Why AMD will perform better than NVIDIA in DirectX 12

[BiggieShady]

Which will limit performance gains in games since draw calls like will be limited to DX11 speeds.

Not necessarily, when you increase draw distance in a game or a vehicle count or similar settings, you automatically increase number of draw calls ... so it's certain when switching to dx12 mode in next year's games, there will be opportunity to increase settings for more eye candy at the same dx11 frame rate. Granted, that's also opportunity for devs to do more advanced (and computationally expensive) effects in dx12 mode that you can't turn off ... after all we have seen both examples with dx9/dx10/dx11 transitions.

 

[TheGuruStud]

At least I can watch videos on my 290x. 980ti will randomly lock (but can game all day OCed). Thanks Nvidia.

Tell me again how superior nvidia is. What a joke.

Nvidia cheaps out on stuff they don't need at present (and to lower power draw), like compute and hardware bits for dx12. That's not a surprise. AMD packs in tons of tech even though it won't be used in the card's lifespan.

 

[TheoneandonlyMrK]

What will be thing, I would expect that games will have DX12 low lvl in them but they will also have DX11 fall back for people still on 7/8. Which will limit performance gains in games since draw calls like will be limited to DX11 speeds. That will help AMD yes but likely won't give them that much of a lead if any. Be like probably BF4 got mantle, did give AMD only like 10% over nvidia on higher end machines.

10% more battlefield next or starwars battlefront or star citizen or Aos could well sell a few cards though.
So Amd invented and patented an Ace engine and has been doubling up on them since day one ,its not going to be easy for Nvidia to make different tech that works better and id say could take a while, they did overtake Amds early lead with tesalation though, so hopefully good times ahead ,no doubt,, because they both really need to start upping their game since 2 cards for high fps 4k is a thing still and not one i like.

 

[arbiter]

At least I can watch videos on my 290x. 980ti will randomly lock (but can game all day OCed). Thanks Nvidia.
Tell me again how superior nvidia is. What a joke.

Sounds like something wrong with your machine.

Nvidia cheaps out on stuff they don't need at present (and to lower power draw), like compute and hardware bits for dx12. That's not a surprise. AMD packs in tons of tech even though it won't be used in the card's lifespan.

Kinda hard to support something that was a locked tech by another company. AMD fans love to point that out when talking about PhysX. Ironic that AMD pulled the same thing.

10% more battlefield next or starwars battlefront or star citizen or Aos could well sell a few cards though.

Thing is with all the questionable marketing and lies AMD has told last 4-5 years, People will be gun shy about touching AMD hardware since you don't know if what they claim about the card is the truth. Example like the claimed performance of Fury X vs 980ti and most likely Nano vs 970(its easy to see the claims won't hold up to what AMD wants people to think)

 

[Aquinus]

I think this needs to be inserted here:

AMD and NVidia both make good products. They each have their advantages and their flaws. NVidia undoubtedly makes the currently best performing GPU out right now. That is not to say that it's the most feature filled or the best cost.

It's true though, nVidia trimmed the edges in order to squeeze out a very performant core for what it can do. AMD isn't bad, however, there are a lot of things you probably wont need... at least, not yet.

When push comes to shove, I think my 390 handles games pretty well. If I got a 390x, Fury, 970, 980, or 980 Ti, I bet I would still be happy.

Without knowing what games in the future are going to demand out of games makes it kind of hard to know if the existence or lack of features will make a big difference or not. In the case of Async processing, AMD already had something, (ACE.) It's nice that AMD has it but, for most modern games it doesn't mean a whole lot.

AMD's only advantage is that they've been thinking about GCN for a while and they intended it to be more than simply a GPU. There are a lot of features that tout GCNs compute ability.

So enough with the bashing. New technology is new. Before we know it, nVidia will have something like ACE on their next lineup of cards. The question will be if AMD will have made any substantial change to combat it. Given their budget, I doubt it.

 

[the54thvoid]

What a few people probably want to know is how does Kepler do with DX12 Async?

I found this powerpoint about CUDA and it's so far over my head I may as well go to bed and dream of ice cream.

http://its.unc.edu/files/2014/11/UNC02_Fundamental_CUDA_Optimization.pdf

 

[TheoneandonlyMrK]

Sounds like something wrong with your machine.

Kinda hard to support something that was a locked tech by another company. AMD fans love to point that out when talking about PhysX. Ironic that AMD pulled the same thing.

Thing is with all the questionable marketing and lies AMD has told last 4-5 years, People will be gun shy about touching AMD hardware since you don't know if what they claim about the card is the truth. Example like the claimed performance of Fury X vs 980ti and most likely Nano vs 970(its easy to see the claims won't hold up to what AMD wants people to think)

The sun does note shine out of nvidias ass, nvidia ,intel Amd ,Qualcomm all make statements upon what they expect they will be selling before its in their hands to sell and they could all be called liers legitimately by zelous fanbois for what they have sometimes delivered.

 

[arbiter]

I am seriously lose for words how people expect Nvidia to handle an locked AMD tech since 7000 series, How does people expect Nvidia to handle that perfectly when likely added at last minute so Maxwell couldn't been updated for it.
It would been like PhysX being made DX12 standard and expect AMD cards to do it perfectly.

 

[Mussels]

I am seriously lose for words how people expect Nvidia to handle an locked AMD tech since 7000 series, How does people expect Nvidia to handle that perfectly when likely added at last minute so Maxwell couldn't been updated for it.
It would been like PhysX being made DX12 standard and expect AMD cards to do it perfectly.

its not AMD locked at all - it was designed around the DX11 specs with extras on top. AMD have a feature advantage this generation, which will even out as soon as nvidia release new cards. thats it, a one generation advantage. no different to any previous DirectX release where one company got out first.

 

[arbiter]

its not AMD locked at all - it was designed around the DX11 specs with extras on top. AMD have a feature advantage this generation, which will even out as soon as nvidia release new cards. thats it, a one generation advantage. no different to any previous DirectX release where one company got out first.

it was locked til DX12 adopted it now its not. But that would be like PhysX being added in, i bet AMD fans would be screaming since amd cards won't be able to run it. But can't expect nvidia card to support something that was looking like wasn't added til after maxwell was finalized design and put in to production.

I bet AMD lobbied to heck to get it added in after maxwell was final knowing it wasn't gonna be able to do as well as their card.

 

[Steevo]

Is it time to drag out the Nvidia has feature level 12.3 bullshit threads yet?

 

[Moofachuka]

Even if they do, and that's a big if, since AMD has had a good headstart on designing hardware for DX12, as opposed to DX 11, it's not a big deal either way. You see the graph, based on the ONE benchmark/game out now, that shows the the Nvidia top dog, not designed for DX12, and the AMD top dog, designed for DX12, are not very far apart.

I foresee AMD improving, and Nvidia having Pascal designed for DX 12, and that there won't be a heck of alot of difference between the competitors at most price points. Personally, I think it's a helluva alot of noise being made on both sides about the issue right now, when there is one DX12 game and only a handful of DX12 games by the time Pascal drops.

I don't think either side's strident and zealous believers can really claim anything at this point.

kinda late reply, but even if Pascal's DX12 will not have a lot of difference than AMD, their DX11 games overall beat AMD. so if Pascal beats AMD in DX11 and DX12 near par against AMD, I would choose Pascal in this case.

 

[arbiter]

kinda late reply, but even if Pascal's DX12 will not have a lot of difference than AMD, their DX11 games overall beat AMD. so if Pascal beats AMD in DX11 and DX12 near par against AMD, I would choose Pascal in this case.

reumors on pascal was 5000-6000 cuda cores so would be a monster on top of HBM2

 

[Makaveli]

it was locked til DX12 adopted it now its not. But that would be like PhysX being added in, i bet AMD fans would be screaming since amd cards won't be able to run it. But can't expect nvidia card to support something that was looking like wasn't added til after maxwell was finalized design and put in to production.

I bet AMD lobbied to heck to get it added in after maxwell was final knowing it wasn't gonna be able to do as well as their card.

What are you talking about???

 

[Aquinus]

You know, for all of you who bash AMD, I must say, I'm fairly happy with what my 390 has delivered to me so far. For the price, it's certainly not a bad card. While AMD's products aren't perfect, I think people are being a little too critical of them lately. It's always been true that GCN has been a compute-oriented architecture. Nvidia made it very clear what it was going for when they slashed double precision math performance and started going gung-ho on the parts of the GPU games were using the most. The gains to be had from async compute makes a lot of sense considering how GCN was designed and I think that as these kinds of uses are realized, we'll see a shift back to some more compute oriented hardware out of nVidia. The only requirement is if the market demands it, which means that these improvements have to show some significantly tangible gain to be realized. If AMD's GPUs could actually handle 8GT/s GDDR5 ICs, I think we would see a much different card as, at least my 390, seems to scale almost linearly with memory clock speed. HBM might have been an attempt to improve the situation but GCN might very well be hungry for lower latency and not higher bandwidth.

My only complaint would be the amount of power the GPU suck down but, idle consumption us a whole lot better over the 6870s, so I can't complain too much since my GPU spends a lot of time idling, such as while I write this post.

 

[cadaveca]

it was locked til DX12 adopted it now its not. But that would be like PhysX being added in, i bet AMD fans would be screaming since amd cards won't be able to run it. But can't expect nvidia card to support something that was looking like wasn't added til after maxwell was finalized design and put in to production.

I bet AMD lobbied to heck to get it added in after maxwell was final knowing it wasn't gonna be able to do as well as their card.

Asynchronous Compute has been part of AMD's design since the X1900 days. It's merely an oversight that had NVidia not optimize for it, but given that the tech itself is actually pretty old, it is not surprising that NVidia have designs that don't work well doing an updated version of old tech.

It is also why AMD performs better than NVidia in such workloads...it has been part of AMD GPU design for the last decade. The whole idea that is being presented by this topic is rather humorous to me, actually, since these design differences are really what makes AMD and NVidia GPUs different from each other. A lot of this "knowledge" about GPU design has been forgotten, it seems.

The R520 architecture is referred to by ATI as an "Ultra Threaded Dispatch Processor".

 

[terroralpha]

It's a synthetic benchmark. Nobody cares.....

that's a game benchmark, not a synthetic benchmark. like the post stated. jack ass.

 

[arbiter]

You know, for all of you who bash AMD, I must say, I'm fairly happy with what my 390 has delivered to me so far. For the price, it's certainly not a bad card. While AMD's products aren't perfect, I think people are being a little too critical of them lately. It's always been true that GCN has been a compute-oriented architecture.

Sadly AMD has given people a LOT of Ammo to end users and reviews for that matter to be critical of them. Look at all the PR claims over last 4 years and the twitter rant of of AMD Roy recently. AMD's PR staff digs the company a hole and has been digging for years.

that's a game benchmark, not a synthetic benchmark. like the post stated. jack ass.

Its an Alpha AMD sponsored game so reality is results for it ATM are not any better then a synthetic benchmark.

 

[cadaveca]

Its an Alpha AMD sponsored game so reality is results for it ATM are not any better then a synthetic benchmark.

THat's a rather moot point, though. You'd have a better stance simply stating that it is just a single title, and doesn't represent all titles, and ergo, does not say that AMD will be better in DX12.

 

[Arjai]

tautology

Nice. This whole post is tautologous.

 

[arbiter]

THat's a rather moot point, though. You'd have a better stance simply stating that it is just a single title, and doesn't represent all titles, and ergo, does not say that AMD will be better in DX12.

i have that said in fact like 1000 times, but yet it seems to mean nothing and gets same response you gave.

 

[cadaveca]

i have that said in fact like 1000 times, but yet it seems to mean nothing and gets same response you gave.

meh. Sorry, but I did not read those posts, or if I did, I did not recall. But nice to know that we agree.

 

[darkangel0504]

hey, I have a question : " will GCN 1.0 get benefit from DX 12 ? "

 

[R-T-B]

Wow did I just walk into the flame-each-other festival or something? Lotsa insults being thrown around. Keep it clean guys.

 

[RCoon]

Wow did I just walk into the flame-each-other festival or something? Lotsa insults being thrown around. Keep it clean guys.

The title says "AMD" "NVidia" and "better". This was always going to turn into an ugly mess of nonsense. I've begun to just assume very little of worth ends up coming out of these "discussions"