[Guru3d] Hitman (2016) DirectX 12 updated benchmarks review

[amenx]

Guru3d Hitman 2016 benches updated to include DirectX 12 results with AMDs 16.4.2 driver.

http://www.guru3d.com/articles_pages/hitman_2016_pc_graphics_performance_benchmark_review,7.html

 

[BlitzWulf]

Wow look at Hawaii go!

The Hotfix released yesterday must have had some magic in it as these results are much better than the benches from PCGH

 

[3DVagabond]

Still no clear advantage for AMD in DX12. :ninja: /s

 

[Erenhardt]

they seem to not have 7970/280x in their graph, which is a shame. But looking at 7870/270/370 giving 960 a run, it makes you think 280x would 'keplerize' maxwells gtx970

 

[Erenhardt]

Still no clear advantage for AMD in DX12. :ninja: /s

Makes you wonder how much more the can squeeze out of those 438 mmsq. We need to remember that is the very beginning of dx12. No one knows what new trick and techniques devs come up with thanks to low level APIs.

 

[3DVagabond]

they seem to not have 7970/280x in their graph, which is a shame. But looking at 7870/270/370 giving 960 a run, it makes you think 280x would 'keplerize' maxwells gtx970

The 380X is ~20% faster than the 970 and matching the 980 @ 1440.

 

[Adored]

No question Hawaii was the card to buy now but damn, Fury just continues to disappoint.

 

[Flapdrol1337]

They tested the 280x with dx11, 2fps faster than the 380 (non x)

770 performs really terrible, which is strange since the 780Ti performs similar to the 970, you'd expect the 770 to keep pace with the 960.

 

[flopper]

No question Hawaii was the card to buy now but damn, .

Nvidia seems they just fell down flat on the ground and what will be the spin from the PR department?
People normally dont upgrade that often but darn its time to BUY Polaris

Time to buy Polaris
shiny and brightly fast

980ti beaten by 390 series, oh such shame.

 

[Shmee]

no crossfire benches?

 

[Flapdrol1337]

Why did they bother with dx12 when even most amd cards are faster under dx11?

no crossfire benches?

It's under fcat.

 

[BFG10K]

Still no clear advantage for AMD in DX12. :ninja: /s

Heh, didn't look at the DX11 benchmarks, did you? http://www.guru3d.com/articles_pages/hitman_2016_pc_graphics_performance_benchmark_review,6.html

DX11 performance is faster in virtually every case. It's another example of a DX12 implementation being a failure given it required more developer resources but yielded performance regression.

Only a moron would run the game under DX12, especially on cards like the 980 Ti.

Whatever's causing AMD's performance advantage (which I fully admit is impressive) has nothing to do with DX12. But it's really amusing to see people desperate to see something that isn't there. :awe:

 

[book_ed]

AMD loses performance under DX12 in 1080p.

 

[Adored]

I'm starting to wonder if AMD fixed some major DX11 driver issue late on 28nm. There's a lot of DX11 titles where they appear to be well ahead (mostly everywhere except the 980 Ti vs Fury).

 

[3DVagabond]

Heh, didn't look at the DX11 benchmarks, did you? http://www.guru3d.com/articles_pages/hitman_2016_pc_graphics_performance_benchmark_review,6.html

DX11 performance is faster in virtually every case. It's another example of a DX12 implementation being a failure given it required more developer resources but yielded performance regression.

Only a moron would run the game under DX12, especially on cards like the 980 Ti.

Whatever's causing AMD's performance advantage (which I fully admit is impressive) has nothing to do with DX12. But it's really amusing to see people desperate to see something that isn't there. :awe:

Not sure how DX11 performance has anything to do with DX12 or what I said?

As far as most cards being faster? I see a lot that that didn't run at all in DX11 but they did in DX12 because of VRAM. Seems strange to say DX12 failed when it runs? Nothing is slower than "0". And it only seems to be nVidia that is suffering from poorer performance. That's not terribly unusual with DX12. Which actually reinforces what I said.

Don't get me wrong, I'm not saying this epitimizes DX12 performance either. Just that AMD continues to outperform nVidia in most DX12 titles. Am I missing something?

 

[BFG10K]

I'm starting to wonder if AMD fixed some major DX11 driver issue late on 28nm. There's a lot of DX11 titles where they appear to be well ahead (mostly everywhere except the 980 Ti vs Fury).

I've been keeping a close eye on DX12, and in almost every case it's a failure (just like Mantle). DX11 often runs faster with both vendors (more-so with nVidia) and without serious problems.

Many games have serious stability and/or rendering issues under DX12 (look at Guru3D above saying they had 5 minute benchmark load times with nVidia).

The burden of fixing this falls almost squarely on the developer, not the driver engineer that intimately knows the hardware. The developer also needs to keep issuing patches to keep up with new hardware, or they'll wind up with the Mantle situation.

Low level APIs aren't the panacea some people think they are. They make sense for consoles given they're fixed hardware, but I personally think it's the wrong direction for PC gaming.

 

[book_ed]

As far as most cards being faster? I see a lot that that didn't run at all in DX11 but they did in DX12 because of VRAM.

Check 1080p, dx11 vs. dx12, AMD is slower as well. Last time I've tried going over the vRAM limits, the game was pretty stuttery. It is above 0, but not I wouldn't consider that a full fat "win".

 

[BFG10K]

Not sure how DX11 performance has anything to do with DX12 or what I said?

As far as most cards being faster? I see a lot that that didn't run at all in DX11 but they did in DX12 because of VRAM. Seems strange to say DX12 failed when it runs? Nothing is slower than "0". And it only seems to be nVidia that is suffering from poorer performance. That's not terribly unusual with DX12. Which actually reinforces what I said.

Don't get me wrong, I'm not saying this epitimizes DX12 performance either. Just that AMD continues to outperform nVidia in most DX12 titles. Am I missing something?

Uh...what? That post is so full of holes I'm not even going to quote it by parts.

Did you even look at the results?

1080p
Nano: DX11 = 85, DX12 = 82
390X: DX11 = 87, DX12 = 82

DX12 shows a clear performance regression, even on AMD hardware. Only a moron would run the game under DX12 when there's a performance regression, wouldn't you agree?

Therefore by extension, any "victories" in DX12 are meaningless.

 

[book_ed]

You see gains in 1440p though. I've also tested it even higher and the were huge gains, but at the moment most likely they don't have the know how required to actually pull this through in a AAA manner.

BTW, even on AMD the game required a few minutes to actually get in the game under dx12 and with settings above the vRAM limit, it just froze after a few moments of gameplay (the image, the sound seemed to still go on). So yeah, it was actually 0 there as well.

 

[Adored]

We don't know if the frame times and minimums are better under DX12 or not. This was frequently the case with Mantle.

Also don't forget that we're looking at the highest-end CPUs. We've already seen what DX12 can do with FX and slower CPUs.

This is what the devs want - their games running better on lower end hardware, not just the high-end. As I've said before, having a $1000 GPU and $500 CPU means nothing to a game dev when you're just paying the same money for their game as somebody with low-end hardware.

 

[seitur]

DX12 shows a clear performance regression, even on AMD hardware. Only a moron would run the game under DX12 when there's a performance regression, wouldn't you agree?

Therefore by extension, any "victories" in DX12 are meaningless.

Trend is clear though. DX12 (and Vulkan possibly) will be getting better and will matter more as time will pass.

High level APIs for graphic demanding titles are not sustainable anymore because of fundamental market changes and increasing cost of hardware on lower processes.
You just cannot keep throwing more powerful hardware at the problem at pace we've been used to in last 20 years.

That's why software and architectural changes to I/O, memory, etc will have to take more burden.

 

[Magic Carpet]

I've been keeping a close eye on DX12, and in almost every case it's a failure (just like Mantle).

Personally, I have witnessed higher minimum framerate with DX12 in the Tomb Raider game, so for me, DX12 is working as expected. It will only get better in time. DX11 is essentially dead for AAA games, come 2017. And no, it's not just like Mantle (to me, Glide was more like it). Microsoft is behind that standard. Feel the difference.

Haswell 4C/8T @ 3.7 | 270 @ max settings, latest patch/drivers | DX11

Haswell 4C/8T @ 3.7 | 270 @ max settings, latest patch/drivers | DX12

I've seen a lot of folks saying their performance dropped using DX12 mode in Rise of the Tomb Raider, so I figured I would run the test. It appears, at least in Nixxes implemention of DX12, that for better performance you'll need a 6-core Intel CPU.
System Specs:
CPU: i7-5820k@4.5ghz
GPU: R9 290x @ 1100mhz(core)/1400mhz(RAM)
RAM: 16GB DDR4 2133 in Quad-Channel
EDIT - Resolution: 2560x1080
Mountain Peak (DX11): Minimum: 36.89, Maximum 108.22, Average: 68.66
Mountain Peak (DX12): Minimum: 48.73, Maximum 78.73, Average: 66.48
Syria (DX11) Minimum 12.68, Maximum 87.69, Average 63.47
Syria (DX12): Minimum 30.49, Maximum 98.22, Average 61.48
Geothermal Valley (DX11): Minimum 13.40, Maximum 76.7, Average 63.06
Geothermal Valley (DX12): Minimum 23.51, Maximum 164.59, Average: 60.76

DX12 & multi-core CPUs are the future, today. Windows Vista/7/8/8.1 are the new Windows XP as well. It will be funny to discuss this topic in say, 2018.

 

[ShintaiDK]

Trend is clear though. DX12 (and Vulkan possibly) will be getting better and will matter more as time will pass.

Its a trend? I must have missed it.

DX12 have everything against it. And its plain simple due to economics and time. DX12 looks good on paper, but its designed for an utopian PC world. Or a console world.

 

[3DVagabond]

Uh...what? That post is so full of holes I'm not even going to quote it by parts.

Did you even look at the results?

1080p
Nano: DX11 = 85, DX12 = 82
390X: DX11 = 87, DX12 = 82

DX12 shows a clear performance regression, even on AMD hardware. Only a moron would run the game under DX12 when there's a performance regression, wouldn't you agree?

Therefore by extension, any "victories" in DX12 are meaningless.

We'll just let people look at the results. Because only a moron would have this discussion with the charts sitting right there.

 

[R0H1T]

Its a trend? I must have missed it.

DX12 have everything against it. And its plain simple due to economics and time. DX12 looks good on paper, but its designed for an utopian PC world. Or a console world.

Like what?

The economics tells us that with DX12 & win10 you can use your decade old CPU & nearly half a decade old GPU to play the latest games, albeit low settings, without much of a stutter fest & btw did I mention that win10 was free of charge

I'm sure you don't exist in that world, you're like the satan for that Utopia D: