Support NeoGAF

[SimplexPL]

Is the DX12 version in the main branch yet?

You could read the patch notes: "DX12 specific changes (*only applicable to the DX12_preview branch)"

 

[CyberPunked]

You could read the patch notes: "DX12 specific changes (*only applicable to the DX12_preview branch)"

That doesn't tell me that DX12 isn't in the main branch, only that there is a DX12 preview branch.

 

[BoosterDuck]

Is the DX12 version in the main branch yet?

it's not yet

 

[Pagusas]

Really wish we had gotten the HDR update the consoles got (((

 

[dr_rus]

They did:

Was there any improvements in how the game runs on those high end videocards?

 

[scitek]

Really wish we had gotten the HDR update the consoles got (((

Pretty silly if they don't end up bringing it to this game since they're bringing it to Rise of the Tomb Raider.

 

[Javin98]

Tried the benchmark on a new 1080p monitor.

i5-6500
GTX 1060 6GB
16GB RAM

DX11
Fullscreen: On
Exclusive Fullscreen: On
Resolution: 1920x1080
MSAA: Off
Refresh Rate: 60 Hz
VSync: Off
Stereoscopic 3D: Off
Texture Quality: Ultra
Texture Filtering: 16x Anisotropic
Shadow Quality: High
Ambient Occlusion: Very High
Contact Hardening Shadows: On
Parallax Occlusion Mapping: High
Depth of Field: Ultra
Level of Detail: Very High
Volumetric Lighting: On
Screenspace Reflections: On
Temporal Anti-Aliasing: On
Motion Blur: On
Sharpen: On
Bloom: On
Lens Flares: On
Cloth Physics: On
Subsurface Scattering: On
Chromatic Aberration: Off
Tessellation: On

The benchmark results are:

Average: 49.1 FPS
Minimum: 39.6 FPS
Maximum: 61.0 FPS


These results seem appropriate for my hardware?

I don't own the game, but I was following this thread for a while after the game launched and this looks about right. Keep in mind that the benchmarks are significantly more demanding than the gameplay scenarios. So, with an average of 49FPS, you should be able to maintain a solid 60FPS with minor drops in gameplay. That's what I read anyway.

Also, I always keep an eye for your posts because we have the exact CPU and GPU, so I know how a game would run on my PC as well.

 

[jorimt]

Was there any improvements in how the game runs on those high end videocards?

I tested the DX12 benchmark again today (i7-4770k/980 Ti):

Nearly identical DX12 performance on my setup compared to my last test on the 24th (http://www.neogaf.com/forum/showpost.php?p=217978451&postcount=2681), only this time around the sound cut out. I ran the benchmark in DX11, and the sound was back. Ran DX12 bench again just to make sure, and sound cut out yet again.

Oh, another little bonus; DX12 was using 45% of my CPU on average, DX11 was using 70%...

Also, if you look at the registry entry for the game's settings, the DX12 Async option is still at "0" with DX12 enabled. Ah well, maybe there is a boost on some AMD setups.

 

[RedSwirl]

There are some parts in the hub area where my framerate starts to chug. Afterburner tells me it's mostly a CPU bottleneck but GPU usage goes up in these sections too. It's still only using like 6.5 GB of RAM.

 

[dr_rus]

I tested the DX12 benchmark again today (i7-4770k/980 Ti):



Nearly identical DX12 performance on my setup compared to my last test on the 24th (http://www.neogaf.com/forum/showpost.php?p=217978451&postcount=2681), only this time around the sound cut out. I ran the benchmark in DX11, and the sound was back. Ran DX12 bench again just to make sure, and sound cut out yet again.

Oh, another little bonus; DX12 was using 45% of my CPU on average, DX11 was using 70%...

Also, if you look at the registry entry for the game's settings, the DX12 Async option is still at "0" with DX12 enabled. Ah well, maybe there is a boost on some AMD setups.

So they basically decided to not fix the "high end cards" performance in DX12. Shocking.

 

[jorimt]

So they basically decided to not fix the "high end cards" performance in DX12. Shocking.

After investigating DX12 benchmark articles for multiple games, I'm starting to think it's only beneficial on setups that are CPU limited. Once the GPU is the deciding factor, it seems DX12 is the worse performer over DX11 in almost every single scenario. Maybe we've been looking at DX12 wrong, and instead of it being the replacement to DX11, perhaps it's actually an alternative for CPU limited systems, and at best, high-end setups will simply see equal performance to DX11.

That said, my knowledge of DX12's inner working is almost nil, but this lack of performance increase on higher-end setups is perplexing.

 

[dr_rus]

After investigating DX12 benchmark articles for multiple games, I'm starting to think it's only beneficial on setups that are CPU limited. Once the GPU is the deciding factor, it seems DX12 is the worse performer over DX11 in almost every single scenario. Maybe we've been looking at DX12 wrong, and instead of it being the replacement to DX11, perhaps it's actually an alternative for CPU limited systems, and at best, high-end setups will simply see equal performance to DX11.

That said, my knowledge of DX12's inner working is almost nil, but this lack of performance increase on higher-end setups is perplexing.

This is true mostly but it's still a clear case of a developer's inability to write good code when a DX12 renderer is considerably slower than a DX11 one in GPU limited situations. In a perfect case it should be faster and in a regular case it should at least not be slower to any serious degree. When it is slower - it's almost always the DX12 renderer's fault.

 

[jorimt]

This is true mostly but it's still a clear case of a developer's inability to write good code when a DX12 renderer is considerably slower than a DX11 one in GPU limited situations. In a perfect case it should be faster and in a regular case it should at least not be slower to any serious degree. When it is slower - it's almost always the DX12 renderer's fault.

Which brings up another point: driver teams vs. game devs. From my (admittedly limited) understanding, one of the main reasons AMD gets a performance boost in DX12 over DX11 (in the rare instances it does) is because of the fact that AMD's driver team is so bad at squeezing performance "optimizations" (I hate this word now, but it will have to do) out of the DX11 API, at least when compared to Nvidia. Seeing as DX12 "optimizations" are almost solely in the hands of the individual dev, instead of the driver teams, who do only this for a living, so to speak, it's probably safe to say DX12 performance on PC titles is going to vary heavily due to individual dev knowledge of the DX12 API, as opposed to DX11, where most of the performance differences between PC setups are evened out by driver releases.

It seems no matter how we spin it, DX12, and all "to-the-metal" APIs, for that matter, are far more suited to a fixed spec environment, like the consoles, as opposed to PC, where there are too many configuration combinations to count. I don't really have much hope for DX12 on PC at this point, at least if we're expecting to see an average increase in performance across all titles that use it. I see more and more (especially smaller) PC dev teams dropping DX12 in favor of DX11 in the future.

That said, as you say, I have no doubt DX12 can be equal or even (slightly) superior to DX11 in GPU limited situations going forward. It's simply up to the developer now to squeeze it out. And realistically, that's going to be the minority of PC devs.

 

[marcboss]

They improved the DX12 performance with the patchs, on my rig it's almost identical to the DX11 version, but i'm still CPU Limited in some places in Prague, DX11 or DX12...
CPU Usage is always on top, my fps is 100-120 average at Ultra Settings, but goes down to 65-80 in Prague when the CPU hits 95-99% usage, so DX12 on this game is just placebo.

I have a I5 6600K at 4.6 Ghz and GTX 1080 2.1 Ghz.

 

[SimplexPL]

Latest patch adds official DX12 support in the main branch:

http://steamcommunity.com/app/337000/discussions/0/343788552537105000/

Official release of DirectX 12 support in Deus Ex: Mankind Divided:

Added support for DirectX 12.
You can toggle DirectX 12 from either the launcher window or from the Display options within the game.
Note that DirectX 12 requires a restart of the game if (de-)activated from within a running game session. (WARNING: Using the DirectX 12 API can offer better performance on some systems, however, it will not be beneficial on all. If you encounter problems with DirectX 12, we recommend turning it off again. There are no visual or gameplay differences between the two DirectX versions.)
Multi-GPU Note: Currently we only support Single GPU on DirectX 12, see below for more details about Multi-GPU.

Awaiting benchmarks.

 

[jorimt]

Awaiting benchmarks.

I just tested it again, same results as in my last post. I don't think they actually improved DX12 support in this patch, they simply added it to the official build.

 

[cartoon_soldier]

I am installing this on SSD via disk.

Is this for real? 5 hours to install the game?

 

[TaintedHotSauce]

Latest preview has DX12 multi-adapter support

 

[loganclaws]

I am installing this on SSD via disk.

Is this for real? 5 hours to install the game?

Depends how fast your dvd drive is?

 

[SimplexPL]

I am installing this on SSD via disk.

Via what? An actual physical disk?

 

[dmix90]

Latest preview has DX12 multi-adapter support

Native DX12 MGPU or just SLI? I think ROTTR used SLI in DX12 mode or at least when i turned off SLI in NVCP only one GPU worked.

 

[Älg]

I am installing this on SSD via disk.

Is this for real? 5 hours to install the game?

Are you on a data cap? Your download speed is almost ceratainly faster than the read speed of a DVD.

 

[flipswitch]

I am installing this on SSD via disk.

Is this for real? 5 hours to install the game?

I had the same issue when installing Fallout 4 from disc. It was so slow I ended up downloading it instead.


Downloading the new DX:MD patch. It currently says 6.0GB.

 

[dr_rus]

Native DX12 MGPU or just SLI? I think ROTTR used SLI in DX12 mode or at least when i turned off SLI in NVCP only one GPU worked.

All mGPU options in DX12 are native. If you're asking about whether it's LDA or UDA or even implicit then it's LDA most likely as for implicit there would probably be a new entry in DX12 SLI bits in Inspector.

 

[EatChildren]

How is it that nearly two months post release the missing first person hand/gun model ambient occlusions when set to very high hasn't been fixed?

 

[paperspace]

How is it that nearly two months post release the missing first person hand/gun model ambient occlusions when set to very high hasn't been fixed?

Stuttering is still an issue, so I'd imagine they're working on that (at least I hope so), unless they've given up on the PC-version completely.

 

[JaseC]

Stuttering is still an issue, so I'd imagine they're working on that (at least I hope so), unless they've given up on the PC-version completely.

It's still being worked on regularly.

 

[scitek]

How is it that nearly two months post release the missing first person hand/gun model ambient occlusions when set to very high hasn't been fixed?

Stuttering is still an issue, so I'd imagine they're working on that (at least I hope so), unless they've given up on the PC-version completely.

My guess is they had to pull people off this and work on getting the Tomb Raider 20th Anniversary content out the door. Now they can likely start tackling problems with this game again.

 

[jorimt]

"v1.10 build 592.1_P10" patch is out:
http://steamcommunity.com/app/337000/discussions/0/341537671996065340/

We have just released another PC patch for Deus Ex: Mankind Divided, v1.10 build 592.1_P10. This patch focusses on fixes for a couple of gameplay and DirectX 12 related issues raised by the community.

This patch will be applied by Steam automatically when you next start the game. If your game does not update, please restart the Steam client.

The following gameplay changes are in this patch:

• Fixed regression that was introduced in last patch: Walking speed is now fixed properly.
• Fixed regression that was introduced in last patch: Talos Rucker no longer spawns twice.
• AMD intro video now has a lower volume.
• Fixed an issue where clicking on some interactive objects, such as computers or keypads, could cause Jensen to unholster his weapon.
• Fixed an issue where some users had to create a new Windows account to be able to start the game.
• Fixed a crash when loading a savegame.
• Fixed an issue where SSAO could introduce some graphical artifacts.
• Fixed an issue in Breach where the Booster Pack open button would not be shown, even though the user had Booster Packs to open.
• Fixed issues with the dialogue selection wheel.

The following DirectX 12 changes are in this patch:

• Improved stability and performance.
• Major improvement to loading times.

I ran another DX12 benchmark today:

September 8th:
http://www.neogaf.com/forum/showpost.php?p=216273844&postcount=2551

DX11: (AVG) 49.1, (MIN) 39.6, (MAX) 60.7
DX12: (AVG) 46.7, (MIN) 37.6, (MAX) 57.1

September 24th:
http://www.neogaf.com/forum/showpost.php?p=217978451&postcount=2681

DX11: (AVG) 49.4, (MIN) 39.2, (MAX) 61.3
DX12: (AVG) 44.2, (MIN) 36.0, (MAX) 53.7

Today:

DX12: (AVG) 47.3, (MIN) 38.0, (MAX) 58.0

There does appear to be a very slight improvement in DX12 performance on my system over the last update (Sep 24th). It's within margin of error of my first test (Sep 8th) though.

"Very High" vs "On" AO still has less viewmodel coverage, but it looks like the flickering may be resolved (though it seemed resolved to me a couple of updates ago).

 

[paperspace]

I really hope they manage to fix the stuttering soon. The stuttering in Prague is insane, and I just can't bring myself to continue when I know I'll end up with a headache. It's a shame DX12 was a higher priority for them, considering the small (to no) gains.

Even with a 6700K @ 4.5GHz, 32GB 3200MHz, GTX1080, Samsung 950 Pro and G-Sync it's a complete stutter fest.

edit: Unplayable on a fresh install of W10 as well. The upgrade from 4790K, 16GB, 970 & mechanical HDD + fixed refresh rate didn't do much, heh.

 

[TaintedHotSauce]

Looks like the scaling is 100% of DirectX 12 Multi-GPU with two AMD RX 480s.

http://radeon.com/en-us/deus-ex-directx-12-mgpu/

 

[Mindman]

I really hope they manage to fix the stuttering soon. The stuttering in Prague is insane, and I just can't bring myself to continue when I know I'll end up with a headache. It's a shame DX12 was a higher priority for them, considering the small (to no) gains.

Even with a 6700K @ 4.5GHz, 32GB 3200MHz, GTX1080, Samsung 950 Pro and G-Sync it's a complete stutter fest.

edit: Unplayable on a fresh install of W10 as well. The upgrade from 4790K, 16GB, 970 & mechanical HDD + fixed refresh rate didn't do much, heh.

There's real way to fix it, but using borderless window mode fixed most of it for me.

 

[dr_rus]

Looks like the scaling is 100% of DirectX 12 Multi-GPU with two AMD RX 480s.

http://radeon.com/en-us/deus-ex-directx-12-mgpu/

AMD's image.

Real test:

https://www.pcper.com/news/Graphics-Cards/DX12-Multi-GPU-scaling-and-running-Deus-Ex-Mankind-Divided

 

[TaintedHotSauce]

AMD's image.

Real test:

Comparing apples and oranges -- two different resolutions. Scaling is better at 1440p than 1080p.

 

[Durante]

Is the "real test" even running the actual game or the built-in benchmark?

Because we already know that the latter has a really different performance profile from the actual gameplay.

 

[jorimt]

Specs:
OS: Windows 10 64-bit (Anniversary Update)
Nvidia Driver: 375.70
Display: Acer Predator XB271HU
Motherboard: ASRock Z87 Extreme4
Power Supply: EVGA SuperNOVA 750W G2
CPU: i7-4770k @4.2GHz (HT Enabled)
Heatsink: Hyper 212 Evo w/2x Noctua NF-F12 Fans
GPU: GIGABYTE GTX 980 Ti G1 Gaming 6GB
RAM: 16GB G.SKILL Sniper DDR3 @1866MHz
SSD (OS): 256GB Samsung 850 Pro
HDD (Games): 5TB Western Digital Black 7200 RPM/128MB Cache

In-game settings:

For what it's worth, I just ran a quick in-game benchmark (G-Sync enabled) using MSI Afterburner and the FRAPS benchmark tool. I started from an identical save at the front of Palisade Bank, did a single lap around the whole area and back to the front of the bank.

The FRAPS overlay wasn't displaying with DX12, but the benchmark function appeared to work regardless.

Afterburner Graphs (1000 ms Hardware Polling):
DX11 (left), DX12 (right)
FPS (top), Frametime (middle), CPU Usage (bottom)

FRAPS Benchmark DX11:
Frames: 6396 - Time: 89032ms - Avg: 71.839 - Min: 58 - Max: 87

FRAPS Benchmark DX12:
Frames: 6423 - Time: 90359ms - Avg: 71.083 - Min: 56 - Max: 84

Not an exhaustive test by any means, but It looks as if DX12 has a slight frametime edge for the most part, and considerably lower CPU usage, at least in this small (nighttime) section of the game.

All told though, the results aren't that different from my previous built-in benchmark tests:
http://www.neogaf.com/forum/showpost.php?p=220524501&postcount=2749

 

[dr_rus]

Comparing apples and oranges -- two different resolutions. Scaling is better at 1440p than 1080p.

Why is that? Does the top OCed i7 limit two 480s in 1080p in DX12 somehow? This is an average scaling result, no matter how you spin it.

Is the "real test" even running the actual game or the built-in benchmark?

Because we already know that the latter has a really different performance profile from the actual gameplay.

That's a good question as well although I doubt that it will be different for mGPU scaling.

 

[Jtrizzy]

How is this running on a 970? SLI work? How bad is the stuttering?

 

[J_Viper]

So I'm wondering if I should buy this on PC, or grab it now for $30 on PS4

Are there still optimization problems on PC? If I'm rolling with an i5-6600K/1070/16 GBs RAM, could I run this without any issues?

 

[Fledz]

^^^You should be fine.

How is this running on a 970? SLI work? How bad is the stuttering?

Ran fine on a single 970, no idea about SLI. Make sure you have enough RAM otherwise prepare for loooooong load times.

 

[Pop On Arrival]

so I just played this game again after I got my 1070 everything turned up with dx12 on and near jensens apartment i noticed black goop while i was walking back n forth in a specific location. Tried to ignore it, but 20 minutes later when I got off the train to an entirely different disctrict the black artfiact/goop was everywhere.

Restarted the game in DX11 it was gone, but I restarted it again in DX12 in the exact same location i was having severe issues and the black shit over my screen was gone.

First time ever experiencing this in any game.

 

[FutureProtagonist]

AMD's image.

Real test:

https://www.pcper.com/news/Graphics-Cards/DX12-Multi-GPU-scaling-and-running-Deus-Ex-Mankind-Divided

https://www.youtube.com/watch?v=86tp-qRKG58

This is the accompanying video to that review where they test 1440p.

Practically identical to AMD's results, but yes this is the built-in benchmark.

 

[Gitaroo]

Did the dev say if this game will ever get hdr support?

 

[Freshmaker]

so I just played this game again after I got my 1070 everything turned up with dx12 on and near jensens apartment i noticed black goop while i was walking back n forth in a specific location. Tried to ignore it, but 20 minutes later when I got off the train to an entirely different disctrict the black artfiact/goop was everywhere.

Restarted the game in DX11 it was gone, but I restarted it again in DX12 in the exact same location i was having severe issues and the black shit over my screen was gone.

First time ever experiencing this in any game.

No good reason to run in DX12 anyway.

 

[Pop On Arrival]

No good reason to run in DX12 anyway.

really? I dunno, maybe because my first experience with DX12 was ROTR and i had zero issues with that game and it ran perfectly only dipping to high 50s during the geothermal valley in setpiece moments where nothing is going. Also, no artifacs/blackshit over my screen

tl;dr are you saying DX12 is a waste in general or just in this game?

 

[nOoblet16]

Edit: oops wrong thread

 

[Freshmaker]

really? I dunno, maybe because my first experience with DX12 was ROTR and i had zero issues with that game and it ran perfectly only dipping to high 50s during the geothermal valley in setpiece moments where nothing is going. Also, no artifacs/blackshit over my screen

tl;dr are you saying DX12 is a waste in general or just in this game?

In this game with an Nvidia card, there's no reason to use DX 12. 11 performs way better.

 

[SimplexPL]

Also, the latest and best form of ambient occlusion, VXAO, does not work in DX12. Go figure.

I also read somewhere that DX12 mode was actually a wrapper for DX11 which is why there are no gains. Sadly I can't find the source.

 

[DemWalls]

Just got the game, probably gonna install it in a few days, but my rig isn't really cutting edge, so... what options should I set to low (or disable) in order to get decent performance without sacrificing too much in terms of eye candy?

 

[SimplexPL]

What are your specs?

Definitely disable Contact Hardening Shadows and MSAA and go from here. Screen Space Reflection are usually taxing, so you may reduce those. You can also cap to 30fps (see OP).