What GPGPU applications are available to ATI users?

[Scali]

Havok has bigger marketshare than PhysX and because Havok is own by Intel you can bet it will never use Cuda for GPU accleration. Yet ATI has good reltaionship with Havok so they'll be using either OpenCL or DirectCompute. Whichever will better benefit Intel's IGPs and the other "Larrabee" projects that are still active.

Larrabee will benefit more from x86/Larrabee code than OpenCL or DirectCompute.
So it's going to be the same story as PhysX, just with Larrabee instead of Cuda.

Adobe will be moving over to OpenCL by CS6 so that will be a blow to Cuda.

Moving over, or adding support while keeping Cuda in place?
If I would hazard a guess, I would say that OpenCL would give less performance on nVidia GPUs than what they are currently getting.
Adobe's customers won't accept a performance hit from what they're currently getting with CS5, so I think the logical plan of action for Adobe is to keep Cuda support in place. This would make OpenCL a second-class citizen.

Microsoft is also using thier own Direct2Draw API for GPU hardware accelration and not Cuda.

Direct2Draw? I think you mean Direct2D, which is for graphics, not GPGPU, and as such is no competitor to Cuda.
DirectCompute is a competitor to Cuda (and OpenCL). However, MS structured their GPGPU-accelerated functionality so that every vendor is free to implement it with whatever API they prefer (just like how video acceleration has always worked in Windows).

 

[Janooo]

...
The last time AMD tried to push their own instructionset, it didn't take off...
...

Really, where did x64 come from?

That's not their own instructionset (as in: unique to AMD), it's shared with other x86-vendors.
It only worked because Intel supports it aswell. Case in point: Microsoft didn't release an x64 version of Windows until Intel had hardware in place.
In the case of 3DNow! it didn't take off because Intel ignored it. Intel is going to ignore whatever AMD might cook up with Fusion aswell.

The bold statement is false. AMD pushed x64 and Intel followed.
It just might happen again. If AMD creates and pushes some APU(GPU) instructions (or macros, or whatever) Intel might follow again. You never know.

 

[Scali]

The bold statement is false.

Not anymore, since I clarified that 'their own' meant 'unique to AMD' in this context.
When Intel adopts it, it's no longer unique to AMD.
And as I also said, x64 didn't take off UNTIL Intel adopted it, so in effect it was still Intel pushing x64 that made it a success, not AMD.

AMD pushed x64 and Intel followed.
It just might happen again. If AMD creates and pushes some APU(GPU) instructions (or macros, or whatever) Intel might follow again. You never know.

Unlikely, as Intel already has GPU extensions for x86. Codename LRBni (Larrabee New Instructions):
http://www.drdobbs.com/high-performance-computing/216402188

It's far more likely that AMD will adopt LRBni than the other way around...
Llano will not have any sharing of instructions at all, as it's just a K10-class CPU with a Radeon 5000-series GPU copy-pasted onto the die.

 

[grimpr]

Come on now...M$ made millions from AMD's 64bit set, Linux skyrocketed and it effectively destroyed Intels Itanium line and raison d'etre, its much more influential in the industry and you dont give enough credit where credit is due. You can bet that Intel hated it with much passion, at least in the beginning.

 

[Scali]

Come on now...M$ made millions from AMD's 64bit set, Linux skyrocketed and it effectively destroyed Intels Itanium line and raison d'etre, its much more influential in the industry and you dont give enough credit where credit is due. You can bet that Intel hated it with much passion, at least in the beginning.

Firstly, I still don't get why anyone needs to 'give credit' when just mentioning technology in a discussion that isn't about 'credit' in the first place.
Secondly, no matter how you try to spin it, x64 is just a 64-bit extension of Intel's x86. x64 cannot exist without x64... so no matter what, x86 will always be more influential than x64. x86 has pretty much destroyed all competitors long before AMD came up with its 64-bit extensions, and those 64-bit extensions cannot exist without x86 itself, so the influence of x86 continues in x64.
Lastly, just because one technology wins over another technology doesn't mean that the technology is also superior. This goes very much for x86 itself, and there are plenty of other examples (VHS vs Betamax being the archetypal example).
Now go cry somewhere else.

 

[Janooo]

Not anymore, since I clarified that 'their own' meant 'unique to AMD' in this context.
When Intel adopts it, it's no longer unique to AMD.
And as I also said, x64 didn't take off UNTIL Intel adopted it, so in effect it was still Intel pushing x64 that made it a success, not AMD.



Unlikely, as Intel already has GPU extensions for x86. Codename LRBni (Larrabee New Instructions):
http://www.drdobbs.com/high-performance-computing/216402188

It's far more likely that AMD will adopt LRBni than the other way around...
Llano will not have any sharing of instructions at all, as it's just a K10-class CPU with a Radeon 5000-series GPU copy-pasted onto the die.

Now you are twisting words.
So what older (prior to SSE4) x86 instructions are unique to Intel that AMD does not have? Why do you expect it from AMD?
The point is Intel followed AMD before it can happen again. We will see what the future brings...

 

[Scali]

Now you are twisting words.
So what older (prior to SSE4) x86 instructions are unique to Intel that AMD does not have? Why do you expect it from AMD?

x86 itself was unique to Intel until companies like AMD started building clones.
Most extensions that Intel introduced (such as 386, 486, Pentium, MMX, SSE) were also unique to Intel for at least a year, before one of their licensees introduced their first CPU with support for this technology.
Unlike 3DNow!, Intel was successful with these extensions before any of its competitors supported it.

And I'm not saying that I expect it from AMD, I'm saying that I think it is unlikely to happen. It has nothing to do with AMD itself, but rather the world they're operating in, and the lack of control they have over it.

 

[Janooo]

^
AMD/ATI has better APU(GPU). Intel might listen here.

 

[Scali]

^
AMD/ATI has better APU(GPU). Intel might listen here.

Unlikely, as the instructionset for a GPU is irrelevant.
nVidia and AMD use vastly different architectures and instructionsets, yet they both support all the same API standards... DX11, OpenGL 4.0, OpenCL.
This is because these are modern programming environments where everything is based on sourcecode or bytecode, and the driver can compile to native code at runtime.
Therefore there is no need to have instructionset-level compatibility between GPUs.
With x86 you cannot do that. Compatibility is at the binary level.

 

[Janooo]

Unlikely, as the instructionset for a GPU is irrelevant.
nVidia and AMD use vastly different architectures and instructionsets, yet they both support all the same API standards... DX11, OpenGL 4.0, OpenCL.
This is because these are modern programming environments where everything is based on sourcecode or bytecode, and the driver can compile to native code at runtime.
Therefore there is no need to have instructionset-level compatibility between GPUs.
With x86 you cannot do that. Compatibility is at the binary level.

What if AMD introduces new instructions that are going to be pushed to APU?
Intel can push the same instruction to Intel GPU. Nobody would care what happens in APU/GPU if the end results are correct.

 

[Scali]

What if AMD introduces new instructions that are going to be pushed to APU?
Intel can push the same instruction to Intel GPU. Nobody would care what happens in APU/GPU if the end results are correct.

As I already said, Intel already HAS GPU extensions for x86, AMD does not.
So it is unlikely at this point that Intel will drop their extensions and go with some theoretical future GPU extensions that AMD may come up with in a few years time.
Intel is just further ahead in the whole CPU+GPU integration game.
At this point we don't even know if x86-extensions for GPU ever become relevant. We could just stick to DirectCompute and OpenCL forever, and never go to a deeper level of integration.

 

[Gikaseixas]

That isn't the point I was making.
The point I'm making is that if AMD introduces an instructionset that Intel doesn't back, it is going to fail.
It isn't about credit. If you want, I can give credit to AMD for being the main drive behind 3DNow! aswell. Who cares about credit, when the technology is a failure anyway?
And why do you demand that people give credit to every technology they mention everytime?
If you want to play that game, it's about time to give nVidia credit for putting GPGPU solidly on the map with Cuda, and any developer will agree that C for Cuda was the 'inspiration' for the OpenCL API design. In other words, if it wasn't for Cuda, we wouldn't have OpenCL today.
Or how about giving Intel credit for inventing the microprocessor, and defining all x86 instructionsets and extensions, which AMD could base their x86-64 on. Happy with credit now?


People tend to see what they want to see.
As I said many times before, my current videocard is actually an AMD Radeon. So am I against AMD? Not really. I am just critical of them, as I am of any company.

I wasn't talking about 3DNow and you know that so don't pull that crap, i believe you're much more inteligent than that. x64 was being discussed and it is not a failure.

And yes, i give credit to Nvidia for having put GPGPU solidly on the map with Cuda and i also give credit to Intel for their success with x86 BUT i also give credit to AMD for x64, HyperTransport and keeping ATI competitve in the fiercely gpu business.

Question is, can you?? I bet not. All you see is Intel and Nvidia. Finally, you say you have a ATI gpu, so what??? Check my sig, i have Intel, Nvidia, AMD and ATI

 

[Scali]

I wasn't talking about 3DNow and you know that so don't pull that crap, i believe you're much more inteligent than that. x64 was being discussed and it is not a failure.

You were responding to something I said and I was not talking about x64.
I said this:
"The last time AMD tried to push their own instructionset, it didn't take off... 3DNow! was not a bad instructionset at all, but there just weren't enough CPUs on the market to get support from compilers and applications."
Oh and look, I actually gave AMD credit there, I said it wasn't a bad instructionset!
In fact, I actually used 3DNow! back in the day... if 3DNow! failed through lack of support, it surely wasn't because of me.
Found this old 3DNow! code of mine still online: http://srcvault.scali.eu.org/cgi-bin/Syntax/Syntax.cgi?sinsin.c

Question is, can you?? I bet not. All you see is Intel and Nvidia. Finally, you say you have a ATI gpu, so what??? Check my sig, i have Intel, Nvidia, AMD and ATI

I have a pile of old videocards and CPUs in my closet somewhere, brands which you probably never even heard of...
You know what the irony of it all is? The CPU architecture that I used to like most was the Motorola 68k. I've never liked x86 or Intel in the first place, it's always been poor technology compared to 68k. In a way it is an insult to me if you think I'm an Intel fan. I could never be a fan of x86 in any shape or form.

It just seems that a lot of people have no concept of where I'm coming from. They can only think in black and white, and I simply cannot explain who I am, what I think, and what I like and don't like, because you do not comprehend the shades of gray it encompasses.
I don't care if you understand me or not, but stop the accusations because it's getting really old really fast.

 

[waffleironhead]

Not anymore, since I clarified that 'their own' meant 'unique to AMD' in this context.
When Intel adopts it, it's no longer unique to AMD.
And as I also said, x64 didn't take off UNTIL Intel adopted it, so in effect it was still Intel pushing x64 that made it a success, not AMD.



Unlikely, as Intel already has GPU extensions for x86. Codename LRBni (Larrabee New Instructions):
http://www.drdobbs.com/high-performance-computing/216402188

It's far more likely that AMD will adopt LRBni than the other way around...
Llano will not have any sharing of instructions at all, as it's just a K10-class CPU with a Radeon 5000-series GPU copy-pasted onto the die.

Your logic makes no sense here. AMD was responsible for x64, it was successful and it was unique to AMD before the adoption. The terms of the success do not matter. You just moved the goalposts after someone proved you wrong. Pretty typical.

 

[Scali]

Your logic makes no sense here. AMD was responsible for x64, it was successful and it was unique to AMD before the adoption. The terms of the success do not matter. You just moved the goalposts after someone proved you wrong. Pretty typical.

I specifically stated 3DNow! in my original post:
"The last time AMD tried to push their own instructionset, it didn't take off... 3DNow! was not a bad instructionset at all, but there just weren't enough CPUs on the market to get support from compilers and applications."

It's not my fault that some people have this uncontrollable urge to try and squeeze x64 into the point I was making (so if anyone was moving goalposts, it wasn't me)... which means it is no longer the point I was making, as x64 and 3DNow! are fundamentally different in that Intel supports one, but not the other.
Can we please drop it now? I am getting tired of having to report all these AMD fanboys.

 

[Scali]

Well, I'd like to add one thing to the matter.
I was an early adopter of x64. I started using Windows XP x64 pretty much as soon as it was released, and ported my codebase over to it.
Want proof? Here it is:
http://bohemiq.scali.eu.org/forum/viewtopic.php?f=4&t=39
Back in April 2006 I had DirectX 9 code running on Windows XP x64.
So if anyone wants to tell ME about the success of x64, they had better been using an x64 OS and x64 OS before early 2006... but I guess that's too much to ask.
As I say, people just can't think in those shades of gray...

 

[waffleironhead]

I specifically stated 3DNow! in my original post:
"The last time AMD tried to push their own instructionset, it didn't take off... 3DNow! was not a bad instructionset at all, but there just weren't enough CPUs on the market to get support from compilers and applications."

It's not my fault that some people have this uncontrollable urge to try and squeeze x64 into the point I was making (so if anyone was moving goalposts, it wasn't me)... which means it is no longer the point I was making, as x64 and 3DNow! are fundamentally different in that Intel supports one, but not the other.
Can we please drop it now? I am getting tired of having to report all these AMD fanboys.

Thats the sign of success. The adoption by intel. thanks for admitting you were wrong.

 

[Gikaseixas]

You were responding to something I said and I was not talking about x64.
I said this:
"The last time AMD tried to push their own instructionset, it didn't take off... 3DNow! was not a bad instructionset at all, but there just weren't enough CPUs on the market to get support from compilers and applications."
Oh and look, I actually gave AMD credit there, I said it wasn't a bad instructionset!



I have a pile of old videocards and CPUs in my closet somewhere, brands which you probably never even heard of...
You know what the irony of it all is? The CPU architecture that I used to like most was the Motorola 68k. I've never liked x86 or Intel in the first place, it's always been poor technology compared to 68k. In a way it is an insult to me if you think I'm an Intel fan. I could never be a fan of x86 in any shape or form.

It just seems that a lot of people have no concept of where I'm coming from. They can only think in black and white, and I simply cannot explain who I am, what I think, and what I like and don't like, because you do not comprehend the shades of gray it encompasses.
I don't care if you understand me or not, but stop the accusations because it's getting really old really fast.

...brands which you probably never even heard of. Are you for real? Why do you think this is even relevant? Makes you feel better i guess eh.

68k was actually good, it had real potential. Motorola should have kept it alive.

 

[Scali]

...brands which you probably never even heard of. Are you for real? Why do you think this is even relevant?

You claim "All you see is Intel and Nvidia".
I say I have a closet full of hardware that proves that I see good technology from whatever company makes it.
So yes, it's relevant.

68k was actually good, it had real potential. Motorola should have kept it alive.

That's where you're wrong.
One of the main problems with x86 is the legacy.
Motorola was right to kill off 68k for regular desktops, and move to the more advanced (and legacy-free) PowerPC architecture.

 

[Gikaseixas]

I specifically stated 3DNow! in my original post:
"The last time AMD tried to push their own instructionset, it didn't take off... 3DNow! was not a bad instructionset at all, but there just weren't enough CPUs on the market to get support from compilers and applications."

It's not my fault that some people have this uncontrollable urge to try and squeeze x64 into the point I was making (so if anyone was moving goalposts, it wasn't me)... which means it is no longer the point I was making, as x64 and 3DNow! are fundamentally different in that Intel supports one, but not the other.
Can we please drop it now? I am getting tired of having to report all these AMD fanboys.

Don't acuse ppl of being AMD fanboys just because they're proving you WRONG. I actually prefer Intel right now and that's why my number one pc has a bad a$$ i7 in it. If AMD releases something better i might swicth camps.

We where talking about x64, stop being dilusional or trying to fool ppl. You brought 3DNow here.

 

[Scali]

Don't acuse ppl of being AMD fanboys just because they're proving you WRONG. I actually prefer Intel right now and that's why my number one pc has a bad a$$ i7 in it. If AMD releases something better i might swicth camps.

We where talking about x64, stop being dilusional or trying to fool ppl. You brought 3DNow here.

No we weren't. Read the thread again.
My remark on 3DNow! was BEFORE someone brought x64 into the mix, as a response.

 

[waffleironhead]

I specifically stated 3DNow! in my original post:
"The last time AMD tried to push their own instructionset, it didn't take off... 3DNow! was not a bad instructionset at all, but there just weren't enough CPUs on the market to get support from compilers and applications."

It's not my fault that some people have this uncontrollable urge to try and squeeze x64 into the point I was making (so if anyone was moving goalposts, it wasn't me)... which means it is no longer the point I was making, as x64 and 3DNow! are fundamentally different in that Intel supports one, but not the other.
Can we please drop it now? I am getting tired of having to report all these AMD fanboys.

Please quit being so defensive and accusing anyone who disagrees with you as being a fanboy. You're thinking too black and white.

 

[Gikaseixas]

You claim "All you see is Intel and Nvidia".
I say I have a closet full of hardware that proves that I see good technology from whatever company makes it.
So yes, it's relevant.



That's where you're wrong.
One of the main problems with x86 is the legacy.
Motorola was right to kill off 68k for regular desktops, and move to the more advanced (and legacy-free) PowerPC architecture.

Even with all the knowledge display you bring here when making your arguments I cannot help but feel that there is something emotionally disfunctional in your discussions.

I actually agreed with you that 68k was good but to your eyes somehow i managed to be wrong again. I just can't make you happy.

 

[Scali]

Please quit being so defensive and accusing anyone who disagrees with you as being a fanboy. You're thinking too black and white.

The only logical explanation of why there is so much emphasis on x64 and 'credit' in a thread that is about GPGPU, where I made a remark referring to 3DNow! to point out that Larrabee's GPU-extensions have a greater chance of becoming the de-facto standard for x86+GPU integration than whatever AMD might cook up in the future... is that there are AMD fanboys involved (trust me, it's not about disagreeing... as I proved, I adopted x64 at a very early stage).

Normal people wouldn't focus on off-topic subjects like that, page after page. And they wouldn't do it with so many personal attacks directed at me.
So don't go complaining when I call you out, you had it coming.

 

[Scali]

Even with all the knowledge display you bring here when making your arguments I cannot help but feel that there is something emotionally disfunctional in your discussions.

As I said, people see what they want to see.

I actually agreed with you that 68k was good but to your eyes somehow i managed to be wrong again. I just can't make you happy.

Then why don't you stop trying?
I don't think you'll ever understand me. Everytime you think you've got me figured out, I'm one step ahead again.