DOOM updated with Vulkan support

[hsjj3]

Kepler is going dooooown boys. And that is tested with async compute disbled(SMAA) so further gains possible for AMD.

So how is a GTX 750 Ti noticeably faster than a GTX 760? Back when I had the 750 the 760 was much ahead.

Is this because the 750 is Maxwell and the 760 is Keplar? Has this ever happened before, the low designation of a series becoming faster than the one above it?

 

[Yakk]

Sure, nvidia was still actively optimizing Maxwell over Kepler so makes sense. In the same way lesser Pascal cards are now receiving more optimizations than Maxwell. In the same way I wouldn't be surprised to see some lesser Pascal cards overtake traditionally slightly faster Maxwell cards.

These Benchmarks seem to be showing how optimization dependant specific nvidia cards seem to be versus AMD cards relying more on straight hardware performance given the same basic code.

Will be interesting to follow how this further develops.

 

[swilli89]

Holy ****, GTX770 vs 7970!! Even R7 370 is much faster. Who would have thought this would happen with 2012 hardware? It's like ancient in computer world yet the best purchase you could ever made in the history of PC gaming. :thumbsup: Nvidia need to push new architecture every year just to keep up with AMD's moderately upgraded architecture ancient architecture.

Wow. Just wow. The 780ti is less than half as quick as its contemporary, the 290X. Its yet another glimpse into future APIs loving GCN hardware and nVidia offering poor support for its older generations.

How does AMD have a bad rap for some ambiguous, anecdotal "bad drivers" when this is further demonstrable, definitive, quantifiable PROOF that nVidia has poor driver support for anything other than its bleeding edge cards???

 

[Yakk]

Wow. Just wow. The 780ti is less than half as quick as its contemporary, the 290X. Its yet another glimpse into future APIs loving GCN hardware and nVidia offering poor support for its older generations.

How does AMD have a bad rap for some ambiguous, anecdotal "bad drivers" when this is further demonstrable, definitive, quantifiable PROOF that nVidia has poor driver support for anything other than its bleeding edge cards???

Expect nvidia counter marketing either finding some obscure detail not really affecting anything but gives them an argument, or then taking up something"in the future" like Conservative Rasterization to distract from their very real looking previous generations Performance Issues.

 

[Ares202]

Wow. Just wow. The 780ti is less than half as quick as its contemporary, the 290X. Its yet another glimpse into future APIs loving GCN hardware and nVidia offering poor support for its older generations.

Shocking really, people would have laughed at you if you told them that the 290X would be twice as fast 1-2 years ago in DX12

 

[tonyfreak215]

This (and other reasons) are why I will only buy AMD.

I like their consumer friendly business practices, even if it's to their detriment.

I'm kicking myself for not buying a 7970 when they first came out :/

 

[RussianSensation]

the r7 370 (7850 OC) is beating the GTX 780 ti

hopefully Nvidia will improve their drivers, that should happen, 780 ti behind 1024SPs GCN 1.0

Kepler is now joining GF5 and 7 as another GPU architecture that aged horribly. Luckily for NV, instead of these gamers learning from the past, they just get the next NV gen product. In a sense, the faster older gen becomes outdated, the better it is for NV. We are also seeing 7950/7970/290 users upgrading to 1070 because AMD has nothing for them. Once again, the longevity of GCN played right into NV's hand. The vast majority of average Joes never cared and won't that AMD's modern cards last longer. NV is king at having top performance over the first 12-18 months of a GPU generation which gets most people to buy their cards. Then they shift driver focus to the next generation.

The irony here is that even if NV provided future drivers, we'll still hear how AMD drivers suck.

 

[Magic Carpet]

RussianSensation,

Kepler is still great at anything but gaming, though. Like if you use CUDA, or multi-monitors, speaking of power. Or maybe if you're stuck in Windows Vista or Windows 8.0 or Windows XP. Nvidia still provides driver support for those legacy OS'es.

 

[krumme]

If this in any way is representative of dx12 vulcan perf what is the reason to upgrade from a 290x to a 1070? its slim at best. Only vega or big die nv make sense then. And big die nv is gone a be a big xxx die to even touch vega in theese kinds of loads.

The question me is now do you want dx11 optimized card or dx12 optimized. Not what color the card is.

 

[Dygaza]

This is actually one of the hardest times to make good logical decisions of what card to buy. So much is based on how quickly DX12 or Vulkan games come to markets. I have no doubt upcoming 1060 for example will beat rx480 in 90% of dx11 titles. But is that good enough reason if you buy the card for next 2-3 years? Sometimes that crystal ball would be really handy.

 

[Red Hawk]

Eurogamer/Digital Foundry did some testing with 1080, 1070, Fury X, 480, 980, and 970:

https://www.youtube.com/watch?v=ZCHmV3c7H1Q

Modest gains for Nvidia, huge gains for AMD. They measured a 50% gain for Fury X in one example.

 

[Dygaza]

Correction, ignore this post.

Spoiler

Seems there is some serious driver installing issue on nvidia that causes Vulkan to not work properly:

https://forums.overclockers.co.uk/showpost.php?p=29777026&postcount=427

Originally Posted by nhornby51743
PERFORMANCE FIX FOR NVIDIA USERS: READ READ READ

Sorry about the caps, that was to draw attention.

To get better performance Vulkan has to be installed on your pc first. To do this go into C:\Program Files\NVIDIA Corporation\Installer2\Display.Driver, then install it from there.

After I installed this my GTX 1070 was putting out an average extra 30fps frames, maxed out at 1080p.

And the end result:

https://www.youtube.com/watch?v=A-tpzjIL7BI

Huge difference. Explains why some people has gains and some are getting nothing.

 

[Bacon1]

https://youtu.be/WOaHpZjQ73M?t=408

Nice, huge perf/watt gains on the 480 due to the increased FPS, but same overall power usage.

 

[Bacon1]

Seems there is some serious driver installing issue on nvidia that causes Vulkan to not work properly:

https://forums.overclockers.co.uk/showpost.php?p=29777026&postcount=427



And the end result:

https://www.youtube.com/watch?v=A-tpzjIL7BI

Huge difference. Explains why some people has gains and some are getting nothing.

As someone pointed out in the comment on that video:

Patched: 2160x1200
1.0.8: 2560x1440

That is a huge difference of 1,094,400 pixels or 30% of the total screen size reduced.

Explains why its running faster on "Patched" version.

 

[DisarmedDespot]

Seems there is some serious driver installing issue on nvidia that causes Vulkan to not work properly:

https://forums.overclockers.co.uk/showpost.php?p=29777026&postcount=427



And the end result:

https://www.youtube.com/watch?v=A-tpzjIL7BI

Huge difference. Explains why some people has gains and some are getting nothing.

Er, am I reading it wrong, or was the patched game in the video running at 2160x1200 and the unpatched at 2560x1440?

 

[Dygaza]

Yeah, it is. I didn't notice either before someone pointed out. Where would you even get 2160x1200 resolution? Told him his error, let's see what follows.

Edit. As expected, no gains.

 

[Bacon1]

Yeah, it is. I didn't notice either before someone pointed out. Where would you even get 2160x1200 resolution? Told him his error, let's see what follows.

Edit. As expected, no gains.

Yeah no gains:

Damn it, no gains for me... Sorry guys and my bad for not checking the resolution


https://forums.overclockers.co.uk/showpost.php?p=29777663&postcount=445

 

[Maverick177]

Epic Fail, lmfao.

 

[Sushisamurai]

Damn, all this talk about 280x makes me want to buy doom just to try this Vulcan/a sync compute out. I have 3-280X's, I wonder how it'll scale...

 

[n0x1ous]

Damn, all this talk about 280x makes me want to buy doom just to try this Vulcan/a sync compute out. I have 3-280X's, I wonder how it'll scale...

It won't. No Crossfire or SLI support.

 

[Yakk]

Damn, all this talk about 280x makes me want to buy doom just to try this Vulcan/a sync compute out. I have 3-280X's, I wonder how it'll scale...

It's not CF enabled (yet), but I can say it runs AMAZINGLY well on just one 280x.

 

[Sushisamurai]

... The ?recent lack of mGPU support has been super disappointing. I've been seriously contemplating EMA with a odd match of fury, Ti coupled with mid ranges cards just for fun since I got 40 PCIe lanes to play with - the lack of support currently is giving me second thoughts though.

Seriously though, it'd be cool to have a Fury, 480, 290X, and 1080 running EMA. Back on topic, that Vulcan gain on AMD cards is nuts.

 

[Osjur]

I tried to test the increase we get from only Async Compute as we know that the path only works with TSAA and no AA.

First I checked my fps with no AA enabled in OGL, then with FXAA. Then I switched to Vulkan and redid the tests. No AA and FXAA (which disables async path). I also took the results from TSAA.

Well, it turns out that I'm actually losing performance when Async Compute path is enabled, at least with Fury X. I redid the tests 3 times and I always got the same results. I restarted the game every time I changed a AA or rendering path.

Test system: 3930K @ 4.5Ghz, 16GB 2133mhz DDR3, Fury X, latest drivers.
DOOM settings: 2560x1600p, ULTRA settings.

First place: Advanced Research Complex, after train stops at the station.

OpenGL:
NoAA: 54fps
FXAA: 53fps
TSAA: 52fps

Vulkan:
NoAA: 71fps
FXAA: 74fps
TSAA: 69fps

Second place: Argent D'Nur, at mission start. (forgot to take pic)

OpenGL:
NoAA: 82fps
FXAA: 81fps
TSAA: 79fps

Vulkan:
NoAA: 102fps
FXAA: 105fps
TSAA: 98fps

Yep, results are not skewed and Vulkan with FXAA (no Async Compute) is actually faster than Vulkan without AA and Async Compute enabled

 

[kondziowy]

I do not agree with this...
In Necropolis(the Crucible mission beggining) on Fury non X 1440p

NoAA: 109fps
FXAA: 97fps
TSSAA: 104fps
also SMAA: 94 fps

FXAA is slower on my end.

 

[Osjur]

I'll try in necropolis as well.

Edit: getting regression there.

NoAA: 94fps
FXAA: 100fps
TSAA: 91fps