AMD @ GDC: Partnership with MS next-generation graphics.

[AnandThenMan]

That article is 3 years old. Also it's hard to crash a Windows machine/D3D application? HAHA no.

 

[3DVagabond]

That article is 3 years old.

OMG! Talk about irrelevant.

 

[frozentundra123456]

I don't know about AnandThenMan but i'm never using DX ever again where Mantle is available. by comparison its horrible.

Well, that rules out a whole 2 games. DX is doomed.

 

[frozentundra123456]

Nope. Thats yet another selfproclaimed hype.

I don't see how mantle fans keep saying DX 12 is a copy of mantle. I don't technically know how similar they are, but there is one *huge* difference between them, and that is that DX is hardware agnostic. However similar they are otherwise, that is a game changing difference if DX 12 can deliver similar performance improvements to mantle.

 

[ASM-coder]

one *huge* difference between them, and that is that DX is hardware agnostic.

But DX12 is not.

 

[3DVagabond]

Well, that rules out a whole 2 games. DX is doomed.

That's miles ahead of one tech demo. A demo from a game that didn't seem to run any better on a GK110 than it runs on an XB1.

I don't see how mantle fans keep saying DX 12 is a copy of mantle. I don't technically know how similar they are, but there is one *huge* difference between them, and that is that DX is hardware agnostic. However similar they are otherwise, that is a game changing difference if DX 12 can deliver similar performance improvements to mantle.

The huge difference is Mantle actually exists on consumers' PC's.

 

[frozentundra123456]

But DX12 is not.

So are you trying to say DX12 will not support amd, Intel, and nVidia graphics hardware, in contrast to mantle?

 

[AnandThenMan]

Well, that rules out a whole 2 games. DX is doomed.

In it's current form, essentially yes. Microsoft would follow the DirectX status quo otherwise as they've done for over a decade.

I don't see how mantle fans keep saying DX 12 is a copy of mantle. I don't technically know how similar they are, but there is one *huge* difference between them, and that is that DX is hardware agnostic. However similar they are otherwise, that is a game changing difference if DX 12 can deliver similar performance improvements to mantle.

Hardware agnostic does hurt a thin API, it is very easy to slip into "that's good enough because it supports everything" mode. That's how we got here in the first place. Either way, if DX12 manages to follow the general model of Mantle (and by all accounts it's extremely similar if not identical) it will be far better than it is now, very welcome.

But we are years out from seeing DX12, a lot can change in that time frame. MS may not get DX12 to work acceptably given the need to support a wide range of hardware. There is bound to be at least some compromise. AMD will run into the same issues going forward with Mantle, which is why they wisely chose to initially cut off Mantle at GCN and not pursue previous architecture support.

AMD is playing both sides of the table here, they are contributing to DX12, and at the same time they are pushing Mantle with game studios and the console crossover titles. That's a lot of influence.

 

[PPB]

I think he has no idea when he talks about api or software for that matter being hardware agnostic. The moment pre GCN and pre fermi is ruled out of dx12 compatibility you arent being hardware agnostic at all.

You can target a wider range of hardware, but you cant be hardware agnostic, its downright silly.

Mantle support is feature dependant, and it currently doest support other vendors because those vendors wont give the credit to amd for making a better api. Thankfully ms just saved nv and intel the hassle of doing that as they got a free license of mantle under a different name.

 

[frozentundra123456]

In it's current form, essentially yes. Microsoft would follow the DirectX status quo otherwise as they've done for over a decade.

Hardware agnostic does hurt a thin API, it is very easy to slip into "that's good enough because it supports everything" mode. That's how we got here in the first place. Either way, if DX12 manages to follow the general model of Mantle (and by all accounts it's extremely similar if not identical) it will be far better than it is now, very welcome.

But we are years out from seeing DX12, a lot can change in that time frame. MS may not get DX12 to work acceptably given the need to support a wide range of hardware. There is bound to be at least some compromise. AMD will run into the same issues going forward with Mantle, which is why they wisely chose to initially cut off Mantle at GCN and not pursue previous architecture support.

AMD is playing both sides of the table here, they are contributing to DX12, and at the same time they are pushing Mantle with game studios and the console crossover titles. That's a lot of influence.

So what is your answer to some compromise in performance for the ability to use all major graphics hardware? I only see two possible solutions.

One is an API war between amd, nVidia, and Intel. The other is a hardware monopoly by one of the 3 hardware manufacturers. Are either of those alternatives attractive?

 

[frozentundra123456]

I think he has no idea when he talks about api or software for that matter being hardware agnostic. The moment pre GCN and pre fermi is ruled out of dx12 compatibility you arent being hardware agnostic at all.

You can target a wider range of hardware, but you cant be hardware agnostic, its downright silly.

Mantle support is feature dependant, and it currently doest support other vendors because those vendors wont give the credit to amd for making a better api. Thankfully ms just saved nv and intel the hassle of doing that as they got a free license of mantle under a different name.

You are putting words into my mouth. Obviously no API will support *all* hardware. When I say hardware agnostic I mean that it will work an proper hardware from all 3 major manufacturers. Sounds like just sour grapes that nvidia and Intel will not be forced to support mantle.

 

[3DVagabond]

So what is your answer to some compromise in performance for the ability to use all major graphics hardware? I only see two possible solutions.

One is an API war between amd, nVidia, and Intel. The other is a hardware monopoly by one of the 3 hardware manufacturers. Are either of those alternatives attractive?

How about we use Mantle today (and for the next couple of years) where we can and DX/OpenGL where we can't and let 2016 decide what we are going to use then? The industry could shift to Linux by then. :sneaky:

 

[AnandThenMan]

How about we use Mantle today (and for the next couple of years) where we can and DX/OpenGL where we can't and let 2016 decide what we are going to use then? The industry could shift to Linux by then. :sneaky:

Free from the shackles of Windows is an invertible step IMO, it's not a matter of if but when. And yes I see no reason why Mantle can't happily coexist with DX/OpenGL. BF4 supports two APIs and the horsemen of the apocalypse didn't show up at my door.

 

[frozentundra123456]

How about we use Mantle today (and for the next couple of years) where we can and DX/OpenGL where we can't and let 2016 decide what we are going to use then? The industry could shift to Linux by then. :sneaky:

That is most likely what will happen. To be clear, I have no problem with mantle as a value add performance increase for amd user so long as amd/developers are willing to devote the resources to support it. That still does not change the fact that only supporting hardware from one manufacturer is a huge disadvantage. I also don't think the huge amount of criticism directed toward DX in these forums is justified. It is not perfect, and undoubtedly needs to be more efficient. On the other hand, it allows me to choose hardware from any manufacturer I want and to play any game I want at decent settings on a relatively moderate system.

 

[Obsoleet]

News Flash, the PC industry is shrinking. Even Intel are struggling to keep up sales of their CPU's.

Its not what it once was, less people are spending large amounts of money on hardware to play games, and who can blame them when all they read in places like this is that the only option they have is to buy a Motherboard and CPU which alone costs as much as a PS4, and then the rest on top of that...

And it will only get worse unless PC Gaming can be more affordable.

Yup. But it is affordable. You can put together an A10-7850K rig for much cheaper than a full blown Intel i5/i7 + standalone card.

With graphics and games kind of plateauing for a while now, I think an A10 is a great choice. There are plenty of 13 year old elitists who disagree that it's 'fine', but it really is. Kaveri benchmarks have it doing well in newer games at 720 and 1080 with no AA/AF. Walking away with a $180 CPU/GPU with Mantle and TrueAudio isn't that bad in my book, definitely can't get that anywhere else.

I'm still using a Q9450 + 5870, and I'm sure I spend more time gaming than 90% of the forum population here. Meaning, I don't need more to meet my gaming needs. So an A10 to start with, and then put in a R7 or R9 if you need more power is a reasonable solution IMO.

CPUs are way overpowered. I don't buy the nonsense that everyone needs a top end Intel CPU + $500 video card. It's not true, I'm proof of that. My next system likely will be a Carrizo. No reason why it won't do the job especially with Mantle support.

 

[PPB]

CPUs are way overpowered. I don't buy the nonsense that everyone needs a top end Intel CPU + $500 video card. It's not true, I'm proof of that. My next system likely will be a Carrizo. No reason why it won't do the job especially with Mantle support.

Yeah, people just dont realize that instead of calling an API (ie DX11 according to the green appologists here) to work "just fine" because you "only" need a $300 CPU to barely stay at 60 fps on the majority of current gen games and blaming the rest of the lower performing CPUs for being "weak", they should start blaming the API for being weak and the one to be pointed at.

People probably need to play more MMOs to know the sad state of DX11 in CPU overhead... its just plain silly to be at 40% GPU load all the time with all the eyecandy enabled and MSAA x4 and not being able to reach the 60fps target. Thats when you realize that most people here really havent gamed in a long time, the "throw more hardware to the problem" became obsolete and people shouldnt be forced to shell more on their hardware to compensate such an awfully optimized API as DX is in its current state.

 

[Carfax83]

People probably need to play more MMOs to know the sad state of DX11 in CPU overhead... its just plain silly to be at 40% GPU load all the time with all the eyecandy enabled and MSAA x4 and not being able to reach the 60fps target. Thats when you realize that most people here really havent gamed in a long time, the "throw more hardware to the problem" became obsolete and people shouldnt be forced to shell more on their hardware to compensate such an awfully optimized API as DX is in its current state.

A low GPU load doesn't necessarily indicate a CPU bottleneck. It could be that the game just isn't very demanding on the GPU, which most MMOs don't tend to be. Or similarly, if V-sync is enabled, it can lessen GPU load since the frame rate is limited.

MMOs if anything though tend to be more CPU demanding, due to all the player status tracking and updating and behind the scenes calculations that the CPU has to do.

 

[sontin]

It seems to me that Mantle has shown that DX is pretty good at most of what it does. Mantle is only showing notable gains in draw calls and multithreading. That is to say, Mantle is fixing a more recent problem with the CPU limiting what can be done.

So I wouldn't say DX was so bad, but it does have a problem that needs fixing now, and it appears they are taking steps to fix it.

Up to now Mantle hasn't shown any benefits to multithreading.

 

[VulgarDisplay]

Up to now Mantle hasn't shown any benefits to multithreading.

what?

 

[f1sherman]

By the way, I found this article from 2011 very interesting. It proved me wrong about how long AMD has been talking about doing something like Mantle, but it also shows some inconsistent statements from AMD. It could very well just be some PR to smooth over things with MS, since Mantle was so far away. Also, check out who may have been working together the whole time, and aren't exactly game studio "wins". It does make me see AMD's motivation for Mantle in a more positive light. :thumbsup: I had my money on Mantle being a rushed response to MS's 11.2 tiled resources/multithreading improvement announcement.

AMD Walks Back Comments On DirectX Performance Impact



http://www.gamasutra.com/view/news/33723/AMD_Walks_Back_Comments_On_DirectX_Performance_Impact.php

Am I reading too much into it, or this debunks the theory that developers had been begging IHVs, and only AMD had responded positively.
To me it sounds more like AMD had been trying to brake away from MS/DX and gathering dev support.

After an interview in which he said DirectX was "getting in the way" of using the true power of PC graphics hardware, AMD GPU division worldwide developer relations manager Richard Huddy has re-stated the company's commitment to Microsoft's PC game creation standard.

In an interview with technology news site CRN, Huddy said only a select few developers had expressed to him a desire to go around the current driver standards, including Battlefield developer DICE and Crysis developer Crytech.

Feel free to represent , but honest answers plx

 

[f1sherman]

Whats the alternative:
DICE and Crytek utterly disappointed in DX go from door to door imploring for more efficient API, even if at lunch it would empower only what... 10% of hardware?

I know this has been beaten to death, but still.... ^^

Huddy pointed out that hardware makers and API makers like Microsoft are constantly putting pressure on each other to make sure both sides are continuing to make the best use of each others' technology.

"AMD needs to innovate, as a gaming company and as a CPU company," Huddy said. "Microsoft needs to do the same thing. We’re making sure that the synergy between us is a highly cooperative one. If they were to say ‘graphics is a done deal’ that would be a big problem. They haven’t said that.”

Q.E.D. (???)

 

[Obsoleet]

Yeah, people just dont realize that instead of calling an API (ie DX11 according to the green appologists here) to work "just fine" because you "only" need a $300 CPU to barely stay at 60 fps on the majority of current gen games and blaming the rest of the lower performing CPUs for being "weak", they should start blaming the API for being weak and the one to be pointed at.

People probably need to play more MMOs to know the sad state of DX11 in CPU overhead... its just plain silly to be at 40% GPU load all the time with all the eyecandy enabled and MSAA x4 and not being able to reach the 60fps target. Thats when you realize that most people here really havent gamed in a long time, the "throw more hardware to the problem" became obsolete and people shouldnt be forced to shell more on their hardware to compensate such an awfully optimized API as DX is in its current state.

Put simply, this is just AMD pushing the hardware industry forward again.
I know praising AMD or their stuff doesn't go down well around here, but they're the Google of the hardware industry. They did it with the Athlon, AMD64, Eyefinity, HSA/APUs and now Mantle. Mantle is a big deal, if it weren't you wouldn't see everyone scrambling the jets. It's a 'ignore at your own peril' thing.

But while everyone else is talking about it, AMD has always done it. I can't think of anything Nvidia has done to push the industry forward at all. Physx? 3D Vision?? Nvidia's main claim to fame is their marketing, it's really, really good and that's not a slight. AMD drops the ball in many areas and it's frustrating to watch. But, I don't consider NV an engineering company at its heart but rather a marketing company. It obviously works out. It does for Apple too.

The 'throw hardware at the problem' days have been over for a while. Things change.
I'm a software developer by trade, and it's going to be hard to compete with SteamOS for efficiency as time goes on. I don't think many people are going to stick with being forced to buy new versions of Windows for DX API updates every few years once SteamOS is up and running. It doesn't make sense. I'm not doing it. I'm on my last version of Windows (7). This or XP will get run in a VM on SteamOS for old Windows95-XP games that I want to play. DOSBox for everything else. New games linux/SteamOS only.

If my rig blew up tonight, I'd replace it with a cheap Kaveri rig. Because games have plateaued for a while now, and PS4/XB1 aren't really going to push things forward either.
I'm in a holding pattern until the consumer Oculus Rift is ready. Or until one finds a good 30" 60hz 4K LCD. Once OR is released, if it's 1440x2 a lot of people are going to be upgrading. Eyefinity provides an incentive if people like it, I never did. Right now though, according to Steam the most popular is just old fashioned 1920x1080. When I play games on my plasma sometimes I use 720 because it's easier to read text.
So most gamers are at pretty low res. I've always been convinced this forum and most of Anandtech besides those in the PC Gaming subforum do not play many games at all. This is a place where people gather who are excited about specs, and enjoy spending money on hardware for the thrill. I was there once, for a long time. There are some gamers here of course, I'm one, but I'm convinced many are simply interested in the hardware. Which is getting stranger over time because most hardware is grossly overpowered, the obvious benefits are gone.
I was more into the hardware update cycle when we actually needed it, for a long time we HAD to have upgrades or games wouldn't even run. I don't have any games that won't run on my Q9450 (6 years old) /5870 (5 years old).
But I can tell you many times that I had to buy very expensive RAM, or a new CPU to play games in the early 90's. Change happens.

Like I said I'm a developer, I can afford whatever computer I want. I priced out a new system on Newegg tonight without monitor and it came to $2200. This is an allout Intel system. Looking at the games I play, I think it's a waste. Kaveri or Carrizo makes more sense to me, but as APUs catch up and HSA gets more software adoption, I'd rather wait it out if I can.
We'll see when Oculus Rift arrives, then maybe $2200USD on a rig to replace my currently working-well system will be worth it.

 

[R0H1T]

Talk about self aggrandizement. If DX12 is just now in alpha stage, why on Earth would Microsoft tell you, a mere beta tester, anything about it?

No one said that they'd tell me about it, only the hardware vendors would've had early access & not even game developers, but you keep on peddling this statement "that it was in development as early as 4yrs back" as a fact is frankly a lie because some bits of it would've made there way through to win8 at some point in time ! The development on DX12 probably started near the time XBONE was initially conceived alongside AMD or perhaps later on when the APU was fully functional, that's the best guesstimate for a timeline during which DX12 would've made it even to the drawing board, anytime before that certainly doesn't make any sense to me.

And talking about an over inflated ego, it does seem to me that you have plenty of it !

 

[3DVagabond]

People have taken this statement from nVidia...

Our work with Microsoft on DirectX 12 began more than four years ago with discussions about reducing resource overhead. For the past year, NVIDIA has been working closely with the DirectX team to deliver a working design and implementation of DX12 at GDC.

...and twisted and spun it into saying, "We've been actively involved with the development of DX12 for four years."



As R0H1T said the most likely scenario is that work between MS and AMD for the close to metal API used by XB1 was the foundation. I believe it evolved into DX12 with additional input from nVidia (and likely Intel) which started last year. This would have been necessary for multi vendor support.

 

[Alatar]

I can't think of anything Nvidia has done to push the industry forward at all.

-(modern) GPGPU
-Variable refresh monitors for gaming
-(modern) multi GPU implementations
-Frame pacing (making multi GPU actually useful)

Just to name a few. You just have to think harder.

It's actually much easier to think of Nvidia innovations when it comes to GPUs.