Early benchmarks indicate major DX12 gains for AMD

if all you're doing is gaming, there's no need for the X-cards. if you do a bit of editing, than the extra stream processors will come as a bonus.

Say, didn't nvidia have a chance to support mantle since it was all kinda FOSS?

AMD gave them all the opportunity in the world to support mantle, but they chose not to. then again AMD wasn't really expecting Mantle to replace DirectX. even though many hoped it would considering the back end of it was really OpenGL.

Yeah i allready posted those benchmarks in the lounge exaly.
Its interessting to see the performance jumps for AMD cards in this particular game.

But ofc its just a beta test, and just 1 game.
It still doesnt tell us that much.

But that AMD cards will most likely get a bigger performance boost from DX12 then Nvidia cards is something that i predict aswell.
Still about the actual numbers that showed in the graphs, iĀ“m personaly a bit skeptical about those to be honnest.

the unreal demo with dx12 also shown my 290x beating 980ti.

1 Like

Do you have that particular article by hand aswell?
I think it would be a good value to this thread.

And iĀ“m interessted to see it.

http://hexus.net/tech/news/graphics/85385-dx12-unreal-engine-4-elemental-demo-download-available/

they didn't show any numbers, but guy who made it, is quite open for conversations.

He used;
Intel i7-4790K CPU @ 4.8Ghz, an EVGA Nvidia GTX 980 Ti @ 1366mhz core / 4100mhz memory and 32GB of DDR3-2400 quad RAM.

I used:
Intel i7-3770k CPU @ 4.2GHz, MSI LE AMD 290x @ 1150MHz core / 1500MHz memory, and 16GB of DDR-3-1600 RAM.

He's tests pointed to around ~60FPS at 1080p while mine were ~90FPS.
[vsync was off on both tests]

~ some of my comments ~
at 1440p its quite buggy dunno why... glare is odd... also going from windowed mode to full screen seems to give me hell off amount of performance boost making my median hit 115FPS

there might be some bugs in NV drivers forcing vsync i've seen some people hitting higher with 980ti but lower than my 290x.

fps supplied are not highest and lowest, they are the median of what i've seen most of the time through out the demo.

seems apex effects work without any issues on this demo on amd gpu's.

Thanks for the article.
Its very interessting exaly, but i also do see allot of various numbers.
Still its a topic on which i will definitely keep my eye on.

No one has said nvidia sucks. the benchmarks didn't show that Nvidia sucks. It just shows that the 390x and 980 are very similar cards when it comes to directx 12. The AMD cards are able to run at their FULL potential under dx12. That is the only thing that the benchmarks are showing. We will see Nvidia and AMD going head to head now and the race is on.

I know no one said "Nvidia sucks" but it's the general vibe I got from some people's posts. I think what you just said is the best way to describe these benchmarks. I do greatly dislike seeing "benchmarks" for games and cards and what not that aren't even out yet. They rarely come out to be the same as what they show.

No but any mantle exclusive features left in the code will give AMD an advantage.

the only real advantage for amd are that their gpu's are already optimized for it. While NV isn't.
NV cards are using serialized data pipe, thus is why they preform so well in dx11 env.
AMD is async data pipes, thus is why there's plenty of untapped potential there.

Even accounting for the above.. (serial vs async)

...And this might just be me being paranoid (which I am :)) but I am a little worried that nvidia might try to stall dx12 adoption by intentionally having dx11 perform better

they have the market share really and what they say goes

Dx11 is very comfortable territory for them so I dont really think they would push an api all that hard which evens the odds with their competitor.

With a competitor that is basically circling the drain all they really need to do is stall.

Anyone know whats going on here then ? Ashes of singularity dx12 bench

Is this also going to effect Vulkan the same way on NV cards?

Theoretically yes, Nvidia has always depended on Driver optimizations. but the problem now at least from what's known, the reason the Nvidia cards are getting annihilated is due to how Maxwell processes things. Maxwell is very serial heavy, Mantle/DX12 and Vulkan like Parallel Processing. Maxwell isn't good at it. the thing with the Fiji cards. is that they developed the cards to take advantage of Parallel processing. even going as far as taking a slight performance hit in DX11. if you read through that post, AMD did everything for the long term as apposed to Nvidia who set Maxwell for DX11, but don't have proper optimization for DX12. (even though it supports it) Mind you DX12 is still in "?" phase. what i mean by that is, We know games are coming for it, but gaming developers haven't said much of anything. the only game that's confirmed to have DX12 support (and I believe Vulkan.. Don't quote me on that) is Deus Ex: Mankind Divided.

Saying Nvidia cards are getting "annihilated" is a bit over dramatic isnt it?

in those DX12 benchmarks no. AMD's increase in performance is sort of expected. Mantle is their technology. it wouldn't be a surprise if AMD won those DX12 benchmarks that's why Nvidia is pissing and moaning. if it was a small amount Nvidia probably would of kept quiet.

I knew that AMD cards (GCN) were very beefy cards without good firm/software, but I never thought they would see 80% jump in performance compared to Nvidia.
If pascal isn't a compute card AMD is going to stomp nvidia with greenland.

Yeah, but with games that use Mantle, that kind of increase isn't observed, iirc. It is strange that they would see a bigger improvement with DX12 than with their own API.

I think it's cause AMD's drivers aren't that good. I'm sure if someone did a test with all the current Mantle-based games. the increase would be much more apparent. on day one mantle showed a good increase but it was a bit finicky.