AMD Charts Path for Future of its GPU Architecture

[Benetanegia]

Still not sure how scalar has a performance advantage tbh, at a glance it should be weaker

It's something I'll need to research more.

CPUs are scalar (+ a vector unit) and GPGPU means running code that typically runs on CPU on the GPU, hence scalar is an advantage for a wider range of code.

Both future architectures from AMD and Nvidia are going to be scalar + vector. For AMD it's the arch in the OP. For Nvidia I'm not sure if it was kepler or Maxwell, but in any case by 2013 both companies will be there.

 

[Wile E]

That's one use, to me, and not one that I personally get any use out of. You falsely inflating the possibilities.

As a home user, there's 3D browser acceleration, encoding accelleration, and game physics. Is there more than that for a HOME user? Because that's what I am, right, so that's all I care about.

Which brings me to my point...why do I care? GPGPU doesn't offer me much.

Then AMD's use of APUs is just as useless. It operates on exactly the same principles.

 

[cadaveca]

Then AMD's use of APUs is just as useless. It operates on exactly the same principles.

Currently, for me, it is useless.

Until we get games that take advantage of what's offered, to me, APUs are nothing more than a XBOX360.

I mean, what does sandybridge on Z68 offer? It'll do the same acceleration that discrete cards can, but...that's it?

Unless it offers me a better gaming experience, I don't care.

They can't run CUDA, Intel's SandyBridge and AMD's APUs both, so I don't get any game benefits, such a Phys-X; so why would I be interested?

 

[Wile E]

Then why come in here spreading misinformation?

 

[cadaveca]

Then why come in here spreading misinformation?

It's not misinformation. It's my opinion. If CUDA was opened to those other platforms, then there might be reason to be interested, hence it hurting the consumer.

I mean, if there was a real APU with an nVidia GPU, that'd be great, but because alot of these chips are intended for desktops, and is you want better 3D performance than what an AMD APU or SB offers, the AMD APU's paired with an AMD GPU are going to be the very best option, performance wise.

But I can't get Phsy-x on that high-performance option...

We know AMD isn't going ot be there on the software side; it's up to the dev's to decide to implement the technologies, but at the same time, when it comes to gmaing, nV is going to be pushing thier options, and that doesn't help.

 

[Wile E]

It's not misinformation. It's my opinion. If CUDA was opened to those other platforms, then there might be reason to be interested, hence it hurting the consumer.

I mean, if there was a real APU with an nVidia GPU, that'd be great, but because alot of these chips are intended for desktops, and is you want better 3D performance than what an AMD APU or SB offers, the AMD APU's paired with an AMD GPU are going to be the very best option, performance wise.

But I can't get Phsy-x on that high-performance option...

It is misinformation when without CUDA, we would not have these new APUs, as there would be no interest in this form of computing. CUDA did not hurt consumers, period. It drove an entire market into being, which is producing new open standards. That is the very definition of good for consumers.

 

[cadaveca]

It drove an entire market into being, which is producing new open standards. That is the very definition of good for consumers.

OK, sure. But with AMD's GPUs following closer and closer to nV's solutions, it makes far less sense for nV to restrict thier software to thier chips alone.

THAT doesn't help anyone, but them.

 

[Wile E]

OK, sure. But with AMD's GPUs following closer and closer to nV's solutions, it makes far less sense for nV to restrict thier software to thier chips alone.

THAT doesn't help anyone, but them.

They are also developing for the open standards, so your point is irrelevant. Sure, it helps only them, but it doesn't hurt anyone in the process.

 

[cadaveca]

so your point is irrelevant

Thanks for your opinion.

 

[Wile E]

Thanks for your opinion.

No, it's a fact. CUDA benefiting nV is completely irrelevant to the topic at hand. The topic at hand is whether or not it hurts consumers. Benefiting nV does not automatically equal harming consumers. It does not, becasue nV still fully supports the open standards as well.

 

[cadaveca]

The topic at hand is whether or not it hurts consumers

No, actually, the topic is what AMD is doing with thier GPUs. I'd like to see them run CUDA, but it's not gonna happen. CUDA sucks.

 

[Wile E]

No, actually, the topic is what AMD is doing with thier GPUs. I'd like to see them run CUDA, but it's not gonna happen. CUDA sucks.

At least it sucks less than Stream.

 

[Damn_Smooth]

Currently, for me, it is useless.

Until we get games that take advantage of what's offered, to me, APUs are nothing more than a XBOX360.

I mean, what does sandybridge on Z68 offer? It'll do the same acceleration that discrete cards can, but...that's it?

Unless it offers me a better gaming experience, I don't care.

They can't run CUDA, Intel's SandyBridge and AMD's APUs both, so I don't get any game benefits, such a Phys-X; so why would I be interested?

Would this be an implementation that you would consider useful?

http://blogs.msdn.com/b/somasegar/archive/2011/06/15/targeting-heterogeneity-with-c-amp-and-ppl.aspx

I’m excited to announce that we are introducing a new technology that helps C++ developers use the GPU for parallel programming. Today at the AMD Fusion Developer Summit, we announced C++ Accelerated Massive Parallelism (C++ AMP). Additionally, I’m happy to say that we intend to make the C++ AMP specification an open specification.

And, forgive my ignorance here, but wouldn't this also render Cuda somewhat obsolete?

 

[Wile E]

Would this be an implementation that you would consider useful?

http://blogs.msdn.com/b/somasegar/archive/2011/06/15/targeting-heterogeneity-with-c-amp-and-ppl.aspx



And, forgive my ignorance here, but wouldn't this also render Cuda somewhat obsolete?

That depends solely on what the API is capable of. CUDA as we know it right now will eventually die off for a more hardware independent approach.

 

[Damn_Smooth]

That depends solely on what the API is capable of. CUDA as we know it right now will eventually die off for a more hardware independent approach.

I really don't know enough to pretend that I know what I'm talking about yet, so thank's for explaining that to me.