First Nvidia & AMD Deus Ex: Mankind Divided DX12 Benchmark appeared

[Mr.Meerkat]

5 minutes ago, Morgan MLGman said:

But ROTTR isn't actually DX12, it's something of a DX11.2 which is DX11 with some of DX12 features, that's pretty much a transferred game from Xbox One.

So we can stop using it as an argument for the 1060 vs RX480 debate?

 

10 minutes ago, Godlygamer23 said:

Something else to bear in mind here is that AMD drivers have a higher CPU overhead versus NVIDIA drivers. By using DX12, the cards are simply allowed to perform better than with previous versions.

True true, so can we debunk that argument as well?  

Probably shouldn't however as it's still a valid argument as DX11>DX12 performance still at the moment  

[Morgan MLGman]

Just now, Mr.Meerkat said:

So we can stop using it as an argument for the 1060 vs RX480 debate?

This has been known for quite a while, I don't consider it a DX12 title. Though the DX12 mode helps FX-line CPUs quite a bit, minimum FPS is MUCH higher when transitioning to that mode.

[Morgan MLGman]

5 minutes ago, AresKrieger said:

Well if this trend keeps up nvidia will either push forward the 5 year plan or they have a special dx12 version of gameworks ready so they are just waiting for the titles to launch

 

Either way at least for me so far I haven't cared about the games nvidia has been stumbling on, hopefully that trend continues but I would just lower shadow quality if it didn't since tbh video game shadows are higher detail than actual shadows often enough.

We'll see what Vega will bring (besides HBM2), cause Polaris is a little weird. It's aimed more at 1080p and efficiency, but it lost a little of good ol' compute compared to Hawaii.

 

My 290X despite being similar at 1080p to a 480 (it's faster by quite a bit as it's OC'd), beats the 480 significantly in DX12 due to its beefy Hawaii XT chip and a wide, 512 bit memory bus (especially coming in handy at higher resolutions). Look at the difference between the 390X and the 480 at the first graph, that pretty much describes it, as it's got the same chip as my 290X, it's just higher factory clocked, so this is pretty much how much faster a 290X is over a 480 in DX12...

[Hakugman]

Wow.. to think my Gtx 1080 would down perform like this with dx12 is absolutely ridiculous. 

[Hakugman]

26 minutes ago, Sauron said:

Wow. Imagine paying 800$ for a 1080 and then having it perform like a fury. Hold on to your wallets boys, pascal ain't worth it.

-edit-

to be fair most games aren't dx12 or vulkan, but still, if at all possible it's a good idea to wait.

The whole DX 12 API is broken as hell atm, don't get your panties in a bunch. I could say that Microsoft should learn from Vulkan and get their game together because so far their marketing blargh is going down the hill. 

[Sauron]

Just now, Hakugman said:

The whole DX 12 API is broken as hell atm, don't get your panties in a bunch. I could say that Microsoft should learn from Vulkan and get their game together because so far their marketing blargh is going down the hill. 

Vulkan is showing a similar story though, maybe slightly less dramatic but still, it's not looking good for pascal.

[Guest]

.

[Mr.Meerkat]

8 minutes ago, Morgan MLGman said:

This has been known for quite a while, I don't consider it a DX12 title. Though the DX12 mode helps FX-line CPUs quite a bit, minimum FPS is MUCH higher when transitioning to that mode.

People are still using it as an argument sadly...they just claim "it's down to the devs for whether AMD or Nvidia performs better in DX12" which is true to a certain extent but in the case of ROTTR, it's not true  

[Valentyn]

6 minutes ago, Hakugman said:

The whole DX 12 API is broken as hell atm, don't get your panties in a bunch. I could say that Microsoft should learn from Vulkan and get their game together because so far their marketing blargh is going down the hill. 

The API is fine, it's entirely down to developers to put in the effort for vendor specific codepaths to get the best performance.

If they don't they're better off sticking to DX11 than hacking together a broad DX12 codepath that doesn't really do the job.

 

Add in overbearing Publishers that can't keep their grubby hands out of developers' work and you get rushed halfassed messes.

[UndueHatred]

Quote

EIDOS Montreal and NIXXES noted that there is a known bug that causes some very high end cards to regress relative to DirectX 11. According to them, this bug is being addressed by the development team, so it will be interesting to see whether the culprit of this underwhelming DX12 performance is this bug.

Source

[lukart]

well looks like 480 keeps being a better option for future titles... just saying

 

By the way, Guru3D didn't even used the latest AMD driver that just came out today with Dx12 support for Deux. 

[Prysin]

49 minutes ago, Morgan MLGman said:

Did AMD sponsor this game? Like WTF...

yes, its a AMD title. Based on the same engine as Hitman was.

Yes it is better optimized towards AMD.

 

Nvidia has had plenty of time to optimize for their own architecture.

[Curufinwe_wins]

Sounds like a really shitty DX12 port again....

 

Only 5% boost on radeon cards? Yea no shit then the Nvidia cards are suffering. Why the fuck can't developers not suck at DX12.

[Curufinwe_wins]

27 minutes ago, Sauron said:

Vulkan is showing a similar story though, maybe slightly less dramatic but still, it's not looking good for pascal.

Really because Doom didn't cause any Nvidia cards to do worse....

 

Pascal on the properly done DX12 titles shows a 5-10% boost. GCN shows a 15-35% boost. The fact that the GCN cards are only getting 5% means this DX12 codepath is probably dogshit.

[Sauron]

Just now, Curufinwe_wins said:

Really because Doom didn't cause any Nvidia cards to do worse....

 

Pascal on the properly done DX12 titles shows a 5-10% boost. GCN shows a 15-35% boost. The fact that the GCN cards are only getting 5% means this DX12 codepath is probably dogshit.

But in Doom the improvements were a LOT lower for nvidia cards.

[Curufinwe_wins]

3 minutes ago, Sauron said:

But in Doom the improvements were a LOT lower for nvidia cards.

AND?!?!?! 

 

They were still beating their competition (outside of the Fury X pulling ahead of the 980ti/1070 stock v stock). End of story.

 

This port showing yet another regression in DX12 is literally proof of shitty developers. Because clearly properly done benefits can be had.

 

Average FPS	RX 480	GTX 970	GTX 980
Open GL	69.7	75.7	86.6
Vulkan	89.7	77.3	90.7
Performance Boost	+28.7%	+3.4%	+4.7%

Average FPS	GTX 1080	GTX 1070	GTX 980 Ti	R9 Fury X
Open GL	134.0	107.7	109.3	88.7
Vulkan	149.0	115.0	115.0	123.7
Performance Boost	+11.2%	+6.8%	+5.2%	+39.5%

 

[WereCat]

So basicaly R9 Fury Sapphire Nitro is €320 which is more or less the same price as GTX 1060 (I paid €310 for mine)... except the Fury destroys the 1060 in DX12 with about 50% more performance and matches the GTX 1070 which costs €120 more.

I already got the GTX 1060 in my PC but I am not regreting it yet as the only DX12 title out there that I want to play is new Tomb Raider and that one doesent scale well in DX12 for both team red and green. Also it performs much better in almost all games that I play in DX11.

[Curufinwe_wins]

2 minutes ago, WereCat said:

So basicaly R9 Fury Sapphire Nitro is €320 which is more or less the same price as GTX 1060 ... except the Fury destroys the 1060 in DX12 with about 50% more performance and matches the GTX 1070 which costs €120 more.

I already got the GTX 1060 in my PC but I am not regreting it yet as the only DX12 title out there that I want to play is new Tomb Raider and that one doesent scale well in DX12 for both team red and green.

The same statement could be made for buying a 980ti for around that price (which is easy to do in the US at least). Both beat the 1060 in everything (except VR, but who cares about that), but the power consumption is well over doubled the 1060.

[WereCat]

4 minutes ago, Curufinwe_wins said:

The same statement could be made for buying a 980ti for around that price (which is easy to do in the US at least). Both beat the 1060 in everything (except VR, but who cares about that), but the power consumption is well over doubled the 1060.

Well, the 980ti costs exactly the same as new GTX 1070 here so there is no point to get it.

[Curufinwe_wins]

Just now, WereCat said:

Well, the 980ti costs exactly the same as new GTX 1070 here so there is no point to get it.

Makes sense then. Local pricing issues

[Morgan MLGman]

24 minutes ago, Prysin said:

yes, its a AMD title. Based on the same engine as Hitman was.

Yes it is better optimized towards AMD.

 

Nvidia has had plenty of time to optimize for their own architecture.

Oh I know that, I mean I understand getting bigger boosts in DX12 etc, I'm in love with my 290X for that, but a Fury X 10FPS above a GTX 1080 at 1080p? I'm like... whaaaaat

 

[NuclearPeace]

Its almost as if switching to a new and radically different API will have its initial hiccups...

 

Also, the reason why DOOM runs so fast on AMD cards is because it uses AMD specific shader intristic functions and the fact that AMD's OpenGL drivers are hot garbage so Vulkan provides a dramatic speed boost.

[Techicolors]

1 hour ago, Tataffe said:

Can we get a comparison between DX11 and DX12 so that we can actually see the differences for each card?

[ivan134]

40 minutes ago, Curufinwe_wins said:

AND?!?!?! 

 

They were still beating their competition (outside of the Fury X pulling ahead of the 980ti/1070 stock v stock). End of story.

 

This port showing yet another regression in DX12 is literally proof of shitty developers. Because clearly properly done benefits can be had.

 

Average FPS	RX 480	GTX 970	GTX 980
Open GL	69.7	75.7	86.6
Vulkan	89.7	77.3	90.7
Performance Boost	+28.7%	+3.4%	+4.7%

Average FPS	GTX 1080	GTX 1070	GTX 980 Ti	R9 Fury X
Open GL	134.0	107.7	109.3	88.7
Vulkan	149.0	115.0	115.0	123.7
Performance Boost	+11.2%	+6.8%	+5.2%	+39.5%

 

This game uses the same engine as Hitman though where AMD is faster even in dx 11 because the it seems to be a compute heavy engine. AMD's gigantic gains in dx12 tend to be in games where they have a really huge CPU overhead in dx 11.

 

Also, the 470 is faster than a 1060 using Vulkan.

[That Norwegian Guy]

1 hour ago, The Benjamins said:

It has a software async so I think it has to spend resources to do that which in turn gives less to graphics tasks, hence a lose. When AMD is hardware level so it will never take a resource hit. (my theory)

Pascal does have onboard Async control now, unlike Maxwell. But it still only has the single shader pipeline vs. AMD's 8, so Pascal's async compute isn't really proper async compute.