AMD CPU speculation... and expert conjecture

[Embra]

http://www.cpu-world.com/news_2014/2014041801_AMD_Beema_low_power_APUs_spotted.html

AMD "Beema" low power APUs spotted

 

[juanrga]

http://www.cpu-world.com/news_2014/2014041801_AMD_Beema_low_power_APUs_spotted.html

AMD "Beema" low power APUs spotted

Dr. Su confirms that shipping to notebook makers started last quarter. Beema will be also offered as update for the new AM1 platform

http://www.fudzilla.com/home/item/34540-amd-shipping-beema-apus

 

[jimmysmitty]

I honestly love how people talk about developers wanting to move away from Microsoft, yet they have been able to. Why have they not moved on to OpenGL which supports most of the features DX11 does and is on Windows, Linux and Mac OSX? That is the question. Why would it take an API from AMD when there already was a viable alternative? Could it be that people are talking about what they think devs want rather than what devs really want?

I also do not want to go to a point where my GPU doesn't support all the features. Honestly one of the downsides (it is also a up side) about PCs is the insane amount of different combinations. It is why no dev can optimize their games for all hardware. I don't want another to be that one GPU doesn't support said API that gives said feature but does have this other feature.

We will see in the long run but I don't see AMD pushing Mantle too much. Honestly they don't have a lot of money and they are not making it hand over fist, Intel is outselling in CPUs and NVidia is outselling in GPUs. They need to make sure they don't waste money on ventures that wont be profitable.

And again, I will wait till I see third party independent results for ARM vs x86 in servers, not something AMD is saying. It is the same as AMD saying their CMT design is superior to another design, sure they are going to say that. You expect AMD to build a bunch of ARM CPUs and then trash talk ARM?

 

[jdwii]

Bro i'm with you jimmy 100% actually not too many people want to deal with the headaches of linux from a software support standpoint and if or when it does happen it will be done by some other source paying that company to make it happen. However i feel like its best not to get into the linux vs windows thing again since i can't keep my mouth shut when certain users come up with interesting theories surrounding the discussion.

 

[jimmysmitty]

Bro i'm with you jimmy 100% actually not too many people want to deal with the headaches of linux from a software support standpoint and if or when it does happen it will be done by some other source paying that company to make it happen. However i feel like its best not to get into the linux vs windows thing again since i can't keep my mouth shut when certain users come up with interesting theories surrounding the discussion.

I agree to stay away from that topic. I just pointed out that developers have always had the ability to move away from DirectX but they only will once it will benefit them in the long run.

Nothing AMD or Microsoft does will really push that.

As well, so many people are still claiming, although we have no way to be sure, that DirectX 12 is a result of Mantle and that Microsoft had given up on PC gaming.

It is all a waiting game though. That is all everything comes down to.

 

[amdfangirl]

I don't want another to be that one GPU doesn't support said API that gives said feature but does have this other feature.

Well it's the just in case principle, isn't it?

Way back when, I remember arguing with a friend who wanted an Nvidia GPU in case PhysX/CUDA took off.

Must be the same with Mantle (though it probably has more merit). Better buy AMD, don't want to miss out on Mantle!

Who wants to miss out? This is a marketing strategy...

 

[etayorius]

I honestly love how people talk about developers wanting to move away from Microsoft, yet they have been able to. Why have they not moved on to OpenGL which supports most of the features DX11 does and is on Windows, Linux and Mac OSX? That is the question. Why would it take an API from AMD when there already was a viable alternative? Could it be that people are talking about what they think devs want rather than what devs really want?

Well you got Notch, Johan, Gabe, Carmack (when he was in ID) and a Myriad of other Indie Devs out there making big noises about wanting to get rid as far as they can from MS... and SteamOS was the answer to that, currently AMD, nVidia and Intel promised to support OpenGL heavily from now on, you also got a whole bunch of Publishers like Activision, EA, Eidos and others promising to support MANTLE in most of their Modern Engines, i think that speaks heavily as to what Devs want.

 

[jdwii]

"indie devs" Its like a well known fact that most programmers like anything beyond windows more. The truth is as long as most PC gamers use windows it will NEVER change. Really what's so bad about microsoft besides their cost? I get mad easily and honestly i only ever gotten made at W8(not 8.1) and xp.

 

[juanrga]

Some generic myths about linux and windows were refuted before, now I will focus on gaming and what devs want:

Valve CEO: Why Linux is the future of gaming

More Game Developers Unhappy With Windows 8

Microsoft Responds to Windows 8 Hate From Game Devs

How Microsoft is trying—but failing—to court indie game developers

Unreal Engine 4 now supports Linux and SteamOS: Is this the year of desktop Linux?

CRYENGINE adds Linux Support as Crytek Prepare to Offer New Possibilities at GDC

Alienware Collaborates with Crytek to Provide New Experiences for Gamers

One Year On: Steam for Linux Content Increases by 900%

Sid Meier's Civilization: Beyond Earth Is Coming To Linux

Star Citizen Will Be Coming To Linux

Yes! I can see every dev loves windows and hates linux

About AMD and its new ARM CPUs... well you must take the claims with a grain of salt, iff you don't know anything about ARM. Using well-known benchmarks of A15 and architectural differences of the new A57, I predicted the performance of AMD Seattle many time ago in this same thread. Some posters did laugh and claimed that ARM will never caught x86 in performance, but the recently released server benchmarks confirmed my prediction and destroyed theirs.

Of course the A57 cores used in AMD Seattle are not the more powerful ARMv8 cores. E.g. Apple cyclone core is so powerful that two of them clocked at 1.3GHz match the performance of 1.5--1.7GHz x86 quad-cores from Intel/AMD, even when the x86 cores had a performance bonus of higher thermal limits.

Performance: Cyclone > A57 > Jaguar/Silvermont

Posterior anand analysis found that the cyclone core is not still showing its full performance (being capped by both available software and surrounding hardware such as the memory subsystem). Anand analysis found:

Apple didn't build a Krait/Silvermont competitor, it built something much closer to Intel's big cores. At the launch of the iPhone 5s, Apple referred to the A7 as being "desktop class" - it turns out that wasn't an exaggeration.

And of course, we know that Apple is preparing to migrate away from Intel desktop processors. Probably when finfets are ready.

We know that the guy behind Apple cores is now the "Chief Architect of AMD's Microprocessor Cores". This is part of a recent interview to him:

AMD is currently undergoing a very public and intensely scrutinized reorganization, shrinking into profitability. This extends to Jim Keller's cores groups; one of the first tasks he completed was his own internal reorganization and responsibility shuffle to align his group's efforts best with the new efforts being made by the company. As we now know, that includes ARM cores, in the form of both IP for hardware security zone functionality and 64-bit ARMv8 for servers; an AMD Opteron branded ARM processor is expected in the next 12 months.

[...]

The best part of interview came in a nice little tidbit about core performance while discussing how much market and application awareness plays a role in core design. Many things are incremental, one of which is legacy performance on new designs. Jim confidently stated AMD are on track to catch up on high performance core, a function of design improvements. We couldn't pin down a timeline for this, but with a time scale of two years core design and one year build and test, it's not going to be immediate. My expectation is 2015.

Only a comment, my expectation is not 2015, but 2016/17.

http://www.rage3d.com/articles/hardware/amd_worldcast/

 

[juanrga]

After measuring CPU and graphics performance, now it is time to measure OpenCL performance in AM1 platform

http://www.phoronix.com/scan.php?page=article&item=amd_am1_opencl&num=2

 

[juanrga]

Numbers as those below are the reason why I have no faith on any Mullins-based tablet from AMD

http://seekingalpha.com/article/1955211-intel-please-give-up-on-mobile

My suspicions again confirmed.

http://www.pcworld.com/article/2145980/amd-steers-clear-of-lowcost-tablet-market.html

 

[de5_Roy]

AMD Mantle and NVIDIA 337.50 Scaling Demonstrated with Star Swarm on AM1
http://www.pcper.com/reviews/Graphics-Cards/AMD-Mantle-and-NVIDIA-33750-Scaling-Demonstrated-Star-Swarm-AM1

amd doing good but the ghost of glofo wafer agreement hangs over...
http://techreport.com/news/26338/amd-posts-another-loss-but-beats-wall-street-forecast

Samsung's 28'' display serves up single-tile 4K at 60Hz for $800
http://techreport.com/news/26337/samsung-28-display-serves-up-single-tile-4k-at-60hz-for-800\
i wonder if this'll have a long term [strike]consequence[/strike] impact.

lanoc's kaveri review
http://lanoc.org/review/cpus/6848-amd-a10-7850k-kaveri

GlobalFoundries licenses Samsung process tech, grants AMD access to FinFETs
http://techreport.com/news/26336/globalfoundries-licenses-samsung-process-tech-grants-amd-access-to-finfets
seems glofo cancelled their 14nm xm process in favor of samsung's 14nm finfet. considering glofo's track record.. prolly a good idea. or 14nm-xm may have been doa.
i hope amd takes advantage of this opportunity. they'll have access to both tsmc and glofo's leading processes this way.

 

[ohyouknow]

Hrmm, when those 4k prices drop a little more I'll be passing over 1440p to 4k from 1080p. Purchasing a new GPU to run it though, will not be so fun.

 

[jdwii]

People can't jump to the conclusion(fallacious argument) its over dev's hating or wanting to move away from Microsoft for one all those dev's are doing that because of steam a central system based on linux, Its still a question on how that will work out i hope it does. That way we can have linux vs windows wars when gaming on both platforms will be used, first we need something that isn't OpenGL since its too bloated and not nearly as good as Mantle or directx and then we need somthing that everything will work on when it comes to hardware like Directx and OpenGL.

 

[Cazalan]

Which begs the question: Does preemption in all cases make sense anymore? Wouldn't it make sense, for multi-core systems, to allow the programmer to specify threads that can NEVER be preempted unless manually specified, for the purposes of maximizing performance?

Isn't that essentially what the new consoles are doing? These 2 CPUs are for the OS, these 4 are for the game, and these 2 are for "background" tasks. But yeah, I'm sure there is plenty of work to do with creating more efficient task schedules. Like how there are GPU profiles today to give optimal graphics performance, you could have CPU profiles to give optimal game performance.

 

[-Fran-]

As any big company wants, they don't want to throw money away. If they develop for "Linux / GNU Linux", they'll have to invest a lot more than for Windows (currently) and it's a very risky approach. That's where Gabe came in telling everyone he'll create a stable ecosystem based on GNU Linux for game developers, just like Google did with Android.

In the specific case of game development, we can stop talking about GNU Linux and start talking about Android/SteamOS/Ubuntu. I'm guessing that will leave less room for misinterpretation and we can drop this stupid "GNU Linux vs Windows" argument.

So, in summary, Devs/Publishers want a safeguard to their investment more than anything else. As long as Microsoft gives a somewhat better ecosystem than Linux, no sane Publisher or Dev will just jump ship to Mac or Linux-based-OSes. The tools around each platform to develop is in like a 3rd or 4th position, so MANTLE being for Linux or not, makes little to change the landscape of "wanting out" from Windows IMO.

Cheers!

 

[juanrga]

AMD Mantle and NVIDIA 337.50 Scaling Demonstrated with Star Swarm on AM1
http://www.pcper.com/reviews/Graphics-Cards/AMD-Mantle-and-NVIDIA-33750-Scaling-Demonstrated-Star-Swarm-AM1

I find lovely that the eXtreme i7 is about twice faster than the cheap Athlon AM1 under DX , but under MANTLE the $30 Athlon offers about the same FPS than the $1000 i7.

amd doing good but the ghost of glofo wafer agreement hangs over...
http://techreport.com/news/26338/amd-posts-another-loss-but-beats-wall-street-forecast

This was given and commented before.

lanoc's kaveri review
http://lanoc.org/review/cpus/6848-amd-a10-7850k-kaveri?showall=&start=2

Lovely that another review find the Kaveri A10 CPU is at i5-2500k level as predicted time ago.

GlobalFoundries licenses Samsung process tech, grants AMD access to FinFETs
http://techreport.com/news/26336/globalfoundries-licenses-samsung-process-tech-grants-amd-access-to-finfets
seems glofo cancelled their 14nm xm process in favor of samsung's 14nm finfet. considering glofo's track record.. prolly a good idea. or 14nm-xm may have been doa.
i hope amd takes advantage of this opportunity. they'll have access to both tsmc and glofo's leading processes this way.

Glofo 14XM was finfet process and AMD was ready to tape out. Glofo has canceled it and licensed Samsung 14LPP/E process, because Samsung schedule was more aggressive and the process has better properties.

This means that AMD will have access to finfets before planned and that the process will be better than expected.

 

[juanrga]

That way we can have linux vs windows wars when gaming on both platforms will be used, first we need something that isn't OpenGL since its too bloated and not nearly as good as Mantle or directx

Of course of course

Who needs DirectX? AMD, Nvidia, and Intel team up to demo ultra-efficient OpenGL

NVIDIA, AMD, Intel Explain How OpenGL Can Unlock 15x Performance Gains

AMD, Intel and Nvidia join forces to push OpenGL for games developers

Some quotes:

They also mentioned during their presentation that with proper implementations of these high-level OpenGL techniques, driver overhead could be reduced to almost zero. This is something that Nvidia's software engineers have already claimed is impossible with Direct3D and only possible with OpenGL (see video below).

The techniques presented apply to all major vendors and are suitable for use across multiple platforms

OpenGL can cut through the driver overhead that has been a frustrating reality for game developers since the beginning of the PC game industry. On desktop systems, driver overhead can decrease frame rate. On mobile devices, however, driver overhead is even more insidious, robbing both battery life and frame rate.

 

[jdwii]

That way we can have linux vs windows wars when gaming on both platforms will be used, first we need something that isn't OpenGL since its too bloated and not nearly as good as Mantle or directx

Of course of course

Who needs DirectX? AMD, Nvidia, and Intel team up to demo ultra-efficient OpenGL

NVIDIA, AMD, Intel Explain How OpenGL Can Unlock 15x Performance Gains

AMD, Intel and Nvidia join forces to push OpenGL for games developers

Some quotes:

They also mentioned during their presentation that with proper implementations of these high-level OpenGL techniques, driver overhead could be reduced to almost zero. This is something that Nvidia's software engineers have already claimed is impossible with Direct3D and only possible with OpenGL (see video below).

The techniques presented apply to all major vendors and are suitable for use across multiple platforms

OpenGL can cut through the driver overhead that has been a frustrating reality for game developers since the beginning of the PC game industry. On desktop systems, driver overhead can decrease frame rate. On mobile devices, however, driver overhead is even more insidious, robbing both battery life and frame rate.

Wonder if Mr. Bill gates can just talk about anything and have people believe it? Beliefs are a funny thing however the ones who take "what if" statements should just join ken ham now and except that the world is only 6 thousand years old.... results or GTFO.

 

[jdwii]

AMD Mantle and NVIDIA 337.50 Scaling Demonstrated with Star Swarm on AM1
http://www.pcper.com/reviews/Graphics-Cards/AMD-Mantle-and-NVIDIA-33750-Scaling-Demonstrated-Star-Swarm-AM1

I find lovely that the eXtreme i7 is about twice faster than the cheap Athlon AM1 under DX , but under MANTLE the $30 Athlon offers about the same FPS than the $1000 i7.

amd doing good but the ghost of glofo wafer agreement hangs over...
http://techreport.com/news/26338/amd-posts-another-loss-but-beats-wall-street-forecast

This was given and commented before.

lanoc's kaveri review
http://lanoc.org/review/cpus/6848-amd-a10-7850k-kaveri?showall=&start=2

Lovely that another review find the Kaveri A10 CPU is at i5-2500k level as predicted time ago.

GlobalFoundries licenses Samsung process tech, grants AMD access to FinFETs
http://techreport.com/news/26336/globalfoundries-licenses-samsung-process-tech-grants-amd-access-to-finfets
seems glofo cancelled their 14nm xm process in favor of samsung's 14nm finfet. considering glofo's track record.. prolly a good idea. or 14nm-xm may have been doa.
i hope amd takes advantage of this opportunity. they'll have access to both tsmc and glofo's leading processes this way.

Glofo 14XM was finfet process and AMD was ready to tape out. Glofo has canceled it and licensed Samsung 14LPP/E process, because Samsung schedule was more aggressive and the process has better properties.

This means that AMD will have access to finfets before planned and that the process will be better than expected.

 

[jimmysmitty]

juanrga, seriously lay off the Linux. You missed my point completely. If developers wanted to use something other than DX it has been in front of them, I was not mentioning a OS specifically as we went through this already multiple times.

As for ARM, lets stick off of it unless we are comparing it directly in a review for say AMDs actual ARM based CPUs vs their or Intels server CPUs.

As for the mantle results posted, all it shows is that Mantle does what it is supposed to and alleviates that CPU bottleneck. In the future that Intel CPU will fare better than the AMD CPU as there will be a point where the CPU will still bottleneck no matter how much of the GPU you can actually utilize.

 

[blackkstar]

The reason why people aren't rushing to OpenGL is that Khronos Group is notoriously slow at adopting new features. Does no one remember when OpenGL got completely crushed by DX in terms of features (like tessellation)?

As a game developer, you had two choices.

1. Go for OpenGL, port to DirectX anyways for Xbox. Lose a ton of features and performance
2. Just go with DirectX anyways.

Mantle is a solution that lets you just recompile a lot of DX code. It also completely bypasses having to go through a third party to get changes to the API.

Imagine Crytek wants to add a new feature to Cryengine. WIth DirectX, Crytek can petition MS to add a new feature and wait for them to do it. Then, once the specification is released, AMD and Nvidia can create the hardware to take advantage of the new features.

With something like Mantle, game developer goes up to AMD and asks for a feature. AMD tells Crytek what they can do since they actually make the GPU and control the complete software stack from low level chip to API. Then it gets added.

It's a lot bigger deal than a lot of you think. Not only is it about escaping MS, but it's about getting better access to changes in the API you're using.

I don't think some of you understand the situation with Windows 8.

It is along the lines if Tom's Hardware decided it was time to optimize their forums for touch. So they completely redesign everything, remove ability to type long posts, and just flat out dumb it down and change it.

Do you still stay? Or do you want out?

You can call it falacious logic if you want (although I fail to see what type of fallacy it is), but the evidence is that game developers are pushing for alternatives to Windows very aggressively right now. They've been able to do this for 10+ years and they never have.

Don't you think that the newfound effort to escape Windows at a time when Microsoft has release a version of Windows that's basically ignoring the form factor PC games target is more than just a coincidence? Specially after gaming on Linux has been possible for a very long time.

If anything, I would say that the folks calling this fallacious are the ones ignoring the arguments being made and are instead just calilng the argument itself fallacious.

If you have evidence that this new found push for Linux gaming, like Mantle, SteamOS, Linux ports, etc has absolutely nothing to do with Windows 8 and Microsoft's current direction with Windows, please show me, because I'd love to see it.

 

[juanrga]

Glofo 14XM was finfet process and AMD was ready to tape out. Glofo has canceled it and licensed Samsung 14LPP/E process, because Samsung schedule was more aggressive and the process has better properties.

This means that AMD will have access to finfets before planned and that the process will be better than expected.

Indeed! Samsung is ready for volume production of 14nm finfet by the end of this year, whereas Glofo will be volume producing on 14nm finfet at early 2015.

http://www.eetimes.com/document.asp?doc_id=1321974

This could give AMD the possibility to skip the previously scheduled 20nm node for some of the products and go for 14nm finfet closing the process gap with Intel. Carrizo APU on 14nm finfet is a possibility.

 

[juanrga]

The reason why people aren't rushing to OpenGL is that Khronos Group is notoriously slow at adopting new features. Does no one remember when OpenGL got completely crushed by DX in terms of features (like tessellation)?

As a game developer, you had two choices.

1. Go for OpenGL, port to DirectX anyways for Xbox. Lose a ton of features and performance
2. Just go with DirectX anyways.

Anand has recent article that explains why developers are pushing OGL. In short: OGL is faster and less complex than directX, whereas adding multiplatform support.

http://www.anandtech.com/show/7890/return-of-the-directx-vs-opengl-debates

 

[8350rocks]

I honestly love how people talk about developers wanting to move away from Microsoft, yet they have been able to. Why have they not moved on to OpenGL which supports most of the features DX11 does and is on Windows, Linux and Mac OSX? That is the question. Why would it take an API from AMD when there already was a viable alternative? Could it be that people are talking about what they think devs want rather than what devs really want?

I also do not want to go to a point where my GPU doesn't support all the features. Honestly one of the downsides (it is also a up side) about PCs is the insane amount of different combinations. It is why no dev can optimize their games for all hardware. I don't want another to be that one GPU doesn't support said API that gives said feature but does have this other feature.

We will see in the long run but I don't see AMD pushing Mantle too much. Honestly they don't have a lot of money and they are not making it hand over fist, Intel is outselling in CPUs and NVidia is outselling in GPUs. They need to make sure they don't waste money on ventures that wont be profitable.

And again, I will wait till I see third party independent results for ARM vs x86 in servers, not something AMD is saying. It is the same as AMD saying their CMT design is superior to another design, sure they are going to say that. You expect AMD to build a bunch of ARM CPUs and then trash talk ARM?

Because OGL is a convoluted P.I.T.A. is why.

DX is FAR easier to use, and get consistent performance. Additionally, drivers far better support DX at this time.

In D3D you have generally 1 best way to do things, whether you like it or not, it is either the fastest way, or yields best results to do it that way.

In OGL, you have 4-5 different ways to do things, and none of them are the fastest or yields the best results, it is just the way that the guy who did it wanted to do it.

So, if you change a coder at some point, say the guy who did most of your OGL code leaves, then the guy who comes in behind him now has to figure out what he was thinking. Considering many coders do not actually leave much documentation, you can lose months worth of time if the new guy wants to do it a different way that is perceived to be better.

Now, OGL offers most of the functionality you get from D3D, however, the API is convoluted, rife with legacy code, and all kinds of other stuff that should have been cleared out and rewritten long ago. Also, OGL and D3D use different languages...so you cannot easily port your work from D3D to OGL...(M$'s blatant intention at that...)

Fast forward to MANTLE, which now allows you to take D3D code, make 20 minutes worth of changes, and plug it right in.

Which one sounds easier to develop a secondary render path on the back end to you?