Deus EX: Mankind Dividied system specs revealed

[Carfax83]

I glanced at the ram graph a few times, didn't notice it go much above 8GB. Didn't play for very long in DX12 mode, but I'll have an experiment with it later on when I get a chance.

Also page file is on my Samsung 950pro.

It won't matter whether the pagefile is on your 950 Pro, as even the fastest SSD is many times slower than your RAM. But if it doesn't hit 10GB or go above, then it could be how Windows is managing your RAM..

What do you have your page file set to?

 

[Carfax83]

Have you tested to see if CHS doesn't break shadow draw distances anymore? I didn't really look for it before, but you were criticizing the game for it. So I'm curious if the patch has fixed it in your estimation.

Nope, haven't tested it yet. Don't know if I even want to, as CHS has a significant performance hit (especially for NVidia cards) for a very marginal IQ gain.. Supposedly the shadow draw distance has been fixed per the latest patch though:

The following fixes are in this patch:
- Fixed a freeze that happened for some players during the subway loading scene.
- Fix for issue where massive stalls would occur in Prague.
- Fixed issue where players could no longer save when they ran out of Steam Cloud space. (When the player runs out of space, the oldest saves will now be removed from the Steam Cloud. They will NOT be deleted and are still available from disk. They will remain present in your in-game ‘load game’ list.
- Improved shadow behaviour in the distance when Contact Hardening Shadows is turned on.
- Fixed issue where tutorial messages could stay/re-appear on-screen after completing the tutorial.
- Fixed an issue where players could lose input control when hacking certain computers.
- Fixed various UI issues.

 

[Carfax83]

Furyx vs 1070 vs 1080
https://www.youtube.com/watch?v=afA9A88hP_M
Furyx is 10% faster than 1070.

That's not really a definitive conclusion if you ask me. Whenever testers use NVidia reference cards, the Fury X looks better than it really is because it has water cooling and as such, is always at it's maximum clock speed. Reference cards on the other hand have significantly worse heat dissipating capabilities than aftermarket versions, and so they perform less than their potential.

If they had used aftermarket parts, it likely would have given the 1070 enough of a boost to match the Fury X..

 

[Deders]

It won't matter whether the pagefile is on your 950 Pro, as even the fastest SSD is many times slower than your RAM. But if it doesn't hit 10GB or go above, then it could be how Windows is managing your RAM..

What do you have your page file set to?

Obviously, but there have been times like when I spread my pagefile over 2 disks in order to make Doom 3 playable at Max settings with only 256MB of ram (required 512 for uncompressed textures) so more bandwidth can help, even if it is low.

Have just loaded up the tf29 headquarters. It didn't take as long as I expected. Prague was much longer. Not much of a difference here.

In dx12, the system ram needed was over 10GB. In DX11 it is just over 6GB.

Vram is just over 7GB, with DX12 it was approaching 8GB.

Page file is set to auto.

 

[Carfax83]

In dx12, the system ram needed was over 10GB. In DX11 it is just over 6GB.

Vram is just over 7GB, with DX12 it was approaching 8GB.

OK that looks about right. And you're still getting more stutter than DX11?

The only other thing I can think of is that it might be your CPU.. If you have your Windows power option set to balanced, and you also have C states enabled in your UEFI bios, then the CPU won't be running at 4.6ghz. It will likely be running at a much lower clock speed, and not only that, it will be changing clock speeds continuously based on the load. Try putting your Windows power option to performance just to test it and see if it helps.. That way your CPU will be running at full bore.

 

[Deders]

I actually do have my CPU set to 4.6 constantly, but that's more to do with keeping the PSU draw within its optimal range (I had been planning for a SLI rig with 2x 250W cards to go with a 34" ultrawide. Looks like I won't need to).

After watching the behaviour and speed of how speedstep reacts compared to say AMD's version, I'm very confident that the only place you are going to see a difference in performance will be when benchmarking. If there is any modern game type load at all on each core, they will all pretty much lock at Max turbo anyway.

In this level the slight stutters when moving weren't really noticeable in DX12. I'll check again when I'm back on the streets.

 

[Azix]

That's not really a definitive conclusion if you ask me. Whenever testers use NVidia reference cards, the Fury X looks better than it really is because it has water cooling and as such, is always at it's maximum clock speed. Reference cards on the other hand have significantly worse heat dissipating capabilities than aftermarket versions, and so they perform less than their potential.

If they had used aftermarket parts, it likely would have given the 1070 enough of a boost to match the Fury X..

this is such a weird argument. The fury X by default comes with water cooling so there is no "looks better than it really is". Its a GPU with water cooling stock. There is no variation on that. There is no other way it actually is and AMD cards do not even have that kind of boost. its 1050 Mhz or bust.

In the 1070s case, reference or aftermarket barely matters. The card in that video is boosting to 1860 minimum from what I saw. An MSI 1070 gaming X on guru3d was boosting to 1974. Just a 6 percent difference. That is not going to make up 10%. Many of even the aftermarket ones will be in that 1800-1900+ range.

 

[Carfax83]

If there is any modern game type load at all on each core, they will all pretty much lock at Max turbo anyway.

Might be for a quad core I guess, but I'm running a hex core and mine rarely ever lock at max turbo..

In this level the slight stutters when moving weren't really noticeable in DX12. I'll check again when I'm back on the streets.

Just did a longer run through of different areas, including the TF29 base and my RAM usage went over 12GB 😀

I predicted a few pages back that DX12 would increase RAM usage substantially..

 

[Carfax83]

this is such a weird argument. The fury X by default comes with water cooling so there is no "looks better than it really is". Its a GPU with water cooling stock. There is no variation on that. There is no other way it actually is and AMD cards do not even have that kind of boost. its 1050 Mhz or bust.

It's not a weird argument. The FuryX is always running at max clock speeds, and it is a horrible overclocker. The NVidia cards on the other hand are good overclockers and are easily capable of squeezing out significantly more performance..

In the 1070s case, reference or aftermarket barely matters. The card in that video is boosting to 1860 minimum from what I saw. An MSI 1070 gaming X on guru3d was boosting to 1974. Just a 6 percent difference. That is not going to make up 10%. Many of even the aftermarket ones will be in that 1800-1900+ range.

I watched the video more closely this time, and the FuryX does not lead the GTX 1070 by 10% the entire time. Sometimes it was more, sometimes it was less, sometimes it was equal with the 1070.. And sometimes the GTX 1070 was leading, ie check 1:12, and in Prague proper.

 

[Deders]

Let's not get partisan about this. I noticed a couple of places the 1070 had a lead like when using smart vision but the Fury was designed to compete with the 980ti which is in the same ballpark as the 1070.

I would have liked to have seen moving performance in the video, but I can see what they chose the parts they showed.

 

[Unreal123]

What PChardware.de and Computerbase has done good is off contract hardening shadow (GPU Open) when they benchmark in gameplay ,which is tanking more then 20% performance and does not do anything for Nvidia cards.

contract hardening shadow off

contract hardening shadow on

People blamed Nvidia for gameworks features but who will blame AMD for this useless feature in this game?

 

[Bacon1]

People blamed Nvidia for gameworks features but who will blame AMD for this useless feature in this game?

It's no different than HFTS in The Division

Huge performance hit for more "correct" shadows. I personally like darker / defined shadows even if they are "wrong".

I don't think either are worth it, but at least AMD's version is cross-vendor and also open source. Its also less of a performance hit (20% vs 25%+) than HFTS.

https://www.computerbase.de/2016-03/the-division-benchmark/

I don't think any major site tested with CHS turned on, most I saw turned it off.

 

[Unreal123]

It's no different than HFTS in The Division

Huge performance hit for more "correct" shadows. I personally like darker / defined shadows even if they are "wrong".

I don't think either are worth it, but at least AMD's version is cross-vendor and also open source. Its also less of a performance hit (20% vs 25%+) than HFTS.

https://www.computerbase.de/2016-03/the-division-benchmark/

I don't think any major site tested with CHS turned on, most I saw turned it off.

Everyone is doing when they set benchmarks to Ultra. All youtube benchmarks, Guru3d etc. If you off it then it is not Ultra settings and you check most recent video on youtube benchmark ,which has contract hardening shadow on. However, their is too much double standard in forums ,therefore it is useless to argue because AMD can never be wrong.

 

[Bacon1]

Everyone is doing when they set benchmarks to Ultra. All youtube benchmarks, Guru3d etc. If you off it then it is not Ultra settings and you check most recent video on youtube benchmark ,which has contract hardening shadow on. However, their is too much double standard in forums ,therefore it is useless to argue because AMD can never be wrong.

Guru3d used High settings not Ultra.

http://www.guru3d.com/articles_page..._graphics_performance_benchmark_review,8.html

PCGH didn't use Ultra either:

http://www.pcgameshardware.de/Deus-...ews/DirectX-12-Patch-Beta-Benchmarks-1207199/

High / Very High settings.

Some reviewers use gameworks / "max" settings and others don't. Thats why you have to look at more than just one source. I'd turn down a lot of the settings for minimal IQ loss but massive FPS gain myself. I wouldn't use CHS or HFTS (if it was available) because they have a huge fps hit for minimal IQ improvement.

How is that a double standard exactly? These features are completely optional, they aren't required by the engine.

 

[Bacon1]

https://www.computerbase.de/2016-08/deus-ex-mankind-divided-benchmark/2/

The performance hit for each setting is basically even between the 1060 and 480. Each losing ~25% between Ultra -> VH and then ~20% for each next preset and ~16% for low.

 

[Unreal123]

Guru3d used High settings not Ultra.

http://www.guru3d.com/articles_page..._graphics_performance_benchmark_review,8.html

PCGH didn't use Ultra either:

http://www.pcgameshardware.de/Deus-...ews/DirectX-12-Patch-Beta-Benchmarks-1207199/

High / Very High settings.

Some reviewers use gameworks / "max" settings and others don't. Thats why you have to look at more than just one source. I'd turn down a lot of the settings for minimal IQ loss but massive FPS gain myself. I wouldn't use CHS or HFTS (if it was available) because they have a huge fps hit for minimal IQ improvement.

How is that a double standard exactly? These features are completely optional, they aren't required by the engine.

I say one last time that i am talking about "Ultra settings". You can check any youtube video or benchmarks expect Pchardware.de and Computerbase ,which they have put contract hardening shadow on in gameplay or in build benchmark.

 

[Bacon1]

I say one last time that i am talking about "Ultra settings". You can check any youtube video or benchmarks expect Pchardware.de and Computerbase ,which they have put contract hardening shadow on in gameplay or in build benchmark.

I don't care about random's on youtube. The major sites aren't using CHS.

Digital Foundry also isn't using Ultra - https://youtu.be/oNF8c9y9GxE

Again its optional and has similar performance hit on both AMD and NVidia hardware, so it doesn't really matter except that its excessive FPS hit for minimal IQ gain.

Did you complain this much about Hyper settings in Mirror's Edge, or Ultra settings in The Division / ROTTR?

All most all AMD fanboys debunk gameworks benchmark just like you do ,therefore, tell what Nvidia user should do here?

I have no idea what this sentence means.

 

[Unreal123]

I don't care about random's on youtube. The major sites aren't using CHS.

Digital Foundry also isn't using Ultra - https://youtu.be/oNF8c9y9GxE

Again its optional and has similar performance hit on both AMD and NVidia hardware, so it doesn't really matter except that its excessive FPS hit for minimal IQ gain.

Did you complain this much about Hyper settings in Mirror's Edge, or Ultra settings in The Division / ROTTR?



I have no idea what this sentence means.

I mean that any site ,which are using gameworks features in benchmark are called shills or payed by Nvidia. They even say gimpworks.

 

[Bacon1]

https://www.youtube.com/watch?v=OhrPJWqMTzU

Guessing the right side is DX12 since it handles CPU load much better. But there is a guy with a 1060 and i5 2500k where its pretty much unplayable in some cases on the left side (15 fps with major stuttering in fights) vs 40s on the other side.

 

[Unreal123]

https://www.youtube.com/watch?v=OhrPJWqMTzU

Guessing the right side is DX12 since it handles CPU load much better. But there is a guy with a 1060 and i5 2500k where its pretty much unplayable in some cases on the left side (15 fps with major stuttering in fights) vs 40s on the other side.

Nixxes already replied yesterday about memory leaks, Ram usage, stuttering and GPU usage. They only said that it is not even beta and it just a preview version at the moment and their will be bug fixes.

 

[AtenRa]

Do you guys play the game with MSI Afterburner enabled in DX-12 ??? if yes , close MSI Afterburner and try again, might be this one that causes the stuttering problem.

 

[Unreal123]

Do you guys play the game with MSI Afterburner enabled in DX-12 ??? if yes , close MSI Afterburner and try again, might be this one that causes the stuttering problem.

Ok i give a try.

 

[Bacon1]

I mean that any site ,which are using gameworks features in benchmark are called shills or payed by Nvidia. They even say gimpworks.

Source?

Also the difference between Gameworks and GPUOpen is well... Closed vs Open Source and freely modifiable, and GPUOpen effects tend to have equal performance hit on both sides, vs Gameworks effects impacting Nvidia far less.

980 went from 76 -> 56 fps. 26% perf hit
290 went from 63 -> 36 fps. 43% perf hit.
780 went from 49 -> 36 fps. 26% perf hit.

26% vs 43% in Hairworks at launch.

Lets look at the original TressFX

980 went from 98 -> 69 fps. 29% perf hit.
290 went from 73 -> 49 fps. 32% perf hit.
780 went from 78 -> 51 fps. 34% perf hit.

So 3% faster to 2% slower than AMD using Nvidia hardware, vs 17% slower using hairworks on AMD vs Nvidia.

And thats averages, minimums were destroyed completely.

The R9 290X enjoyed a massive 75% performance jump with HairWorks disabled, making it 17% slower than the GTX 980 -- it all started to make sense then.

We should reiterate that besides disabling HairWorks, all other settings were left at the Ultra preset.

http://www.techspot.com/review/1006-the-witcher-3-benchmarks/page6.html

Here is what the Witcher developers CDPR said regarding Hairworks:

Many of you have asked us if AMD Radeon GPUs would be able to run NVIDIA’s HairWorks technology – the answer is yes! However, unsatisfactory performance may be experienced as the code of this feature cannot be optimized for AMD products. Radeon users are encouraged to disable NVIDIA HairWorks if the performance is below expectations."

http://www.overclock3d.net/news/gpu...nvidia_hairworks_unoptimizable_for_amd_gpus/1

AMD was able to fix it by forcing it to only render @ 16x tessellation levels not 64x, but there is a major reason why its called Gimpworks.

Also many people here are calling Gaming Evolved Gaming Devolved so its not like its one sided.

 

[Bacon1]

Nixxes already replied yesterday about memory leaks, Ram usage, stuttering and GPU usage. They only said that it is not even beta and it just a preview version at the moment and their will be bug fixes.

I was assuming that the better version was the DX12 version, since he has an old i5 2500k and the right side was using all threads @ 100%. Would be nice if people labeled their videos.

 

[Deders]

Might be for a quad core I guess, but I'm running a hex core and mine rarely ever lock at max turbo..

Just did a longer run through of different areas, including the TF29 base and my RAM usage went over 12GB 😀

When you had speedstep enabled, what was the CPU% set to in the power profile? I used to have it at 5%. Sometimes windows sets it differently. Maybe a number in between that and 100% would give a greater variety of frequencies.

I'll check again when I'm back on the streets.

Had a run through the streets and found that it did have ever so slight judders every now and then. Mostly near crossroads. Disabling CHS made everything really smooth.