Fresh Vulkan benchmarks. Lyin' Novidya rekt !
digital foundry joins the ride:
https://www.youtube.com/watch?v=ZCHmV3c7H1Q
We did it 4chan
This is just proof that AMD's drivers are pathetic.
>>55556062
>this fury x score
what the actual fuck
Does DOOM support asynchronous compute when running on the Vulkan API?
Asynchronous compute is a feature that provides additional performance gains on top of the baseline id Tech 6 Vulkan feature set.
Currently asynchronous compute is only supported on AMD GPUs and requires DOOM Vulkan supported drivers to run. We are working with NVIDIA to enable asynchronous compute in Vulkan on NVIDIA GPUs. We hope to have an update soon.
>>55556235
>>55556286
>We are working with NVIDIA to enable asynchronous compute in Vulkan on NVIDIA GPUs. We hope to have an update soon.
Never ever. Nvidia cards can't into async compute.
>>55556286
>working with NVIDIA to enable asynchronous compute
how can you enable something that isn't even there to begin with?
>>55556286
>>55556318
>>55556338
Nvida doesn't have the hardware to do asynchronous compute.
>>55556338
>>55556318
>>55556286
stop spamming you fagget
>>55556318
>>55556338
repeating it won't make it real. it is not magic.
Nvidia are going to fix it with new drivers and BTFO AMD.
>>55556394
YES! DRIVERS WILL FIX EVERYTHING AND HOPEFULLY MAKE AMD CRY!
>>55556286
>>55556318
>>55556338
nVidia: The Way It Is Meant To Be Waited™
>Nvidia
>energy effecient
>>55556412
you retard by the benchmarks you can clearly tell that AMD only supports only one card at the moment.
fucking fuckboi
do you get off by being wrong or what
>>55556434
>AMD only supports only one card
and which card is that?
>>55556434
now tell me fuckboi
where's the maxwell drivers you've been waiting for, huh?
give us something good or gtfo
>>55556286
The nvidia cards get a 5/10% boost from vulkan currently so the pascal cards should get a nice gain from the new drivers. I remember the 1080 hitman video digital foundry did with dx12 and the 1080 gained as much as 30 fps in some scenes so I'd expect as much from the new update.
should I buy a 480 now and a second one later?
>>55556455
gee I wonder which card has enabled async compute.
surely it's not the fury.
>>55556458
I'm not waiting for any drivers.
I'm just kinda using my gpu as god intended and don't even care about 1% up or down.
do you guys simply have nothing better to do than scream at benchmarks?
>>55556455
The Nvidia Killer.
>>55556493
>new drivers
so you are waiting for, what, 6 months? where are the drivers for Ashes enabling async?
>>55556529
all of the amd cards have async enabled you retard.
>>55556530
so, all of the amd cards?
>>55556521
If only the 1070 was available for the MSRP ($379), it would be a no-brainer.
when will we be able to purchase one for the MSRP?
>>55556521
Nah, wait for the aftermarket cards to come out because the reference card is power constrained and has a really shitty cooler (so as a result they'll be much faster). If you then in the future decide to buy another card the new multiGPU functionality in DX12 isn't tied to vendor support the second card doesn't even have to be an AMD card.
>>55556325
>>55556648
if I can afford it, is it better to wait for the 1070 to drop to MSRP? or still get an aftermarket 480?
>>55556639
It's already pretty cheap in the uk. You can pick one up for 399 which is only about £20 more than the msrp. I'm going to be getting one soon when I upgrade my whole pc as the 1060 is barely upgrade over my current card and amd are too incompetent to make a decent gpu that has consistent performance on the current standard api which is DX11. I don't want to have to wait 2 months for an update like with DOOM just to use an amd card at decent framerates.
>>55556706
don't buy anything that doesn't support async and new apis (dx12 and vulkan) properly.
>>55556709
>You can pick one up for 399
literally where? pass a link, I wanna ship to the netherlands. they're sold out on amazon:
https://www.amazon.co.uk/EVGA-Nvidia-GeForce-Gaming-Graphic/dp/B01GVHNWUK/
>>55556739
so what should I buy?
>>55556706
A 1070 is a faster card, but in terms of value it's not as good of a deal as even the reference 480. If you want a faster card, go ahead, but regardless even at MSRP you're paying almost twice as much for nowhere near as big of a boost in performance.
>>55556639
Look at jet.com if you want a cheaper one. You can get some of the more basic models for 380-390 usd, I bought the EVGA FTW edition for 427.
>>55556062
Hab den gerade auf CB gelesen. Wird gut. Unity hat das kaum auf der Roadmap, also kein KSP mit Vulkan in der nächsten Zeit :(
>>55556759
until nvidia comes out and ACTUALLY give a decent answer or a driver to work on async you should stay away from them...
remember dx12/vulkan is oriented on the compute path.. meaning the tflops you see have way more meaning now than on dx11.. thats why you see the fury x going nuts on doom being 5 fps behind 1080
>nvidia offer better performance in the vast majority of games
>for a lower power consumption
>with better drivers that work on day one
>and have the best background video capture
But oh my god there's that one DX12 game where ATI win and now there's a whole another game, that's like 2x more games! Nvidia is finished!
>>55556795
yeah but I'm not going to wait a year for developers to jump on the new API bandwagon
>>55556782
>jet.com
doesn't want to load anything
>>55556820
i dont know pretty sure they already jumped on unless ofc you are one of those that thinks any dx12 games released so far are irrelevant or "benchmarks"
>>55556810
>But oh my god there's that one DX12 game where ATI win
All of them.
>>55556739
What if Nvidia comes with a driver update? Will AMD be kill?
>>55556834
>beating a 980
wtf. if only AMD wouldn't suck ASS on DX11
>>55556810
nvidia has burned more cards down than paul pot kill count
http://modcrash.com/nvidia-display-driver-damaging-gpus/
http://www.zdnet.com/article/warning-nvidia-196-75-drivers-can-kill-your-graphics-card/
http://gadgets.ndtv.com/laptops/news/windows-10-is-killing-nvidia-graphics-cards-720443
they even burned down monitors too
https://forums.geforce.com/default/topic/862417/geforce-drivers/windows-10-geforce-drivers-are-killing-samsung-and-lg-notebook-lcd-display-panels/
kek
>>55556853
They already compete - this attitude that AMD is terrible at DX11 has no basis in reality.
>>55556759
http://www.geforce.co.uk/hardware/10series/geforce-gtx-1070/
>>55556795
A driver is not going to fix the fact that Nvidia still can't do async compute in hardware. The devs behind AotS found out that their software implementation actually leads to a regression in performance and when they asked Nvidia about it they asked them to not use async compute on their hardware.
AMD spent years and probably billions in R n' D on low level APIs with features like async when Microsoft and the Khronos group weren't really interested in it. At the same time Nvidia knew that APIs were not going to go low level or support stuff like async compute for a long while and put their R n' D money elsewhere and reaped the benefits. Now APIs have rapidly adopted low level functionality like async compute based on AMD's implementations and Nvidia's short sighted planning is about to screw them very badly.
Engineering projects as big as GPU architectures don't exactly turn on a dime so we shouldn't just expect Nvidia's initial implementations of hardware async compute to fail to match AMD's, it'll be a while before they even come out.
>1060 3gb
>470 4gb
Both $150, NVIDRONES BTFO
>>55556891
Consider that pascal's hardware was beyond the point of any real change mid 2015 at the latest shows exactly how fast the software adoption has been in relation to the hardware.
>>55556820
This. These new api won't be fully adopted until ~2018/19. They might get dx12 support but the fact they'll still be built in dx11 means it's fine to have a card that's over 5 years old or current nvidia cards. Also fuck current dx12. Devs aren't competent enough to make good dx12 support yet and there's always stupid problems with the dx12 mode like on hitman it takes 30 minutes to boot up in dx12 mode for me. Dx11 at least works as intended.
>>55556835
Nvidia doesn't have async compute in the hardware. driver update changes nothing.
>>55556935
Oh shit, Nvidia is kill!
>>55556921
I wouldn't be so sure about those Doom Vulkan benches being that descriptive seeing how async compute is far from the only new feature in Vulkan and DX12. However unfortunately there's no way to turn async compute on and off in any Vulkan or DX12 games so we don't know for certain.
Sure is POO in here...
>>55556891
oh i know which is why it will make it even more hilarious
but if it was amd on nvidia place it would have been an uproar class action lawsuits for blatanly lying bla bla..but somehow sheeps are fine with a 1 year old card already being DOA
>>55556950
> However unfortunately there's no way to turn async compute on and off in any Vulkan
For doom you can by changing anti-aliasing (async is only active with tssaa or no aa - anything else disables it) and iirc in AoTS you can fuck with an .ini to disable it.
>>55556950
async is the single feature that can actually bring more perfomance on dx12
the rest just helps
>>55556416
>Last gen
>vs
>Current gen
Nice node shrink fuckup though.
>>55555555
>>55556416
Dude, I'm pretty sure the 970 isn't 14nm FinFET.
>>55556062
>>55556297
>>55556416
>>55556834
DELETE THIS
>>55557002
>>55557053
kek
damage control
when amd beats 970 is
>last gen
when 970 beats amd
>woah check it out QQ epic omg lel nvidia best
>>55556925
>All big third party engines (UE, CryEngine, Source and even Unity) have have all ether already released implementations or are working on them
>Even publisher in-house engines like Frostbyte (EA), id Tech (Bethesda) and Glacier (Square Enix) are in the process or have already adopted them
You may have been right a year ago, but as time goes on you can be sure that more and more games will support these APIs. There's not that many developers who build their own engines anymore, they ether get a third party one like UE or then one from their publisher and both groups are moving to low level APIs.
>>55556394
>Nvidia are going to break fan speeds with new drivers and BTFO their customers.
>>55557087
Projecting your own mindset there, m8?
Node drop = massive increase in efficiency. 22 or 28nm cannot beat 14nm under any circumstance, 14nm is objectively better.
Might wanna google that shit you tech illiterate fool.
>>55557002
You want to see some last gen demolition? How about this for size.
>>55557124
then why is the 480 such a powerhog?
>>55557134
How would I know? I'm not AMD.
I'm waiting for Volta.
>>55557154
kek
Also, this is the first DX12 game. Same repeat game changing performance difference.
>>55557094
Yes but games will still be built in dx11 for a good few years yet. We won't see a point where games will specifically be made in dx12 only for some time. Fanboys keep going on about dx12 as if it's only comprised of one thing which is asynchronous compute. Even dx11 supported async but dx12 brings a lot more requirements to the table like conservative rasterization which older cards and amd cards in particular don't fully support. Hence why amd GCN is classified as dx12_0 support and nvidia maxwell/pascal is classified as 12_1 because of the Conservative rasterization support. It's on the wiki.
>>55557154
Why not Vega?
If i mostly play GTA V, Fallout 4 and don't care for DX12 titles, what card should i get?
I really want to give AMD a chance, but i heard those games run better on Nvidia.
Oh dang, I just got a 1070 too.
Oh well.
>>55557188
what's your price range?
>>55557124
i could go literally all technical on your ass and make you log off from the internet but its not hilarious this way..
first of all tsmc claims x2 perf/watt from 28nm to 16nm
if you do the calculations nvidia actually managed x1.4 for some reason they didnt get everything out mostly because they are pathetic as always
glofo claims x2.3 from 28 nm base perf
amd got on 470 x2.8 from their previous gen (r9 380) and x2.1 from 480
considering that amd naturally has a lot of hardware comparing to nvidia paxwell actually didnt gain anything in reality that is why they see such a massive regression on dx12 again and i believe its even more than it was on aots (12%)
>>55557182
Nah, I'm doing fine with my 770 at the moment. I might wait to see what Navi brings though.
>>55557188
During the entire development of GTAV on PC, Rockstar didn't use a single AMD card. That should answer your question.
>>55557209
~$200 range.
>have 290
>ever since 3xx series came out, my card is never on any benchmarks
c-can I just look at 390 and subtract 5% fps? pls ;__;
>>55556310
Except they can.
They just can't do it on the cards in hardware, management of the different resources has to be done on threads on the CPU.
With a decent core count and clockspeed, it won't even matter and could even potentially be faster as time goes on (software can be updated, hardware can't)
However for a majority of people, that will mean AMD will win the async contest.
Expect to see 5960X's used a lot more in benchmarks though...
>>55557132
All those 700 cards.
>>55557251
Software never beats dedicated hardware for a job.
>>55557225
Nice off topic tirade you got there champ. You're not only not as smart as you think you are, but you are also an illiterate monkey only capable of reciting numbers without understanding the context or the conversation.
>>55557234
well you have 3.5gb 970, 4 gb rx480 and soon to be released 3gb 1060.
>>55557177
>Yes but games will still be built in dx11 for a good few years yet
Please make like like an Abit motherboard and just stop posting...
Games themselves are not "built" in any API. Renderers, which are a central part of engines and generally not built by the developers themselves, are built USING various APIs and right now they ether have been re-built using the new APIs or are in the process of doing so. If the engine a developer is given to work with supports the new APIs, then the games they make with it will also support.
>>55557188
Ask yourself one question: Are these the games you'll be playing for the whole lifespan of the card?
If yes, buy Nvidia, if not, hold off to see how non-reference 480's or Vega performs.
>>55556062
Nvidia is finished and faggot?
>>55557276
Except video decoding apparently...
>>55557278
He's posting in 2 threads lol. I've been following his posts for like 20 mins now and he keeps talking the most bullshit.
>>55557284
Just recapping your point, so I can post-
The renderer is like the easier part of the actual engine to complete, things like accurate and *cheap* collision detection are far harder.
>>55556639
Even then it's not. The 480 is better than the 1070 (more memory, actual async, much faster in some titles, equal in others), while being much cheaper too.
>>55557284
thats literally what was mentioned in the post you replied to retard. youre repeating the same shit
>>55557371
You're eating the same shit, but I'm not getting all upset about that now am I?
>>55557366
>equal in others)
nope. the 1070 kicks 480's ass in DX11
>>55557278
you were the one claiming node drop=massive increase on effeciency while this is never true since the node chance only accounts for 30% of the perfomance you can gain..
>>55557406
False. The 1070 is never better than the 480.
>>55557426
you brand loyal retards need to kill yourself. both nvidiots and amdrones
anyone had benchmarks for the legacy cards like amd hd 5000 6000 and 7000 series?
>>55557410
Yea, that's right. That's what I claimed and it's universally true. You always get massive performance increase for the same power consumption assuming no changes are done to the architecture to deliberately make it worse.
Then you went and started talking about your dinner and what kind of foot rubs you like. Freakin' idiot.
All the retarded fanboyism and trolling aside, anybody else think the boost 480 and other AMD cards get from Vulkan is fucking impressive? This may well be the concrete start of the new graphics API era.
>>55557436
Exactly. They trust benchmarks that show much lower numbers for the competition than any other benchmark ever produced just to defend their precious nvidia.
>>55557461
What I find impressive is just how shit their drivers for DX11 are. The performance they get from Vulkan isn't magic. Their hardware is decent it's just held back by trash drivers.
>>55557463
I'm talking about this amd-drone: >>55557426
>>55557473
>d-d-damage control!
nvidiots on sucide watch!
I like how every benchmark is different. every thread. same dumb shit.
>ITT
>>55557470
If the performance comes from async computation, then it's basically magic. If not, either their dx12 drivers are god-like or the dx11 drivers severely... bad.
I love how /g/ screams about ZOMG WIN10 BOTNET WIN7 4EVAR
Then they do a full 180 and scream DX12! DX12! DX12!
>>55557586
>DX12
You mean Vulkan, right? DX12 is shit.
>>55557596
>CTRL+F DX12
>30+ results
Hurr durr
>>55557583
Their DX12 drivers just don't really get in the way. That's all. I mean their DX11 drivers are single threaded, so the ironic thing is that using AMD cards with AMD processors is almost impossible without sacrificing a lot of performance.
Want AMD processor? Oops, looks like you need nvidia card.
>>55557132
Note, this specific benchmark is with SMAA.
While the OP's benchmark uses TSSAA.
SMAA doesn't use async compute, TSSAA does. So if your shot used TSSAA, then 1080 would be blow away most likely.
>>55559203
Pics related.
Same review.
>>55556211
There's absolutely nothing that can be done about DX11 and OpenGL not being able to use AMD ACEs at a driver level. They bet the farm years ago on bare metal APIs like Mantle, Vulkan and DX12 becoming the norm, and now the time has come and so have I.
>>55556062
GTX 1080 still faster.
>>AMD BTFO
>>55556062
Every day both companies are finished at least 20 times. I don't fucking care anymore.
>>55556235
an amzeing card, held back partially by drivers, mostly by the process.
once you remove the drivers almost entirely from the mix, the process becomes less of an issue as well.
now imagine this on the process it was built for, the 20nm... god amd left tsmc for a god damn reason.
>>55559870
For all practical cases keeping a fury x saturated is really, really hard because of lel 64 rops. The massive shader array is always stalling so the more work you shift onto its compute units (of which it has many) the better the card runs.
Hell you basically can't cause a fury x to choke on fp32 compute workloads outside of intentionally trying to bring the card down.
