19.787ms
Average Turn Time: 20.39 secs

Clearly a drop in performance with DX12, not sure if there's any improvement in visual fidelity.
 
Seems like people with less than 6 cores are reporting DX12 is worse, whereas for me it performed better. Anyone out there with an 8-core Haswell-E or something that can run the benchmarks and give us more data?
 
Seems like people with less than 6 cores are reporting DX12 is worse, whereas for me it performed better. Anyone out there with an 8-core Haswell-E or something that can run the benchmarks and give us more data?

Yeah, on my 6-core 3930K the benchmark went from just over 34 fps to almost 45 fps @ 4K on a Radeon card. I think you also notice the difference more at higher resolutions like mine.

The other good side effect of the new DX12 driver was that I can now get the Task Manager properly.
 
IDK what you guys are talking about I have better performance from DX12 than with DX11.

DX11
30-32FPS max settings
35FPS Med Settings
40FPS Low Settings

DX12
60-80FPS Max Settings
75-80 FPS Med Settings
100 FPS Low Settings

PC Setup:
AMD FX 9570 8 core Turbo OFF base Clock 4.7Ghz
16GB DDR3 RAM
ASUS Crosshair Formula V Mother board
2x R9 295x2 @ Base settings in Crossfire ( get the same FPS in single Card Config also)
Benckmark ran at 1080P

My only downside is that the game crashes randomly after about 1-2hours of playing. It varies widely though. I just wish I could start the game in DX12 mode through the shortcut instead of the pop-up menu from steam.

P.S. CPU and GPU are water cooled.
 
i5-2500k 390X 8GB, 8GB RAM
  • 3440x1440 borderless window
  • all advanced options on/max
DX11
  • Graphics: 99th 29.817ms
  • AI: 19.14
DX12
  • Graphics: 99th 28.436ms
  • AI: 18.89
About 5% gain with DX12 on my rig
 
Seems to be little to no difference in performance with DX11 and DX12 other than slightly more fluctuations in frame rate in DX12.

At 3840x2160 and all settings on maximum I get a frame time average of 16.6ms with a max of 17.8ms (seems to hover pretty much at 60 fps). There's one spike in DX11 but 2 or 3 in DX12.

System specs:

I7 4790K 4.0GHz
16GB DDR3 1600MHz
GTX 1080 2050MHz Core 10500MHz RAM

Updated results
3840x2160 (4k) max settings
GFX Bench
DX11
Avg Frame Time - 13.681
99th Percentile - 21.288
DX12
Avg Frame Time - 15.397
99th Percentile - 19.159

DX12 appears marginally more even in performance overall, but DX11 gives a higher frame rate (73fps vs 65fps).

7680x2160 (Multi-display)
DX11
Avg Frame Time - 17.634 (56.7 fps)
99th Percentile - 21.967
DX12
Avg Frame Time - 19.832 (50.4 fps)
99th Percentile - 24.852
 
Last edited:
My only downside is that the game crashes randomly after about 1-2hours of playing. It varies widely though. I just wish I could start the game in DX12 mode through the shortcut instead of the pop-up menu from steam.

Put this into the launch window in Steam instead (Steam properties - general tab):

"D:\Steam\steamapps\common\Sid Meier's Civilization VI\Base\Binaries\Win64Steam\CivilizationVI_DX12.exe" %command%

It launches the Dx12 version everytime from any shortcut placed on your desktop. Change the line to reflect your own Steam location of course.
 
....nnnnnnnecro post.

But I'll agree with an earlier post in the thread: The only reason I use the DX12 version is because it eliminates the weird flicker I get on the DX11 opening menu.
 
Since we are reviving the dead. Do we have some more comparisons using both a multi core (6+ thread) and gtx 1060 equivalent card or higher?

This thread was a year old and we now have more multi core options and many more gpus.

I am curious if DX12 performance has improved over DX11 to justify it permanently.

Ill also admit that I am going to move from an I5 3570K 4cores no extra threads to an I8700K (6+6) and think we may see a good enough in performance gains to justify staying in DX12.
I was going to hold off the expac until rig is built. Once done as a hard core Civ Fantatic Ill get the rest of the DLC and the expac and game on..
 
Last edited by a moderator:
I upgraded since last time I benched this game. My current system now is:

8700K @ 5GHz
16GB DDR4 3200MHz CL14
2xGigabyte 1080ti Aorus Extreme SLI

At 4K and 1080P my scores for the AI benchmark times are the same 12.11s in DX11. In DX12 my AI times drop to 12.60s. As for the GPU tests, I find that DX12 is significantly slower than DX11, both single card and even moreso in SLI/multi-gpu. All tests run with custom maximum settings.

Single GPU 3840x2160
DX11 - 92.09fps, 64.00fps 99th Percentile
DX12 - 78.55fps, 62.36fps 99th Percentile

SLI/Multi-GPU 3840x2160
DX11 - 117.79fps, 92.28fps 99th Percentile
DX12 - 94.05fps, 75.17fps 99th Percentile
 
Back
Top Bottom