AMD CPU speculation... and expert conjecture

[kebbz]

mantle will destroy nvidia

 

[The Q6660 Inside]

mantle will destroy nvidia

Not per say, we cannot judge right now...

 

[designasaurus]

Regarding Mantle: I have been thinking, there are only 2 gpu companies these days for gaming (I can't believe any gamer really counts Intel). If game publishers wish to expand beyond Windows, they'll have to program for OpenGL, or OpenGL+DirectX. However, everything I hear about OpenGL is how unreliable and bass-akwards the organization behind it is, which makes me doubt its possibilities for further adoption. So, with only two gaming gpu setups to program for, why wouldn't companies choose to use Nvidia's API and Mantle rather than DirectX and OpenGL?

 

[juanrga]

So much anger over Mantle. If AMD is going to pay DICE and whatever to add Mantle support, I don't see how it's DoA when PhysX has been around forever.

The writing is on the wall. Game developers want out of Windows, and Mantle is a tool to help make that possible. A huge problem with gaming in Linux is DirectX support. WINE does DX9 and that's it. There is no good way of making it work. Valve has a wrapper to translate API calls from DX to OpenGL but it's just needless overhead, specially compared to Mantle.

GamerK, I do feel you're missing the point. Linux has a less than 1% market share, yet Valve is pushing Steam on Linux so hard that they've gone as far as to make their own distribution of Linux.

EA and DICE are already talking about how it just would take ONE big game to get Linux as a viable gaming OS. And EA has had some foresight and used QT for Origin, which is a very easy to port platform. Why did they not use .net since Visual Studio is so good? They used QT so they could port it to different OSes. Even Android has some sort of QT support. It's (obviously) there on Windows, and it is good on BSD, Linux, OSX, iOS, Windows CE, and there are even external ports for more obscure OSes like Haiku.

So, if MS and Windows environment is so great, why did EA go out of its way to use QT?

Do you see now? A lot of developers want out of Microsoft. They are scared of what is happening with Windows 8 and how that is going to affect PC gamers. They see how Intel and Nvidia are pushing for a closed ecosystem of disposable devices which can't be upgraded (hurray for profitability through restrictions!) and they want to keep the gaming PC platform alive. It is growing at a staggering rate, yet MS and Intel want it DEAD because it's not making them as much money as you could make selling fruity toys (note how Apple is worth more than Intel right now, mainly due to a phone while Intel's technological achievements are far greater).

Mantle makes it a lot easier to get your games out of the Windows ecosystem. Companies like Valve don't care about market share, they want out of Windows badly and they'll take any tools they can to get away from where Windows is heading.

It also works out quite well for Intel as well. Their CPUs have been perfectly acceptable in high end gaming rigs for the last 5 years (meaning an i7 920 with a little overclock could still power a good gaming system just fine without massive bottlenecking). For every CPU sale Intel sees in a gaming PC, they more than likely see at least two graphic card sales go to Nvidia and AMD.

You're here ranting about how market share is the defacto goal of game developers right now, and you're doing so because you're (incorrectly assuming like always) that because game developers have always went for the bigger market shares first, that they always will.

If that is still the case, and we're not at the foothills of a mountain of change, then why is both Nvidia and AMD investing so much in Steam Machines to run Steam OS? Why are we seeing AMD go as far as to create a new API that's compatible with DX in some ways, yet is cross platform (ergo Linux and OSX if the consoles are removed from the equation).

You are completely missing what is going to happen. AMD gaming PCs is a huge market for them. They have very little in professional workstations. They don't have much in mobile outside of Jaguar. Their bread and butter is PC gaming. Even their APUs are sold under the guise of affordable gaming platforms.

You are really, really missing the point of Mantle and what Nvidia and AMD are BOTH trying to do. They want out of the Windows future. Where Microsoft is directly competing with hardware that don't make a good platform for what both Nvidia and AMD depend on for large amounts of money (dGPUs and APUs).

So please, really, stop with the "THIRD PLACE LOL!" and market share remarks. We're seeing companies make moves to get away from Windows, and to do so means that they're going to have to take the risk of pushing markets with little market share.

I don't really care where the marketshare of GCN ends up. In gaming PCs, it's definitely got much more market share than Linux does, and Valve is pushing Linux really, really hard. Mantle is a tool to help with this, and AMD went as far as to keep it open so Nvidia can jump on it and implement it themselves if they want.

We already have gone from you implying this is some sort of massive API where games have to be completely coded from scratch to use it, and that no one will use it, to it being much more closer to a recompiler.

I really don't think you understand how much EA and Valve want out of the Windows ecosystem. They won't go OpenGL because it sucks for the most part, but they will definitely go Mantle if they can just recompile with different libraries and get a cross-platform package. It's even easier than an OpenGL port that way and it's going to be massively faster.

It's quite fun watching all the Wintel Nvidiots cry over Mantle and what's happening. I am quite sure you were one of those folks saying Steam on Linux was a disaster and DoA as well. I'm going to enjoy coming back to this post in a year or so and reading how someone like you can be wrong so many times.

AMD has made it quite clear that they are talking with game developers about what they want, and they're asking for tools to get away from Windows. Another massive assumption that you're making is that AMD is forcing this on game developers and studios. For all you know, they could have approached AMD themselves and said they wanted help getting away from DX and Windows.

We simply don't know and you're entire argument rests upon this assumption of which you have no proof of.

So, I hope you can kind of see what game studios want from this. They want cross platform. Indie developers want cross-platform (notice how fast they were to jump on Steam on Linux?) No one likes what MS wants and no one wants to be bound to an ecosystem where your host for that ecosystem is trying to build a vertical stack that runs from hardware to software in house. Specially if you're a hardware manufacturer. Specially if you're a game developer releasing on a platform that is so ill-received by gamers as Windows 8 with metro and knowing that you're basically at the mercy of MS's choices and you have no other options.

So really, just stop already. You have a huge problem with assuming that things that happened in the past will continue to happen no matter what. You do it with multi-threading, you do it with APIs, you do it with EVERYTHING. I realize it works sometimes, but if that's all you do you're limiting your foresight massively.

Excellent remarks. I already explained before that Windows is not the future.

I already mentioned that AMD announced, early this year, that they are abandoning Windows exclusivity and will develop for other OS.

I already mentioned how Intel during its annual conference, a pair of months ago, praised Google and criticized Windows.

I already gave the link with HP claiming a week ago that Windows is not the future and that they are moving away.

We know that Valve and other game developers are abandoning windows.

I find amazing that some people still believes that Microsoft and Windows are driving the market or the industry. I expect Windows 7 will remain for users slow on migrating to anything new. Just as XP is still used still by lots of Windows-oriented people.

Rest of world will move to Linux. About market share, please don't spread the 1% myth. Share is about 8%--12%, which put linux at same level than Windows 8. This explain why Valve and others are ready to switch to linux bandwagon.

 

[juanrga]

Nice post blackstar, but gamerk and palladin do have an important point: from the inside of the development cycle and a business point of view, it doesn't make sense to have yet another API that you have to prepare developers for. It does cost time and money to use a new API in terms of development, testing (QA) and time-to-market.

But the key here is that MANTLE can be considered like a _port_ of the XB1 low-level API to PC. It is not an entirely new API that developer have to learn from zero. As mentioned in ExtremeTech article the differences between both APIs are minimal. Those differences account for the difference between a Windows PC and XB1.

Contrary to Gamer/Palladin idea MANTLE simplifies porting from console to PC. This is what AMD has said in response to all this confusion:

Mantle is NOT in consoles. What Mantle creates for the PC is a development environment that's *similar* to the consoles, which already offer low-level APIs, close-to-metal programming, easier development and more (vs. the complicated PC environment). By creating a more console-like developer environment, Mantle: improves time to market; reduces development costs; and allows for considerably more efficient rendering, improving performance for gamers. The console connection is made because next-gen uses Radeon, so much of the programming they're doing for the consoles are already well-suited to a modern Radeon architecture on the desktop; that continuum is what allows Mantle to exist. ^RH

It is about reusing code and retaining performance when porting to PC.

 

[juanrga]

Additionally, the list of titles being done on Frostbite alone is mind bending.

Dragon Age series' next installment
BF4
Need for Speed series' next installment
Command & Conquer series' next installment
Mirror's Edge 2
SW Battlefront 3
Mass Effect series' next installment (ME4?)

EA uses this engine *A LOT*, for lots of AAA titles, which may very well drive a big partnership with MANTLE and push it heavily in the console to PC porting side of their publishing house. AMD was not selecting poorly when they chose the *LARGEST* publisher in all of video games as a partner to implement this with. EA is always looking for an advantage as well, so this could benefit both sides greatly.

EDIT: If they got just 1 more major publisher to push this...like Activision Blizzard or Bethesda or Valve, for example...they would likely ensure future adoption at a high level.

Any EA publishing partner would have access to this, and likely be persuaded to use it...

About 20 games with MANTLE support are being developed and will be announced soon.

Activision has already announced that will support MANTLE. Crytek is now under the radar. Unreal is Nvidia oriented, but who knows?

 

[juanrga]

AMD is replacing the jaguar-based servers by ARM-based servers. This is happening at 2014.

AMD has claimed how all the future servers will be ARM-based. Feldman has explained why customers want ARM to win.

AMD has also claimed that they are releasing the new x86 based Warsaw servers only for institutional customers that will be slow on migrating to ARM. To be read as, once all our customer migrate to ARM we will not release x86 servers anymore.

AMD will be also replacing jaguar-based tablets by ARM-based-tablets this or next year.

Show me one piece of material from AMD directly that is worded "ARM will replace X"...I can't find one...sounds like more tinfoil hats and speculation. This thread has become full of that lately.

Rocks, you continue using _double standards_. For instance, you didn't provide us any piece of material from AMD saying "We will release FX Steamroller for AM3+ platfform", but you have been saying to everyone on the forums how one could purchase a FX Piledriver plus an AM3+ mobo now and upgrade to Steamroller FX 8-cores latter using the same mobo.

I have not provided you a quote with the exact selection of words that you want to hear, but I have provided you (i) official roadmaps and server plans showing how Seattle replaces Opteron-X, (ii) the "ARM will win" secret slide from AMD, and (iii) quotes such as

He [Feldman] also told us that “Warsaw”, AMD’s follow up in the 2P/4P server market, will cater for a declining but still large market with institutional customers that will take longer to migrate to an ARM ecosystem.

What part of migrating to ARM you don't still understand?

What part of taking longer don't you understand. As I stated before, what is this "longer"? 2, 4, 7, 10 years ... in the mean time in order for this to happen, x86 needs to stop advancing.

your acting like this is taking place in 2014&2015. its not. In your little world, amd is going to stop advancing x86 and just wait for ARM to catch up. That only makes sense to you, no one else, not even AMD is going to just stand still and wait for something that may take decades to happen.

While (Juanrga = wrong) {printf("JUANRGA IS NEVER WRONG")};

No. You are completely wrong once again.

ARM will be not replacing x86 in 2014/2015. In the first place it will took longer, in the second place I already explained you that x86 will remain for legacy workloads as a niche market. I gave the analogy with windows XP as well...

What is being replaced in 2014 are the jaguar servers by the new ARM servers.

Also nobody said you that x86 will stop advancing. What happens is that ARM is advancing faster than x86. As I already said before, but you don't read:

ARM is the fastest-growing CPU architecture in history

A diagram comparing performance growing for both x86 and ARM was given, but you missed this as well. ARM achieves in one generation what x86 achieves in three or four approx.

Now please ignore what is being said to you, make silly things in your mind again and post another laughable reply. You have been doing this for weeks, therefore it would be effortless for you

 

[juanrga]

That was made by a user on here called juangra and is about 60-70% off reality.

I think you mean "about 6000-7000% off reality", because you are living in your own distorted reality where Intel chips are 2000% faster than AMD and 35000% more efficient. Intel is cheaper than AMD also in your own reality.

 

[juanrga]

superpi, gimp image editing, office macro benches. Its so funny how people make 4 core kaveri look like they compete against i5's, they will struggle to beat the i3 haswell cpu.

Super PI, x87, trololollol. You realize AMD is ~40% behind in single-threaded performance, Steamroller should get a 25-30% increase, AKA near Sandy levels, so along with the slighly higher clocks, AMD can get some serious heat on Intel. Your 110% is B$ pulled out of you arse, literally everybody here on this thread has went over and over on how that is untrue, even at Nehalem per core levels, AMD will do seriously good .

You say this as you know I am right.

No. He says because he knows you are plain wrong.

You have been corrected in innumerable occasions, but you return with the same nonsense again and again.

 

[gamerk316]

That was made by a user on here called juangra and is about 5% off reality.

Fixed.
Steamroller will be better than you think, AMD is only 30-40% behind Intel per core.

That's only, what, 18 FPS less in a game getting 45 FPS on average, if the cores are clocked at the same speed?

Not really making the best argument here...

 

[gamerk316]

Regarding Mantle: I have been thinking, there are only 2 gpu companies these days for gaming (I can't believe any gamer really counts Intel). If game publishers wish to expand beyond Windows, they'll have to program for OpenGL, or OpenGL+DirectX. However, everything I hear about OpenGL is how unreliable and bass-akwards the organization behind it is, which makes me doubt its possibilities for further adoption. So, with only two gaming gpu setups to program for, why wouldn't companies choose to use Nvidia's API and Mantle rather than DirectX and OpenGL?

Two reasons:

First off: Remember the days before DX9.0a and SM2.0, back when AMD and NVIDIA had separate pixel models. DX Incorporated several modes prior to DX9, but for the most part, PM 1.1 = NVIDIA and PM 1.4 = AMD. Depending on which one you programmed with, you would bias game performance one way or the other. When SM 2.0 came out, a lot of the performance bias in benchmarks started to go away [some still exists due to architecture differences, but for teh most part, cards line up in order of expected performance].

Using low level native API's will bias benchmarks, significantly. And since benchmarks are a large factor when purchasing a card, whichever company pushes their API the best will basically win a near monopoly on the GPU market. And given the differences in cash NVIDIA can basically do this with ease.

In short: If it becomes a war of low-level API's, AMD looses since NVIDIA can just buy the market.

Secondly, on PC's, there isn't a need for Mantel. The worse case overhead for DX11 is "maybe" 5% performance loss, which for a game running 60 FPS equates to ~3 FPS worth of performance loss, which is acceptable. There simply isn't much overhead left in the API.

So is re-writing the software and adding a secondary graphical backend, that only benefits the minority player in the GPU market, worth maybe a 5 FPS gain for developers? No, it isn't.

 

[8350rocks]

Regarding Mantle: I have been thinking, there are only 2 gpu companies these days for gaming (I can't believe any gamer really counts Intel). If game publishers wish to expand beyond Windows, they'll have to program for OpenGL, or OpenGL+DirectX. However, everything I hear about OpenGL is how unreliable and bass-akwards the organization behind it is, which makes me doubt its possibilities for further adoption. So, with only two gaming gpu setups to program for, why wouldn't companies choose to use Nvidia's API and Mantle rather than DirectX and OpenGL?

Two reasons:

First off: Remember the days before DX9.0a and SM2.0, back when AMD and NVIDIA had separate pixel models. DX Incorporated several modes prior to DX9, but for the most part, PM 1.1 = NVIDIA and PM 1.4 = AMD. Depending on which one you programmed with, you would bias game performance one way or the other. When SM 2.0 came out, a lot of the performance bias in benchmarks started to go away [some still exists due to architecture differences, but for teh most part, cards line up in order of expected performance].

Using low level native API's will bias benchmarks, significantly. And since benchmarks are a large factor when purchasing a card, whichever company pushes their API the best will basically win a near monopoly on the GPU market. And given the differences in cash NVIDIA can basically do this with ease.

In short: If it becomes a war of low-level API's, AMD looses since NVIDIA can just buy the market.

Secondly, on PC's, there isn't a need for Mantel. The worse case overhead for DX11 is "maybe" 5% performance loss, which for a game running 60 FPS equates to ~3 FPS worth of performance loss, which is acceptable. There simply isn't much overhead left in the API.

So is re-writing the software and adding a secondary graphical backend, that only benefits the minority player in the GPU market, worth maybe a 5 FPS gain for developers? No, it isn't.

That's all well and good, but I have a question here:

So, why is the number I am hearing tossed around inside the industry 30-40%+ performance gain? I got wind of some stuff last night, one of the guys I work with was talking about MANTLE. As we are investigating it on several different levels, he has heard that initial tests in BF4 were looking at a 30-40%+ performance gain for MANTLE over DX on AMD cards in Windows. He mentioned they were experimenting with other OS as well; however, he had no information about relative performance for them.

Last point:

NVidia's low level API is dead in the water because they have no foothold in the consoles. No one will develop on that API just like no one really uses PhysX much anymore. However, MANTLE is an open standard, which means that NVidia can use it as well. This would likely be the most intelligent thing that NVidia could ever do, as their low level API uses proprietary code, and MANTLE uses HLSL. It's literally 2x the work to use the NVidia API (and their market share is shrinking over the last 12 months), with MANTLE, it's a few minor tweaks and a recompile.

You and I both know developers are LAZY! They won't do 2x the work to use NVidia's low level API regardless of how much money is tossed around. However, if they have to change a few lines of code and recompile (an hour worth of work, perhaps, plus compile time), then you might actually convert some of them to your cause.

The issue with NVidia is, everything is proprietary...nothing is cross compatible. *THAT* is why MANTLE will succeed where they failed. NVidia's low level API will suffer the same fate as GLIDE.

 

[8350rocks]

http://www.amazon.co.uk/review/R6NLTFRGDJECE/ref=cm_cr_pr_cmt?ie=UTF8&ASIN=B00CV3E3WK&linkCode=&nodeID=&tag=#wasThisHelpful
read the review, an amd user going to intel and suprise suprise he gets double the performance over a phenom 2 and we all know phenom 2 ipc is slightly better then piledriver so with my 2.1xipc theory the i5 4570 with turbo is on par with an fx8350.

i5 4670k:
http://www.amazon.co.uk/review/R29REJDY4R8AUR/ref=cm_cr_dp_cmt?ie=UTF8&ASIN=B00CO8TBOW&channel=detail-glance&nodeID=340831031&store=computers#wasThisHelpful
http://www.amazon.co.uk/review/R3KDIPLP26G9QD/ref=cm_cr_pr_cmt?ie=UTF8&ASIN=B00CV3E3UC&linkCode=&nodeID=&tag=#wasThisHelpful

Now lets see a 4 core kaveri get to 80% of an i5 haswell cpu at same clock or 70% of an i5 ivy bridge cpu. Even that will be on amds at best target for steamroller.

This forum really needs a F&%*($& ignore button for garbage like this...

Can we get one of those? Please...?

 

[Ags1]

Why all the hate for OpenGL? I admit I am no game developer, but I do write 3D games as a hobby. Like any games developer would, I use a game engine that basically hides all the technical muck of OpenGL from me. I work with a scene graph and write GLSL shaders - and I hear very few complaints about GLSL itself.

The technical details of the 3D API are only of concern to those implementing a game engine.

Are all these OpenGL haters just people used to D3D and D3D tools? I think so.

 

[gamerk316]

Regarding Mantle: I have been thinking, there are only 2 gpu companies these days for gaming (I can't believe any gamer really counts Intel). If game publishers wish to expand beyond Windows, they'll have to program for OpenGL, or OpenGL+DirectX. However, everything I hear about OpenGL is how unreliable and bass-akwards the organization behind it is, which makes me doubt its possibilities for further adoption. So, with only two gaming gpu setups to program for, why wouldn't companies choose to use Nvidia's API and Mantle rather than DirectX and OpenGL?

Two reasons:

First off: Remember the days before DX9.0a and SM2.0, back when AMD and NVIDIA had separate pixel models. DX Incorporated several modes prior to DX9, but for the most part, PM 1.1 = NVIDIA and PM 1.4 = AMD. Depending on which one you programmed with, you would bias game performance one way or the other. When SM 2.0 came out, a lot of the performance bias in benchmarks started to go away [some still exists due to architecture differences, but for teh most part, cards line up in order of expected performance].

Using low level native API's will bias benchmarks, significantly. And since benchmarks are a large factor when purchasing a card, whichever company pushes their API the best will basically win a near monopoly on the GPU market. And given the differences in cash NVIDIA can basically do this with ease.

In short: If it becomes a war of low-level API's, AMD looses since NVIDIA can just buy the market.

Secondly, on PC's, there isn't a need for Mantel. The worse case overhead for DX11 is "maybe" 5% performance loss, which for a game running 60 FPS equates to ~3 FPS worth of performance loss, which is acceptable. There simply isn't much overhead left in the API.

So is re-writing the software and adding a secondary graphical backend, that only benefits the minority player in the GPU market, worth maybe a 5 FPS gain for developers? No, it isn't.

That's all well and good, but I have a question here:

So, why is the number I am hearing tossed around inside the industry 30-40%+ performance gain? I got wind of some stuff last night, one of the guys I work with was talking about MANTLE. As we are investigating it on several different levels, he has heard that initial tests in BF4 were looking at a 30-40%+ performance gain for MANTLE over DX on AMD cards in Windows. He mentioned they were experimenting with other OS as well; however, he had no information about relative performance for them.

Last point:

NVidia's low level API is dead in the water because they have no foothold in the consoles. No one will develop on that API just like no one really uses PhysX much anymore. However, MANTLE is an open standard, which means that NVidia can use it as well. This would likely be the most intelligent thing that NVidia could ever do, as their low level API uses proprietary code, and MANTLE uses HLSL. It's literally 2x the work to use the NVidia API (and their market share is shrinking over the last 12 months), with MANTLE, it's a few minor tweaks and a recompile.

You and I both know developers are LAZY! They won't do 2x the work to use NVidia's low level API regardless of how much money is tossed around. However, if they have to change a few lines of code and recompile (an hour worth of work, perhaps, plus compile time), then you might actually convert some of them to your cause.

The issue with NVidia is, everything is proprietary...nothing is cross compatible. *THAT* is why MANTLE will succeed where they failed. NVidia's low level API will suffer the same fate as GLIDE.

Mantel is NOT an open standard; it REQUIRES a GCN based GPU. Which means mantle dies when AMD eventually replaces GCN.

Secondly: Ignore the marketers. Remember what AMD claimed about BD performance?

Thirdly: Let the API wars commence: http://www.guru3d.com/news_story/nvidia_does_5_million_deal_with_ubisoft.html

NVIDIA now throwing money around. Lets see who buys Activation (probably NVIDIA).

 

[8350rocks]

Regarding Mantle: I have been thinking, there are only 2 gpu companies these days for gaming (I can't believe any gamer really counts Intel). If game publishers wish to expand beyond Windows, they'll have to program for OpenGL, or OpenGL+DirectX. However, everything I hear about OpenGL is how unreliable and bass-akwards the organization behind it is, which makes me doubt its possibilities for further adoption. So, with only two gaming gpu setups to program for, why wouldn't companies choose to use Nvidia's API and Mantle rather than DirectX and OpenGL?

Two reasons:

First off: Remember the days before DX9.0a and SM2.0, back when AMD and NVIDIA had separate pixel models. DX Incorporated several modes prior to DX9, but for the most part, PM 1.1 = NVIDIA and PM 1.4 = AMD. Depending on which one you programmed with, you would bias game performance one way or the other. When SM 2.0 came out, a lot of the performance bias in benchmarks started to go away [some still exists due to architecture differences, but for teh most part, cards line up in order of expected performance].

Using low level native API's will bias benchmarks, significantly. And since benchmarks are a large factor when purchasing a card, whichever company pushes their API the best will basically win a near monopoly on the GPU market. And given the differences in cash NVIDIA can basically do this with ease.

In short: If it becomes a war of low-level API's, AMD looses since NVIDIA can just buy the market.

Secondly, on PC's, there isn't a need for Mantel. The worse case overhead for DX11 is "maybe" 5% performance loss, which for a game running 60 FPS equates to ~3 FPS worth of performance loss, which is acceptable. There simply isn't much overhead left in the API.

So is re-writing the software and adding a secondary graphical backend, that only benefits the minority player in the GPU market, worth maybe a 5 FPS gain for developers? No, it isn't.

That's all well and good, but I have a question here:

So, why is the number I am hearing tossed around inside the industry 30-40%+ performance gain? I got wind of some stuff last night, one of the guys I work with was talking about MANTLE. As we are investigating it on several different levels, he has heard that initial tests in BF4 were looking at a 30-40%+ performance gain for MANTLE over DX on AMD cards in Windows. He mentioned they were experimenting with other OS as well; however, he had no information about relative performance for them.

Last point:

NVidia's low level API is dead in the water because they have no foothold in the consoles. No one will develop on that API just like no one really uses PhysX much anymore. However, MANTLE is an open standard, which means that NVidia can use it as well. This would likely be the most intelligent thing that NVidia could ever do, as their low level API uses proprietary code, and MANTLE uses HLSL. It's literally 2x the work to use the NVidia API (and their market share is shrinking over the last 12 months), with MANTLE, it's a few minor tweaks and a recompile.

You and I both know developers are LAZY! They won't do 2x the work to use NVidia's low level API regardless of how much money is tossed around. However, if they have to change a few lines of code and recompile (an hour worth of work, perhaps, plus compile time), then you might actually convert some of them to your cause.

The issue with NVidia is, everything is proprietary...nothing is cross compatible. *THAT* is why MANTLE will succeed where they failed. NVidia's low level API will suffer the same fate as GLIDE.

Mantel is NOT an open standard; it REQUIRES a GCN based GPU. Which means mantle dies when AMD eventually replaces GCN.

Secondly: Ignore the marketers. Remember what AMD claimed about BD performance?

Thirdly: Let the API wars commence: http://www.guru3d.com/news_story/nvidia_does_5_million_deal_with_ubisoft.html

NVIDIA now throwing money around. Lets see who buys Activation (probably NVIDIA).

1.) MANTLE was advertised as an open standard API in Hawaii, as it currently sits, it is written for GCN. Though AMD would not do NVidia's work for them now would they (nor could they for that matter)...

2.) That was word from DICE, not AMD. AMD have been tight lipped about word for the most part.

3.) Again, wait and see...NVidia is going to fight, I don't doubt it. Though who has more money AMD + EA + Activision/Blizzard, or NVidia + Ubisoft?

I would bet EA has more cash than all other parties combined, personally...just saying...

Additionally, I would like to point out, that AMD chose a deal with the largest publisher in the world over Ubisoft, and NVidia is picking up the "leftovers" after AMD walked away from the deal.

EDIT: Just FYI, I thought this interesting...

When I was contacting different companies for quotes on server infrastructure for the upcoming MMO to gauge cost/performance...I found out from SeaMicro that Activision/Blizzard use SeaMicro servers with Opterons for all their game servers for D3 and WoW. They mentioned that they had a contract for "future projects" with them as well.

 

[blackkstar]

Regarding Mantle: I have been thinking, there are only 2 gpu companies these days for gaming (I can't believe any gamer really counts Intel). If game publishers wish to expand beyond Windows, they'll have to program for OpenGL, or OpenGL+DirectX. However, everything I hear about OpenGL is how unreliable and bass-akwards the organization behind it is, which makes me doubt its possibilities for further adoption. So, with only two gaming gpu setups to program for, why wouldn't companies choose to use Nvidia's API and Mantle rather than DirectX and OpenGL?

Two reasons:

First off: Remember the days before DX9.0a and SM2.0, back when AMD and NVIDIA had separate pixel models. DX Incorporated several modes prior to DX9, but for the most part, PM 1.1 = NVIDIA and PM 1.4 = AMD. Depending on which one you programmed with, you would bias game performance one way or the other. When SM 2.0 came out, a lot of the performance bias in benchmarks started to go away [some still exists due to architecture differences, but for teh most part, cards line up in order of expected performance].

Using low level native API's will bias benchmarks, significantly. And since benchmarks are a large factor when purchasing a card, whichever company pushes their API the best will basically win a near monopoly on the GPU market. And given the differences in cash NVIDIA can basically do this with ease.

In short: If it becomes a war of low-level API's, AMD looses since NVIDIA can just buy the market.

Secondly, on PC's, there isn't a need for Mantel. The worse case overhead for DX11 is "maybe" 5% performance loss, which for a game running 60 FPS equates to ~3 FPS worth of performance loss, which is acceptable. There simply isn't much overhead left in the API.

So is re-writing the software and adding a secondary graphical backend, that only benefits the minority player in the GPU market, worth maybe a 5 FPS gain for developers? No, it isn't.

That's all well and good, but I have a question here:

So, why is the number I am hearing tossed around inside the industry 30-40%+ performance gain? I got wind of some stuff last night, one of the guys I work with was talking about MANTLE. As we are investigating it on several different levels, he has heard that initial tests in BF4 were looking at a 30-40%+ performance gain for MANTLE over DX on AMD cards in Windows. He mentioned they were experimenting with other OS as well; however, he had no information about relative performance for them.

Last point:

NVidia's low level API is dead in the water because they have no foothold in the consoles. No one will develop on that API just like no one really uses PhysX much anymore. However, MANTLE is an open standard, which means that NVidia can use it as well. This would likely be the most intelligent thing that NVidia could ever do, as their low level API uses proprietary code, and MANTLE uses HLSL. It's literally 2x the work to use the NVidia API (and their market share is shrinking over the last 12 months), with MANTLE, it's a few minor tweaks and a recompile.

You and I both know developers are LAZY! They won't do 2x the work to use NVidia's low level API regardless of how much money is tossed around. However, if they have to change a few lines of code and recompile (an hour worth of work, perhaps, plus compile time), then you might actually convert some of them to your cause.

The issue with NVidia is, everything is proprietary...nothing is cross compatible. *THAT* is why MANTLE will succeed where they failed. NVidia's low level API will suffer the same fate as GLIDE.

Mantel is NOT an open standard; it REQUIRES a GCN based GPU. Which means mantle dies when AMD eventually replaces GCN.

Secondly: Ignore the marketers. Remember what AMD claimed about BD performance?

Thirdly: Let the API wars commence: http://www.guru3d.com/news_story/nvidia_does_5_million_deal_with_ubisoft.html

NVIDIA now throwing money around. Lets see who buys Activation (probably NVIDIA).

You still do not understand what an API is and what an implementation is.

Mantle is an open API. You are free to make your own implementations of it to run on whatever hardware you like. If I wanted to, I could write a Mantle implementation that worked on Ardeno GPUs. It doesn't matter.

You seem to think that because the only existing implementation is for GCN, that that's the only implementation that will exist. AMD could write a compatible Mantle implementation that worked on their old VLIW cards. Nvidia can make one for Kepler or Maxwell. Intel could even make their own.

This is the part that you don't understand.

DirectX is closed, it is not open. Look at how long it has taken WINE project to reverse engineer DX and get a working implementation of it. It has taken a VERY VERY long time and they still don't have DX support beyond DX9.

You do not understand much out-side of gaming related projects.

Mantle is going to make it extremely easy to make Linux and OSX ports of your games as the API is under the control of the hardware manufacturer instead of the OS vendor. Meaning that it is AMD (and whoever else decides to implement it)'s decision on what OS it will work on.

This is the point you are missing. I wish you could crawl out from underneath your "single core is king Wintel" rock and understand what Valve, EA, and tons of indie developers are trying to do, which is get as far away from Windows dependence as possible.

Understand this: Mantle being an easy port to OSX or Linux and only limiting it to GCN cards is much better than DirectX being a difficult port to OpenGL that anyone can use. Specially regarding Steam Machines that run GCN GPUs.

 

[de5_Roy]

since this post has a on-tpoic reference (technicality! 😀)... you are gonna miss it because you just skipped over to this reply. wiser people are actually reading the reply after mine ... i assume..

Okay amd users explain where improving lets say an a10-6800k or 5800k by 30% will get to an ivy bridge i5 at lets say 3.2ghz performance wise with a steamroller 4 core kaveri.

not an amd user, and your comparison is way off. a10 apus are for entry level pcs, core i5 is for upper mirange to high or lower high end pcs.
kaveri isn't out yet, so comparison is meaningless.

I know its will be 70% slower going by amds projected gains but prove this is wrong. You simply can't as thats why you ignore the benchmarks and make made up figures to make amd look closer then reality.

how do you know if the product is not even out yet? are you yet another addition to the group of people who claim having inside contacts?

No one would buy a core i5 if the fx8320 was only 30% slower per core then an i5 ib. That would make the fx8320 beat the i5 ib cpu lets say both at 3.5ghz easily. An i5 at same clock speed would equal an fx8320 with 5 cores 100% loaded and 20% of the 6 core loaded and 0% load of the other 2 cores. That would make the fx8320 sell like hotcakes if this was the case but we all know the i5 equals the 8 cores if at same clock.

no.
if fx8320's cores are in use, it'll beat the 4 core i5 as long as the workload is using 6-8 cores. by how much - will depend on the type of task. turbo, imc performance, heat dissipation will factor in as well. your percentage calculation is off.

So if you theory is correct an i5 at 3.5ghz would be equal to 5.2 cores of a piledriver cpu at 3.5ghz. Thats means its 60-70% slower then an fx8320.

Now the reality is ivy bridge is 2x better per core then piledriver likewise haswell is 2.1x better per core then piledriver which makes the i5 ivy bridge at 3.5ghz 4 cores equal 8 cores of amd piledriver cores at 3.5ghz and richland is not any better really.

ivy bridge isn't 2x better (context not given, thus a blanket term) than piledriver.
you keep disregarding the cpu prices. that makes your already wrong calculations..er... more wrong...

 

[noob2222]

No wonder why amd fanboys don't get ivy bridge i5's are 2x better per core. Did you not see an fx6300 lose to a 1st gen 2.8ghz i5 at battlefield 4. Now ivy bridge is gigantically better then that i5. IB i5 4 core 3.5ghz= PD 8 core at 3.5ghz multithreaded performance. Now intel has half the cores so in order to be as fast it needs to be 2x better per core and it is to match it multithreaded wise. My i5 4430 haswell is as fast as an fx8320 multithreaded wise do you know why 2.1x better ipc.

Pull your head out of the blue koolaid.
Watch the if i5 4670 @4.4 ghz lose to the 8350 @ 4.5 ghz.

http://forums.pureoverclock.com/video-card-overclocking/22523-post-your-cinebench-r15-scores.html

ivy i5's get smoked by the 8350 and is even with the ivy i7s. Only one faster is the 4770 and the 1000+ intel cups that don't even include a cooling fan.


http://www.overclock.net/t/1431032/top-cinebench-r15-cpu-scores

 

[8350rocks]

noob222 funny how an i5 4670k running at 3.4ghz scores 634 vs the fx8350 at 4.03ghz scoring 643.
http://www.cbscores.com/

WTF is cbscores? That's not a reputable site that I have ever heard of...

How is it all the "way off the beaten path" review sites I come across referenced on this forum are in your posts???

How about this:

http://www.hwbot.org/compare/processors#2741,2493,2743,2495,2689-57,94

I trust hwbot.org more than your site...

 

[-Fran-]

Hey hafi, that site says the 8350 is better than the i5... And the 8350 is even cheaper.

We can close the discussion now, kthxbai.

Cheers!

 

[etayorius]

No wonder why amd fanboys don't get ivy bridge i5's are 2x better per core. Did you not see an fx6300 lose to a 1st gen 2.8ghz i5 at battlefield 4. Now ivy bridge is gigantically better then that i5. IB i5 4 core 3.5ghz= PD 8 core at 3.5ghz multithreaded performance. Now intel has half the cores so in order to be as fast it needs to be 2x better per core and it is to match it multithreaded wise. My i5 4430 haswell is as fast as an fx8320 multithreaded wise do you know why 2.1x better ipc.

What you don`t seem to get is that Intel is only 40-50% faster, not 100 or 200%... i think you are comparing Clock Speeds of both Companies CPU`s, which is pretty much POINTLESS.

AMD just decided to use a different approach, AMD offers the Same Multithreaded performance and about 40% slower Single Threaded performance. Kaveri will shorten the lead Intel has over AMD in regards to IPC or Single Threaded performance.

It`s not 100-200%, all benchmarks and Games only see a 30-40% Lead in Intel CPUs, so ignore the Clock Speeds because at the end of the days they are irrelevant.

I will go as far as to say that i would buy a 6 Ghz CPU with better performance compared to any Intel CPU overall (Single and Multi Threaded) as long as it is priced the same or less than most Intel CPUs, i would care less if AMD CPUs has much higher clock compared to Intel, if you care about that then you may just be still living in the late 90s and early 2000, you`re a decade behind.

 

[8350rocks]

No wonder why amd fanboys don't get ivy bridge i5's are 2x better per core. Did you not see an fx6300 lose to a 1st gen 2.8ghz i5 at battlefield 4. Now ivy bridge is gigantically better then that i5. IB i5 4 core 3.5ghz= PD 8 core at 3.5ghz multithreaded performance. Now intel has half the cores so in order to be as fast it needs to be 2x better per core and it is to match it multithreaded wise. My i5 4430 haswell is as fast as an fx8320 multithreaded wise do you know why 2.1x better ipc.

What you don`t seem to get is that Intel is only 40-50% faster, not 100 or 200%... i think you are comparing Clock Speeds of both Companies CPU`s, which is pretty much POINTLESS.

AMD just decided to use a different approach, AMD offers the Same Multithreaded performance and about 40% slower Single Threaded performance. Kaveri will shorten the lead Intel has over AMD in regards to IPC or Single Threaded performance.

It`s not 100-200%, all benchmarks and Games only see a 30-40% Lead in Intel CPUs, so ignore the Clock Speeds because at the end of the days they are irrelevant.

I would go as far as to say it's closer to 15-20% single threaded deficit on average, and typically they only lose to the i7's in multi threaded performance by a lesser margin than that. In games it's typically a difference of 5-15% depending upon the game...

 

[8350rocks]

http://vr-zone.com/articles/amd-fx-8350-vs-intel-core-i7-3770k-4-8ghz-multi-gpu-gaming-performance/17494.html/1
http://vr-zone.com/articles/amd-fx-8350-vs-intel-core-i7-3770k-4-8ghz-multi-gpu-gaming-performance/17494.html/2

These 2 links prove ivy bridge is 2x ipc.

Those link have nothing to do with IPC.

Thanks for playing.

Now go home to your bridge until you understand what you're talking about. Have a nice day.

 

[8350rocks]

Did you not notice in games that don't use much threads the intel cpu scores double so 2x better fps. I wonder why as both are at 4.8ghz. Its the 2x ipc.

Maybe you misread my last post...

THE 2 LINKS YOU POSTED HAVE NOTHING TO DO WITH IPC.

THANKS FOR PLAYING.

NOW GO HOME TO YOUR BRIDGE UNTIL YOU UNDERSTAND WHAT YOU ARE TALKING ABOUT. HAVE A NICE DAY.

Is that clearer for you?

IPC is a hard wired limitation of the architecture itself, it has nothing to do with games.