DX12 comes to The Division

SiDeards73 · 9 Nov 2016 at 06:12

Seems The Division just added a DX12 renderer on their PTR build for their latest build, it was omitted from the patch notes but it is there.

Will be interesting to see how it develops

peterhc · 9 Nov 2016 at 22:30

Lets hope it improves performance

, new patch is not that bad for lonewolfs !
Any idea when patch with dx12 drops ?

Twirlywoos · 10 Nov 2016 at 00:16

Didn't do anything for me

https://forums.overclockers.co.uk/showpost.php?p=30196651&postcount=2778

Dicehunter · 10 Nov 2016 at 02:09

Depends on the scene but it can have a gain -

Running with on a GTX 1080 running at 2.1GHz on the core, 11GHz on the memory coupled with a 5930K running at 4.50GHz, 32GB of memory running at 3000MHz and the game is on a Corsair 960GB XTi SSD, Run at the games Ultra preset at 1080P.

SiDeards73 · 10 Nov 2016 at 07:00

Twirlywoos said:
Didn't do anything for me

https://forums.overclockers.co.uk/showpost.php?p=30196651&postcount=2778

Posted in that thread Twirly, but for clarity your running a decent OC on 3770k and a decent Nvidia GPU, if you had a lesser AMD card and an AMD CPU you would see better gains, this iteration of DX12 they are using seems to be cutting CPU overheads pretty well.

Also Maxwell does not have proper Async support and if they have used any of that in their iteration of DX12 you wont see much gains either.

Personally i dont think they have done much with the renderer so far hence not even mentioned it yet. I expect once its implemented fully most people will get gains of some sort or other.

peterhc · 10 Nov 2016 at 07:43

So if I'm running same cpu i73770k at 4.4 with 980 I wont see any improvements ?

Yaayuh! · 10 Nov 2016 at 08:08

Cool, so we can now grind the same **** and get demolished by hackers in DZ with a couple more FPS maybe

SiDeards73 · 10 Nov 2016 at 09:34

peterhc said:
So if I'm running same cpu i73770k at 4.4 with 980 I wont see any improvements ?

Too early to really say what gains are possible, seen mixed reviews, someone with a 1070 and a decent cpu got decent gains, others with decent cpu and nvidia cards got nothing.

Nexus got roughly 10% uplift on his cpu and R9 290, cant remember his CPU but it was Intel, i think an early i7.

Twirlywoos · 10 Nov 2016 at 10:23

SiDeards73 said:
Posted in that thread Twirly, but for clarity your running a decent OC on 3770k and a decent Nvidia GPU, if you had a lesser AMD card and an AMD CPU you would see better gains, this iteration of DX12 they are using seems to be cutting CPU overheads pretty well.

Also Maxwell does not have proper Async support and if they have used any of that in their iteration of DX12 you wont see much gains either.

Personally i dont think they have done much with the renderer so far hence not even mentioned it yet. I expect once its implemented fully most people will get gains of some sort or other.

Strange how some people get gains and I am not. Would love to know what setups people have when running these tests.

Gamble has 980ti SLI and a 5280K I think (not sure about his OCs or gaming resolution/graphical settings) and got gains.

Gregster · 10 Nov 2016 at 10:27

First time I have seen DX12 acttually give me (as a current NVidia user) any gains. Nice gains as well and very welcomed and allows me to go that extra notch on settings.

Twirlywoos · 10 Nov 2016 at 10:42

So only people who get gains are:

1. Pascal users
2. SLI Users any generation
3. People with Old Generation CPUs with low overclocks

???

Gregster · 10 Nov 2016 at 10:59

No idea Twirly but you can see a Pascal user getting gains.

Rroff · 10 Nov 2016 at 11:04

Gregster said:
First time I have seen DX12 acttually give me (as a current NVidia user) any gains. Nice gains as well and very welcomed and allows me to go that extra notch on settings.

Might be the difference in settings? but the DX12 side shows a ton more stutter than the DX11 side :S and some other symptoms of uneven frame pacing.

Gregster · 10 Nov 2016 at 11:22

Rroff said:
Might be the difference in settings? but the DX12 side shows a ton more stutter than the DX11 side :S and some other symptoms of uneven frame pacing.

I don't see any of that stutter when gaming but I am on G-Sync and everything in the game is ultra smooth for both in truth. Exact same settings as well and double checked.

Nexus18 · 10 Nov 2016 at 11:45

SiDeards73 said:
Nexus got roughly 10% uplift on his cpu and R9 290, cant remember his CPU but it was Intel, i think an early i7.

It was a beastly i5 750 @ 4GHz

Twirlywoos · 10 Nov 2016 at 11:59

Gregster said:
First time I have seen DX12 acttually give me (as a current NVidia user) any gains. Nice gains as well and very welcomed and allows me to go that extra notch on settings.

Why does your GPU Core Clock fluctuate between the two tests?

The DX12 max was 1962Mhz
The DX11 max was 1949Mhz

During the second segment

DX12 held it's clocks at 1949Mhz whilst at the same scene
DX11 held it's clocks at 1936Mhz

Towards the end scene both cards remain at 1936Mhz.

Also

neil_g
this is confusing. Nvidia list the 980ti Maxwell as compatible.

http://www.geforce.co.uk/hardware/technology/dx12/supported-gpus

So my card does do DX12 and isn't a Pascal or NV only feature. So I am still confused why I don't get any gains. Any other 980ti users who play at 1440p seeing any benefit?

JediFragger · 10 Nov 2016 at 12:01

Best thing about the DX12 renderer for me is that it shows that they've got long-term plans for future DLC (that and the fact they put so much effort into re-balancing the game) :cool:

Rroff · 10 Nov 2016 at 12:01

Twirlywoos said:
[/SPOILER]

Why does your GPU Core Clock fluctuate between the two tests?

The DX12 max was 1962Mhz
The DX11 max was 1949Mhz

During the second segment

DX12 held it's clocks at 1949Mhz whilst at the same scene
DX11 held it's clocks at 1936Mhz

Towards the end scene both cards remain at 1936Mhz.

GPU Boost 3 behaviour will vary a lot with temperatures and GPU load, etc.

Dicehunter · 10 Nov 2016 at 12:32

Twirlywoos said:
[/SPOILER]

Why does your GPU Core Clock fluctuate between the two tests?

The DX12 max was 1962Mhz
The DX11 max was 1949Mhz

During the second segment

DX12 held it's clocks at 1949Mhz whilst at the same scene
DX11 held it's clocks at 1936Mhz

Towards the end scene both cards remain at 1936Mhz.

GPU Boost 3.0.

Twirlywoos · 10 Nov 2016 at 12:45

Rroff said:
GPU Boost 3 behaviour will vary a lot with temperatures and GPU load, etc.

Dicehunter said:
GPU Boost 3.0.

Sorry if I am looking into this a bit too much

But to rule out GPU Boost 3.0 resulting in increase in avg FPS for DX12 he should have let the GPU cool down as you you can see throughout the video (especially at the start) that the DX11 tests were done when the GPU had slightly higher temps. I didn't realise their was a small temp margin before GPU Boost 3.0 kicks in.

Temp
Upto 57c 1962Mhz
Upto 63c 1949Mhz
63c + 1936Mhz

Most of the DX12 tests were done at a higher core clock compared to DX11 which was at 1936Mhz most of the time. For the sake of a fair test keep the fans at 100% and monitor the clock speeds and test again.

I'm going to underclock my CPU to 4.4Ghz in the afternoon and redo my tests just to rule out my current OC being the cause for no FPS gains.