ARM And AMD Partner On OpenCL
I have to admit that I was hoping for much more news from ARM's participation in AMD's currently held Fusion software developer conference in Seattle.
There is the persistent rumor that AMD may be licensing ARM architecture to make its way into the smartphone tablet space, but there was no confirmation of such a move at the conference. Instead, ARM and AMD are partnering in the OpenCL space to promote the craetion of GPU-accelerated apps.
At its conference, AMD announced a set of new OpenCL development tools that cater specifically to its Fusion APUs. The most interesting part of this announcement is the fact it was made by Manju Hegde, AMD's corporate vice president of AMD's Fusion experience program. Some readers may remember Hegde as the founder and CEO of Ageia, the company that invented the PhysX chip. Ageia was acquired by Nvidia in early 2008 and Hegde is now at AMD pitching OpenCL support, which is in direct competition to Nvidia's CUDA.
ARM's Jem Davies delivered a keynote at AMD's Fusion event and while there are obvious competitive edges between ARM and x86 products, the executive stresses that ARM and x86 are the only remaining "relevant" CPU architectures. Davies also pitched a hybrid processor approach that outlined CPU cores, parallel arrays and circuits that are dedicated to very specific functions, which obviously would favor highly parallel software that is written in, for example, OpenCL. It is a somewhat surreal experience to see ARM speaking at AMD's (x86) developer event and AMD could have simply invited ARM to annoy Intel. To see the partnership evolve is interesting, but the benefit to developers at the event was very limited.
I still believe that there is much more to come and those AMD-ARM rumors have some substance.
- Deals for June 15: Fallout New Vegas CE $12.50
- Crysis 2 Pulled from Steam, Now Origin Exclusive (Updated)
- Deals for June 14: LA Noire (PS3, Xbox 360) $40
- Newegg Rejects Best Buy's Copyright Infringement Claim
- Free to Play Games Now Available on Steam
- NEC Says Its Memory Saves Info Without Power
- BioWare Repies to Star Wars: The Old Republic ''WoW Clone'' Comments
- Samsung's Stunning Flexible AMOLED for Q2 2012
- Bill Gates Gladly Admits That He's a Geek, And His Kids Are Still Banned From Apple Products
- Mass Effect 3 May Have 4-Player Co-Op
- BioWare Hacked via Neverwinter Nights Forums
- Deals for June 16: Alienware M11x Core i3 $699
- Microsoft Hints To An August 2012 Win 8 Launch
- Pitchford: People Love Duke; St. John: Clueless Critics
- Deals for June 17: 42'' LG 120Hz 1080p LCD $499
- LulzSec: We Didn't Hack Sega; So We'll Defend Them!
- System Builder Meme Promo
- Intel 710 Lyndonville, 720 Ramsdale SSD Specs







If only the programs were to use OpenCL. I wish all the best to AMD! You will catch up!
"and AMD could have simply invited ARM to annoy Intel."
Didn't AMD, in the past, drive to press event(s) meant for Intel, and whisk 'em off (the press) to AMD's own showing right under intel's nose? I think they were lured off b/c AMD had actual working hardware. Might of been the early battles of 64 architecture... I cant recall.
I just want to say I completely condone and support this type of shenanigans.
so opencl is better than opengl?
one step forward to event intel vs the world
so opencl is better than opengl?
Open Compute Language v Open Graphics Language....former for doing general processing on a graphics card, the latter a graphics standard that used to compete with DirectX. Different standards for different purposes.
OpenCL is an API for parallel computing on both CPUs and GPUs, while OpenGL is an 3D graphics API. They have nothing in common other than that they leverage GPUs to get their work done.
I have been waiting a long time to finally see some more emphasis on OpenCL; it is a very good step in the right direction.
here we go.!
time to play ball.
This can be awesome. OpenCL would be über-nice if it reaches a good state-of-the-art. CUDA is only for NVIDIA cards while OpenCL can run on all cards and platforms (DirectCompute is DirectX only).
No thanks, there's no way OpenCL will become mainstream. If they want to leverage distributed computing on GPUs, do so at the compiler level so that we can leverage the already existing concurrent programming languages like OCaml, Scala, Erlang... There is no way more than a few individuals will bother going to OpenCL. Instead I'm simply going to use Tilera with its 64 cores, or Intel's Knight's Corner, which will allow me to use existing standard programming languages on their many-core hardware. Intel will simply ramp up its production of Knights Corner to 128, 256... cores, and that will be the end for OpenCL and CUDA.
opencl will succeed because of its competition.
opengl was great despite its competition.
opengl would still be widely used (i have a few games that use it, and are far better than the dx counterpart) but because microsoft produced a fud campaign, they stole the show.
as of now, opengl is about at dx10 and 11 quality graphics, but works on EVERY SYSTEM. while dx10 and 11 are windows 7 only (vista is no more... not worth mentioning)
Stop the misinformation about OpenGL. OpenGL was originally made for CAD/CAM and other industrial purposes. It was adapted for 3D games later as "3D Video Accelerator" cards became more common in PC's. As life went on and the industry matured Microsoft was able to quickly update, modify and fix DirectX to match up with what the HW manufacturers were pushing out and what software developers were wanting. Then OpenGL standard is governed by a body of representatives from various companies, most of them industrial in nature. They didn't want any updates to OpenGL as it would require that they update their own software / hardware, often costing hundred's of thousands to millions of dollars to purchase. Their voice prevented OpenGL from maturing and further then it had circa 1990's. It got so bad that companies like Nvidia had to write "extensions" that were additional libraries that provided features that OpenGL didn't have natively. Developers were encouraged to use these "extensions" whenever possible. When you have to use a third party's propriety "extensions" on your "open" standard for it to be competitive, then you know there is a problem.
Eventually another group took over maintaining / modifying the OpenGL standard and they had some very good ideas. But the original industrial group had veto power over any chances, and they threatened to use it liberally. So the fight over the "next" version of OpenGL went back and forth for a few years, all the while MS is perfecting DirectX and software developers had all switched to DirectX. When the OpenGL group finally was able to release OpenGL 3.0, it was lacking in many of the originally desires features and DirectX 9 was dominating everything. Now OpenGL has mostly been rendered irrelevant and its most common use is as a wrapper for Wine's DirectX emulator, or as the engine behind various GUI's in Linux. The industrial developers are still using their old OpenGL library's and they have ~zero~ incentive to move beyond that.
The "rule by committee" method doesn't work in a rapidly changing world. You can't have a counsel of representatives deciding the future of a standard, especially when their at cross purposes and in competition with each other. That would be like having Intel, AMD and Via all on a counsel trying to decide the next ISA. It would be a total and utter failure as each (especially Intel) would try to create the standard in such a way as to give itself an advantage over the others.
No thanks, there's no way OpenCL will become mainstream.
Whatever ARM supports will be the de facto mainstream which is why AMD is wise to partner up with them for development of openCL. ARM processors account for around 90% of all processors sold in the world, so their influence is not to be taken lightly.
You have a typo. In the First paragraph/second sentence you have the word "creation" spelled "craetion".
good news, alway's wait bulldozer out
OpenGl is now at version 4.1 and continues to develop. It maybe playing catch up a little to DiretX after all the debating however, as it's name states, it's open, not closed source and under Microsoft rule like DirectX. It has started evolving a lot more over the last few years and is still a very good api to the point where you can use it as an alternative to DirectX etc. http://www.opengl.org/documentation/current_version/
OpenGl is now at version 4.1 and continues to develop. It maybe playing catch up a little to DiretX after all the debating however, as it's name states, it's open, not closed source and under Microsoft rule like DirectX. It has started evolving a lot more over the last few years and is still a very good api to the point where you can use it as an alternative to DirectX etc. http://www.opengl.org/documentation/current_version/
Yes this is true that it is open, and it is evolving, but so is Direct X. I doubt that the industry has much incentive to roll back to OpenGl at the moment. Most games are still on the MS platform for PC, and those that are on the mac show significant less quality, with the differance being Open GL.
The fact is Open GL works great for work loads, but it does not work so great for games. MS has already developed a flexable platform that supports alot of features, and can be easily used to port over to thre own consols easily, and from my understanding most other consols use propiatory APIs, meaning the open standerd of OpenGl gives them no significant advantage, and so companies come up with something that might.
The fact is there Is not much advantage using OpenGL, unless they come out with a way to port applications from the PC directly to smart phones, which is the only mass consumer item that I actually spot in the wild that can use open GL. Even with this feature you have problems with scaling from large moniter to small, and compatability across the board since you move to an entirely differnat form facter.
OpenCL stands a chance its only real compatition is locked down to a single hardware manufacturer, so I dont see it getting ported across the board to support its competiters. The project is supported by industry, and the industry is still developing. Not to mention AMD has alot to gain from moving more of the workload to the GPU then any one else out there, so you can bet there going to do everything they can to draw developers.
Since there is no Game industry equivalant to drive OpenCL at this point they still have wiggle room to fumble a little bit, it is more understanding to have delays on a up and comming technolagy then it is for something that is perceved as masterd by there respective companies. we dont expect air bus or boing to have years of delays on a refresh of one of there desings, but tell them to make a light saber, and we think... "6 months to a year delay? Only thats really good."
same here people will see games, and companies wount want to experament with what works well for them, however if you introduce something that allowes them to shoot more of that code at the GPU then you might spark there intrest.
you can also bet that the new Llano chips are also turning heads. They support OpenCL, and they last a long time with that GPU under load. If this chip takes off the use of OpenCl can potentially do alot to drop system requirements on what would otherwise be hard to run softwear.
Stop the misinformation about OpenGL. OpenGL was originally made for CAD/CAM and other industrial purposes. It was adapted for 3D games later as "3D Video Accelerator" cards became more common in PC's. As life went on and the industry matured Microsoft was able to quickly update, modify and fix DirectX to match up with what the HW manufacturers were pushing out and what software developers were wanting.
Yes, OpenGL was created for CAD and professional visualization. And yes, Microsoft was able to quickly improve Direct3D (DirectX is not just graphics). However, Microsoft seemingly improved by light-years because Direct3D was SO BAD to begin with. Microsoft had NO IDEA what game developers wanted from 3D graphics, that is why many in the mid 90s went with OpenGL. And if you actually look at the way the Direct3D API has evolved you will notice one thing: each version became more and more like OpenGL!
Also, you did not mention that Microsoft has tried to cripple OpenGL at every chance. For example, backing away from Fahrenheit which would have united OpenGL and Direct3D, reversing their commitment to provide DirectDraw bindings for OpenGL, and removing OpenGL support in their basic drivers.
The fact is Open GL works great for work loads, but it does not work so great for games.
Sorry, but anyone that has actually used OpenGL for games knows that is a false statement.
Also, I guess all those id Software based games since Quake II are not so good on the graphics? Have you seen the latest demo of idTech 5 engine?
One more thing: the OpenGL API is available natively on Windows, Linux and other Unix-like OSs, iOS (iPad, iPhone, etc.), OSX, Android, and even PlayStation 3. And with third-party libraries you can use it on Xbox, Wii, and Windows CE/Mobile/Phone. With the push for 3D games and apps on portable devices Direct3D is at a BIG disadvantage.
There is no way more than a few individuals will bother going to OpenCL. Instead I'm simply going to use Tilera with its 64 cores, or Intel's Knight's Corner, which will allow me to use existing standard programming languages on their many-core hardware. Intel will simply ramp up its production of Knights Corner to 128, 256... cores, and that will be the end for OpenCL and CUDA.
Just making a multicore CPU x86 compatible does not mean existing code will utilise all cores. Programmers will still need to rewrite and optimise their code but just not as much as a complete overhaul. Other aspect is not all situations are ideal for the x86 instruction set. Also it would be nice to broaden the industry in high performance computing and allowing the 40+ chip companies that make non-x86 chips compete against the two that do could really benefit everyone in the long run.
wasnt this in the news just a few months ago, when the rumor about amd switching to arm processors was laid to rest?
Sorry, but anyone that has actually used OpenGL for games knows that is a false statement.Also, I guess all those id Software based games since Quake II are not so good on the graphics? Have you seen the latest demo of idTech 5 engine?One more thing: the OpenGL API is available natively on Windows, Linux and other Unix-like OSs, iOS (iPad, iPhone, etc.), OSX, Android, and even PlayStation 3. And with third-party libraries you can use it on Xbox, Wii, and Windows CE/Mobile/Phone. With the push for 3D games and apps on portable devices Direct3D is at a BIG disadvantage.
I've always found id's graphic engines to be very watery in appearence or rather the textures look muddy. Although the game engines have impressive features, I've never liked the visuals much when compared to say Havak and other engines out there doing DirectX. For me, Direct3D > OpenGL - but that may not be saying much.
hmm a pair or quad of x86's with a pair of arm chips and a pair of gpu's would be an interesting if it was all on one cpu die.
http://blogs.msdn.com/b/somasegar/ [...] d-ppl.aspx
On graphics for games, I would say DirectX is the only choice. OpenGL just isn't there for consumer graphics cards. I don't think its a mystery that consumer graphics cards render OpenGL worse then Workstation cards. While directX works better on consumer cards.
The issue here is 2 fold. Consumer Graphics cards have to be designed for OpenGL rather then DirectX in the first place. This will most likely result in a higher cost at first. When you are talking about playing any games in 3D, you are talking about using a nVidia or AMD graphics card as well. So the smartphones are not yet a factor until both companies start to make talks with ARM architecture. Even then the transition will take a couple years.
Second DirectX 11 is simply a much more robust API for games. You can do more with the shaders, its backwards compatible, and its not as huge a performance hit.
I think this will definetly be a positive. I view OpenCL as a necessity over its alternatives. Maybe DirectCompute will be better. Cuda really should not be taken seriously outside of single platform usage.
Yes, OpenGL was created for CAD and professional visualization. And yes, Microsoft was able to quickly improve Direct3D (DirectX is not just graphics). However, Microsoft seemingly improved by light-years because Direct3D was SO BAD to begin with. Microsoft had NO IDEA what game developers wanted from 3D graphics, that is why many in the mid 90s went with OpenGL. And if you actually look at the way the Direct3D API has evolved you will notice one thing: each version became more and more like OpenGL!
Also, you did not mention that Microsoft has tried to cripple OpenGL at every chance. For example, backing away from Fahrenheit which would have united OpenGL and Direct3D, reversing their commitment to provide DirectDraw bindings for OpenGL, and removing OpenGL support in their basic drivers.
All I hear is "rah rah I hate the Evil Microsoft" and so forth. Bashing MS was old in 2003
The original DirectX sucked, everyone knows this. The point was that MS was able to quickly fix and adapt it for what developers wanted and for what HW manufacturers provided. OpenGL just stayed there and didn't do anything. Eventually DirectX 7 was released and for the first time it overcame OpenGL is capabilities, then DX8, then DX9. The entire time OpenGL was going back and forth with changes stuck "in committee". By the time OpenGL was updated DX10 and 11 were already on the horizon and the industry had already chosen DirectX as the standard. NVidia even tried to save OpenGL by releasing "Extensions" to provide features that the standard didn't support yet. This ultimately failed as NVidia is a single manufacturer and their extensions only support their own HW. MS didn't need to pull any shenanigans, the OpenGL community did themselves in with the constant bickering and in-fighting of what to add and not add. You want someone to blame, then blame the industrial companies that controlled the OpenGL standard, they are the ones that did not want it updated.
DirectX is the standard now because it won the battle. Its an "open" platform, anyone can go and download the binaries, anyone can use it in developing software. Its controlled by a single entity and thus only available on the platforms that entity wants. But anyone's free to reverse engineer it and provide support for DirectX on other platforms.
And BTW, MS didn't handicap OpenGL in any way, your drinking the koolaid. OpenGL acceleration support is provided by the HW manufacturers themselves. The OS only has software rendering support for use when there is no HW rendering available. This is a requirement by the OpenGL standard but not by DirectX (although there is DirectX software rendering engine). MS's OpenGL Software Rendering sucks, everyone knows this, so does their Software DirectX Rendering. Neither of those is capable of playing a "3D" game with any notion of speed, we're talking single digit FPS at low resolution settings. Instead games use the HW provided OpenGL rendering engine, which MS has absolutely nothing to do with. Anyone who used to do OpenGL back in the late 90's and early 2000's would know this. NVidia provided excellent OpenGL support while ATI provided absolute sh!t OpenGL support.
But please, don't let any of this stop you from spreading the hateraid around.
OpenCL stands a chance its only real compatition is locked down to a single hardware manufacturer, so I dont see it getting ported across the board to support its competiters.
This is the only thing I really have to take issue with you on. AMD can license the physx engine from nvidia they just chose not to. Plus dont forget nvidia gives out their engine for free to sell more cards because the more games that use their engine the more cards they can sell to accelerate those games. The entire hardware acceleration was brought from an ASIC to the gpu via the cuda platform that nvidia cards run on. If amd wanted to run physx they could license it from nvidia but amd is backing bullet and openCL.
I do find it interesting that the former head of ageia is now pushing for openCL... sour grapes mayhaps? If his company wanted to stay relevant they should have had a pci-express version of their card because pci with its 33mB/sec transfer rate just isnt fast enough to send all that data to the cpu and then to the gpu... but they didnt and their company was turned into an after thought that was just low hanging fruit so to speak for a big company to pick up. Nvidia bought them and AMD (or even ATi back then) didnt. Now if Nvidia were to come out with a pci-e say 1x or 2x or even 4x physx ppu card for amd users to put in their systems could you foresee the massive influx of dollars into nvidia's bank account? That right there might kill bullet and openCL and all it would take would be taking the ageia PPU and updating it to pci-e.
All I hear is "rah rah I hate the Evil Microsoft" and so forth.
I am sorry you felt I was bashing Microsoft. Most of my programing over the years has been based on Windows platforms, and I still prefer Visual Studio as my main IDE. However, I have kept my eyes wide open, have kept myself open to other platforms, and have actually lived through most of Microsoft's bad choices. Did you?
The entire time OpenGL was going back and forth with changes stuck "in committee". By the time OpenGL was updated DX10 and 11 were already on the horizon and the industry had already chosen DirectX as the standard. NVidia even tried to save OpenGL by releasing "Extensions" to provide features that the standard didn't support yet.
Another correction here: Extensions have always been part of the OpenGL API. SGI realized that hardware advances might outpace updating the API. This is an advantage over D3D, where hardware manufacturers must petition Microsoft to add features to the next version.
And BTW, MS didn't handicap OpenGL in any way, your drinking the koolaid. OpenGL acceleration support is provided by the HW manufacturers themselves. The OS only has software rendering support for use when there is no HW rendering available...Anyone who used to do OpenGL back in the late 90's and early 2000's would know this. NVidia provided excellent OpenGL support while ATI provided absolute sh!t OpenGL support.But please, don't let any of this stop you from spreading the hateraid around.
Again, the broken Microsoft promise of DirectDraw bindings greatly hurt OpenGL full-screen games in the late 90's. It is VERY difficult to reliably change desktop resolution and bit depth without using DirectDraw.
Now, beginning with Vista Microsoft provided an inefficient OpenGL 1.4 wrapper over D3D and encouraged graphics manufacturers to no longer provide their own ICD. Microsoft did this because Aero uses D3D to render the desktop and they provided no interface to allow OpenGL windows rendered by an ICD.
Again, I do not hate Microsoft, otherwise I would purge my home and business of Microsoft products, but I do hate some of their actions and poor choices they have made to try to protect and hold on to their monopoly and power they exert over the PC industry.
On graphics for games, I would say DirectX is the only choice. OpenGL just isn't there for consumer graphics cards. I don't think its a mystery that consumer graphics cards render OpenGL worse then Workstation cards. While directX works better on consumer cards.
The only difference between a consumer and workstation graphics card, at least those based on NVIDIA and ATI GPUs, is the driver, not the hardware.
Workstation OpenGL drivers must pass MANY more extensive tests to insure they render perfectly and are compatible with the major professional CAD and graphics packages. This extensive driver testing is actually the major reason workstation cards cost so much more than consumer cards, and the reason that often OpenGL games will run slightly FASTER on a consumer card than the equivalent workstation card.
This is good news, hope a deeper partnership will evolve between AMD and ARM.
Hope AMD would acquire an ARM license so I imagine an ARM based bulldozer processor.