• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Directx 12 and async compute

Soldato
Joined
19 Feb 2011
Posts
5,849
For the games you play a 480 is a much better value proposition than a 1080, especially going forward. The 1080 is a fantastic piece of hardware but ultimately Nvidia is not seeing gains in DX12/Vulkan but AMD does, and this automatically ups the value of an AMD card by comparison. So while the 1080 is absolutely the faster card, is the price you pay for it worth it? Imo no, especially for 1080p 60fps. I think a much smarter option is to go for a custom 480 this month and then once the Vega cards come out and maybe the 1080ti you can sell the 480 off and buy one of those instead or a (by then) cheaper 1070/1080.

If nothing else the money you keep from going 480 instead of 1080 can let you upgrade CPU which will mean much more for those games.

Exactly this, for the price your going to pay for a 1080, you can upgrade your GPU and CPU, or GPU and Monitor etc, id personally drop in an i7 if i was you, or if your happy with your current CPU then opt for a new monitor, i recently bought a Freesync screen and let me tell you, its night and day the difference between having Freesync and not having it.
 
Associate
Joined
3 Oct 2009
Posts
596
Everything is pointing towards Nvidia still having no answer for AMDs performance in DX12, especially in titles likes the newly updated tomb raider. I say this as a MSI 1080 Gaming X owner.

The smart money is on a 480, anyone buying a 1080 currently is spending on something which is painfully expensive and doesn't offer great value when you consider future games if the trend was to remain true for DX12.

Only buy a 1080 if you have the money, can comfortabley afford it and don't mind the fact you're paying far to much for what it is.. Anyone who thinks otherwise is kidding themselves.
 
Associate
Joined
27 Dec 2014
Posts
1,686
Location
Southampton
Everything is pointing towards Nvidia still having no answer for AMDs performance in DX12, especially in titles likes the newly updated tomb raider. I say this as a MSI 1080 Gaming X owner.

The smart money is on a 480, anyone buying a 1080 currently is spending on something which is painfully expensive and doesn't offer great value when you consider future games if the trend was to remain true for DX12.

Only buy a 1080 if you have the money, can comfortabley afford it and don't mind the fact you're paying far to much for what it is.. Anyone who thinks otherwise is kidding themselves.

you sound very familiar, too familiar even so I am not going to say you started a new account just so you know say the same thing under a different name ... and sorry I have to call it, you own a 1080 like I own a pink unicorn.
 
Associate
Joined
8 May 2014
Posts
2,288
Location
france
Hi all,

I ordered a Gigabyte g1 1080 a while ago and have a delivery date for the end of next week.
Since it's been on order i have read on the net that it doesn't do Directx 12 or async compute properly, where as amd 480 does. Would i be better cancelling and waiting for the quicker amd stuff?
My system currently has a 2500k processor and im gaming at 1080p, sold my hd5770 so without a card at the moment but could wait till the end of he year early next. Stuff i will be playing will be the command and conquer series, dawn of war series,company of heroes series, supreme commander series, homeworld series, ashes of a singularity, grey goo and metal gear solid v. i may also upgrade to a 1440p or 4k monitor in the future.

any advise appreciated:)

i see you are more an RTS gamer, much more cpu hungry btw in these kind of games, the 1080 is complete overkill for what you are going to be playing also, for 1080p resolution and old games, the 480 wouldn't have any issue pushing 120 fps.
personally i would have ordered the 480, and used the extra money to upgrade my cpu to something like an i7, keep in mind that most RTS will try to be on dx12/vulkan specificaly to laverage the multi-threading with these API, more so than other adventuer, fps, racing etc, so chances are best up coming RTS games would be on either of these APIs, so arm yourself with a good multi core cpu, and a 480, that you can upgrade later to 490 or whatever, knowing it's price it would barely lose 50£ of it's price in 2nd hand market.
major RTS studios added dx12 to their engine
firaxis(civ series)
sega, creative assembly, relic ( total war, company of heroes, warhammer series )
oxide games ( ashes )
microsoft ( halo wars )
total war warhammer just patched dx12, halo wars 2 will be dx12, civ 6 will be dx12
thats my opinion, cheers!
 
Last edited:
Soldato
Joined
27 Feb 2015
Posts
12,596
As far as I am concerned as long as I am playing games in dx9 and dx11, then dx12 is irrelevant.

Whilst I do have a passing interest in dx12 and vulkan development, they have zero impact on my purchase decision.
 
Associate
Joined
3 Oct 2009
Posts
596
you sound very familiar, too familiar even so I am not going to say you started a new account just so you know say the same thing under a different name ... and sorry I have to call it, you own a 1080 like I own a pink unicorn.

You do realise my account is 7 years old right?

I guess my ranked 9th score on the Fire strike Ultra thread with an MSI 1080 Gaming X is fiction also?

Gotta love a conspiracy theory... I wonder who he thinks I am lol
 
Associate
Joined
27 Dec 2014
Posts
1,686
Location
Southampton
You do realise my account is 7 years old right?

I guess my ranked 9th score on the Fire strike Ultra thread with an MSI 1080 Gaming X is fiction also?

Gotta love a conspiracy theory... I wonder who he thinks I am lol

just look around the forum a little bit, you'll see quickly who I think you are, sorry can't help it, it's like looking in the mirror :).

I just don't see how anyone can still be saying stuff like this when all you have to do is run the game. Then you wouldn't need to point to any "evidence" but your own. Plus if you really believed what you are preaching, you would have sent that "1080" back and got an 480. Just saying ...
 
Associate
Joined
3 Oct 2009
Posts
596
just look around the forum a little bit, you'll see quickly who I think you are, sorry can't help it, it's like looking in the mirror :).

I just don't see how anyone can still be saying stuff like this when all you have to do is run the game. Then you wouldn't need to point to any "evidence" but your own. Plus if you really believed what you are preaching, you would have sent that "1080" back and got an 480. Just saying ...

You think after 7 years I'd go around creating new accounts to say the same thing? Do you wear a tin foil hat aswell? No idea who you have me confused with but I do find it kind of amusing.

The 480 was a disappointment for me, I had a 290x and as a GPU it doesn't fall into the category I'd buy. Then again, DX12 improvements and lack there of for Nvidia are making it a more appealing card for many.

Whichever way you slice it, up, down, left or right. The 1080 is not a great buy, it just isn't. It's a £400-£450 card being sold for £580-£700. Anyone buying it is mad suggesting otherwise. This is further compounded but Nvidias inability to address the DX12 gap with Pascal, with worrying trends emerging ln recently patched games. It is however the only option currently for somebody who wants a top end card, as AMD are offering nothing in its category.
 
Soldato
Joined
14 Jul 2004
Posts
2,548
The difference is RX480 is at the price it will be in 12-18 months time still because it's priced as a 230mm^2 finfet card should be(they are unfortunately more expensive than a 230mm^2 28nm card would be, and exchange rate exaggerates the difference).

The 1080 is currently priced as a 500mm^2+ core, but in 12-18 months it will be £400 or less, meaning you're overpaying for a 1080 by 50% currently, while with a RX480 you aren't.

We know. I've never had a flagship that didn't lose half its value in a year. It's always been that way. You pay a big premium for top performance, but on the other hand you get a product that could last 2 - 3 years. I paid a frankly daft price for a GTX 780 in 2013, but I've only just replaced it. In the end it was phenomenal value, and come to think of it, even now its not that far off a 480 in terms of grunt.

The simple truth is a 480 will start to struggle much earlier than a 1080. Its way cheaper, but like you said, many of us would view it as a stopgap for Vega, which will probably cost a bomb when it finally shows up.

It's not about what your card is gonna be worth in 1 or 2 years, it's about playing games with the best performance you're willing to pay for at the time. We could argue that the guy who buys flagship GPUs every year is a dumbass, but it's his business at the end of the day. An enthusiast will find a way to justify it pretty easily.
 
Associate
Joined
30 May 2016
Posts
620
The fact that the RX48-0 still see a big boost when using async shaders merely suggests that AMD still have major problems achieving anything close to 100% utilization under DX11 and so the architecture improvements in the new GCN 1.3 are pretty minimal, which si backed up by the lack luster performance.

The fact that Pascal doesn't see a bug improvement with async shaders is much less interesting than the fact that the RX480 does. IMO.

AMD will never achieve 100% utilisation in DX11. The same goes for OpenGL. This was a decision AMD made a while back and it is not going to change. It was all about COST: the didn't have enough money to develop top-notch drivers.

Think about it like this: if it costs $100 to produce a card that you sell for $150, then your profit is $50 per card, right? But that's not the end of it. You need to release drivers. Let's say you sell 1 million of those cards over 2 years. That's $50m profit. But how many man-hours of software development do you spend over those two years to make drivers and optimise per game, or to 'sponsor' games so they work well on your cards? If this costs $40m then you're left with $10m profit. These are random numbers but I'm just trying to make the following point:

AMD decided that it can't afford to pay $40m for drivers/sponsoring. So DX11/OpenGL fell back compared to NVidia and the latter was much faster. What did AMD do instead? They said let's build cards that cost $130 to produce but still sell them for $150. After selling 1 million of those cards we'll have $20m profit which is more than $10m we'd have by investing in DX11/OpenGL. Basically they threw more hardware at the problem to keep up with the better NVidia performance (due to better NVidia software) and even then their cards lost on average. When you compare like for like (e.g. GTX1060 is at 4TFLOPS vs RX480 that is over 5TFLOPS ) the AMD card has more processing power but the NVidia card is still faster in DX11/OpenGL games (it's only because of the extra processing power that the AMD card can stay close).

Enter 2016: what is (finally) changing is that games are starting to support DX12 and Vulkan. As games start using these new APIs, the disadvantage of AMD drivers is removed and furthermore the 'extra unutilised hardware' starts getting utilised. This is true even for older AMD cards.

So basically, async compute is just a small part of the equation and the reality of DX12/Vulkan is much more complex. Async just gives an extra advantage to AMD for certain workloads (which are actually not rare, so it's significant).

As far as the original question (should I still get a GTX1080 after seeing how well the RX480 does): if you can afford the GTX1080 then it'll give you the best performance right now. It is in a different class than the RX480 so even with async/Vulkan/DX12 the RX480 will never get close to the 1080.

Is the GTX1080 a good investment? Not so much. It's way overpriced because there is no competition for it since Vega is still 6 months away.

You will enjoy the 1080 for the next 6 months. Once Vega is released, you will start thinking that you paid way too much for it. But if money is not an issue for you, it won't matter that much.

EDIT: like other people pointed out, there's also a 'missed opportunity' cost. With the money you sacrifice on the 1080 you could instead invest in a better monitor or CPU or whatever. I would seriously consider their advice. Lots of good ideas through above (e.g. an RX480 with a FreeSync monitor combo).
 
Last edited:
Associate
Joined
8 May 2014
Posts
2,288
Location
france
AMD will never achieve 100% utilisation in DX11. The same goes for OpenGL. This was a decision AMD made a while back and it is not going to change. It was all about COST: the didn't have enough money to develop top-notch drivers.

Think about it like this: if it costs $100 to produce a card that you sell for $150, then your profit is $50 per card, right? But that's not the end of it. You need to release drivers. Let's say you sell 1 million of those cards over 2 years. That's $50m profit. But how many man-hours of software development do you spend over those two years to make drivers and optimise per game, or to 'sponsor' games so they work well on your cards? If this costs $40m then you're left with $10m profit. These are random numbers but I'm just trying to make the following point:

AMD decided that it can't afford to pay $40m for drivers/sponsoring. So DX11/OpenGL fell back compared to NVidia and the latter was much faster. What did AMD do instead? They said let's build cards that cost $130 to produce but still sell them for $150. After selling 1 million of those cards we'll have $20m profit which is more than $10m we'd have by investing in DX11/OpenGL. Basically they threw more hardware at the problem to keep up with the better NVidia performance (due to better NVidia software) and even then their cards lost on average. When you compare like for like (e.g. GTX1060 is at 4TFLOPS vs RX480 that is over 5TFLOPS ) the AMD card has more processing power but the NVidia card is still faster in DX11/OpenGL games (it's only because of the extra processing power that the AMD card can stay close).

Enter 2016: what is (finally) changing is that games are starting to support DX12 and Vulkan. As games start using these new APIs, the disadvantage of AMD drivers is removed and furthermore the 'extra unutilised hardware' starts getting utilised. This is true even for older AMD cards.

So basically, async compute is just a small part of the equation and the reality of DX12/Vulkan is much more complex. Async just gives an extra advantage to AMD for certain workloads (which are actually not rare, so it's significant).

As far as the original question (should I still get a GTX1080 after seeing how well the RX480 does): if you can afford the GTX1080 then it'll give you the best performance right now. It is in a different class than the RX480 so even with async/Vulkan/DX12 the RX480 will never get close to the 1080.

Is the GTX1080 a good investment? Not so much. It's way overpriced because there is no competition for it since Vega is still 6 months away.

You will enjoy the 1080 for the next 6 months. Once Vega is released, you will start thinking that you paid way too much for it. But if money is not an issue for you, it won't matter that much.

EDIT: like other people pointed out, there's also a 'missed opportunity' cost. With the money you sacrifice on the 1080 you could instead invest in a better monitor or CPU or whatever. I would seriously consider their advice. Lots of good ideas through above (e.g. an RX480 with a FreeSync monitor combo).

it's not that simple AMD's design was a hybrid to be good in compute and graphics, because it was designed with consoles in mind, the semi-custom business depanded on it, and doesnt matter how good the drivers are, they just cannot overcome a the limitations of a crappy API... i thought dx12/vulkan proof of concept showed ppl by now, but apparently not clear enough for everyone!
 
Last edited:
Associate
Joined
30 May 2016
Posts
620
it's not that simple AMD's design was a hybrid to be good in compute and graphics, because it was designed with consoles in mind, the semi-custom business depanded on it, and doesnt matter how good the drivers are, they just cannot overcome a the limitations of a crappy API... i thought dx12/vulkan proof of concept showed ppl by now, but apparently not clear enough for everyone!

I'm not sure we disagree. Yes, semi-custom depended on this because consoles had very weak CPUs (hence it was essential to be able to do compute on GPU side like you said).

However, DX11 and OpenGL were NOT a requirement for consoles. Both PlayStation and XBox used custom APIs for that exact reason. So they never had this problem.

It was only on PC where the problem arose, because DX11/OpenGL left the 'optimisation' task to the DRIVER (which AMD/NVidia write) and AMD decided it's not worth the effort. I think they made the right decision and 2017 will be the year everyone realises it, because games are being pushed to DX12/Vulkan (where the driver now matters much less).
 
Associate
Joined
8 May 2014
Posts
2,288
Location
france
I'm not sure we disagree. Yes, semi-custom depended on this because consoles had very weak CPUs (hence it was essential to be able to do compute on GPU side like you said).

However, DX11 and OpenGL were NOT a requirement for consoles. Both PlayStation and XBox used custom APIs for that exact reason. So they never had this problem.

It was only on PC where the problem arose, because DX11/OpenGL left the 'optimisation' task to the DRIVER (which AMD/NVidia write) and AMD decided it's not worth the effort. I think they made the right decision and 2017 will be the year everyone realises it, because games are being pushed to DX12/Vulkan (where the driver now matters much less).

again it's not that AMD decided not to bother with the drivers, it literally cannot do otherwise, the cost of developing another architecture is not sustainable, because the market isn't big enough to cover it, for that to be profitable they would need like 100% of market share for both pc and console.
not that driver matters much less, but rather devs now have access to features and hardware control they didnt have before, and that drivers couldn't solve, it had to be the API.
 
Soldato
Joined
19 Feb 2011
Posts
5,849
OP should read this

https://www.reddit.com/r/nvidia/com...ed_about_asynch_computeshaders_nvidia/d59c0qs

Basically, right now, in DX11, Nvidia cards are ahead of AMD cards because AMD cards have hardware sitting unused as their software drivers simply cannot tap into it, however this hardware under Vulkan and DX12 is able to be utilised, giving much much much more performance, on Nvidias side, they are already using their hardware fully in DX11, that is why under DX12 and Vulkan there is little gain, due to the hardware already being utilised and the method these 2 API's leverage the hardware AMD has sitting unused.

So basically, if you want a card for the here and now, buy an Nvidia, if you want a card thats a bit more futureproof and forward looking, buy AMD, i think eventually Nvidia may take the route AMD has or a hybrid of the 2, especially when DX12 and Vulkan games become the norm and they start to lag behind.
 
Associate
Joined
15 Jul 2015
Posts
32
Stuff i will be playing will be the command and conquer series, dawn of war series,company of heroes series, supreme commander series, homeworld series, ashes of a singularity, grey goo and metal gear solid v.any advise appreciated:)

None of these games require the power of a GTX 1080, especially not in 1080p. Besides if you're not in a rush why purchasing a card for premium price?
 
Soldato
Joined
25 Jun 2011
Posts
5,468
Location
Yorkshire and proud of it!
Regardless of relative merits, you're gaming at 1080p which means the Nvidia 1080 (any variant) is massive overkill. You could buy a 480, still have good results, and put the saved money by to resell the 480 when Vega is released and get a probably much better card for only a small loss and almost zero risk. Or you could keep that money back and put it towards a HDR monitor when those start to emerge later in the year as I think that (with a compatible GPU such as the 480) will deliver you far better gains than using a 1080 to power your current resolution. I would FAR rather go from non-HDR to HDR than move from 60fps to 120fps. The latter of which I'd barely notice.
 
Last edited:
Back
Top Bottom