Is this the end for AMD?

[Cerb]

I wonder if it is too late for AMD (or anyone buying AMD) to do their own proprietary GPGPU API/Implementation for the Radeon GPU?

Is Nvidia reaching a saturation point in their CUDA development? (Thus allowing AMD to close the gap while they are on the fast learning curve)

TBH, there isn't much need to add another one. CUDA will slowly die off, should AMP become accepted across platforms, which is likely to happen, IMO. While nVidia certainly took advantage of having something nobody else could run on GPUs, CUDA is significantly more capable than anything but AMP, right now, and AMP is, of course, brand new.

AMD has lagged in software support. There's no lack of talent to hire for that kind of thing, and it shouldn't take large teams of programmers, either. AMD's management did not put a focus on making the technology they had sing and dance. nVidia made theirs do Broadway shows, even during times the hardware was largely inferior. This is not to say NVidia has been entirely fair. They are a cunning business. But, they got the job done, while AMD just kept talking about how great it would be to do the job.

 

[cbn]

TBH, there isn't much need to add another one. CUDA will slowly die off, should AMP become accepted across platforms, which is likely to happen, IMO. While nVidia certainly took advantage of having something nobody else could run on GPUs, CUDA is significantly more capable than anything but AMP, right now, and AMP is, of course, brand new.

AMD has lagged in software support. There's no lack of talent to hire for that kind of thing, and it shouldn't take large teams of programmers, either. AMD's management did not put a focus on making the technology they had sing and dance. nVidia made theirs do Broadway shows, even during times the hardware was largely inferior. This is not to say NVidia has not been entirely fair. They are a cunning business. But, they got the job done, while AMD just kept talking about how great it would be to do the job.

Thank you very much for letting me know about MS AMP.

 

[dguy6789]

Just want to throw in that Llano blows away anything Intel has by such a margin that I wouldn't say Intel is capable of competing in that segment at all right now. Llano would not be possible if AMD did not buy ATI.

 

[frozentundra123456]

Just want to throw in that Llano blows away anything Intel has by such a margin that I wouldn't say Intel is capable of competing in that segment at all right now. Llano would not be possible if AMD did not buy ATI.

I would say this is a gross overstatement. Bobcat is superior to atom, but Llano is a trade off at best. Better graphics for an outdated CPU. It might make sense in the mobile space, but in the desktop, it has a very limited appeal.

And I know the next agrument, Well AMDs cpu performance is "good enough". Maybe, but especially on the desktop, just get an Intel cpu and add a discrete 50 dollar card to have both superior CPU and graphics performance. I just dont feel that Llano's graphics performance is yet good enough to compensate for the CPU deficit. The integrated gpu on Sandy bridge is probably "good enough" for almost anything except gaming. Llano is far superior, but would you really want to game on Llano??

 

[exar333]

Just want to throw in that Llano blows away anything Intel has by such a margin that I wouldn't say Intel is capable of competing in that segment at all right now. Llano would not be possible if AMD did not buy ATI.

CPU performance = Intel
GPU performance = AMD
Power consumption = Intel

How exactly does Llano 'blow away' anything from Intel? Its a solid product, but saying anything more is really ignorant. If you just need a HTPC, its good, but still worse than a cheaper CPU + cheap discrete. if you game, Llano is still too slow. Its power-hungry and CPU performance is lagging for the price.

Edit: A cheap MB + SB-celeron is plenty capable. if you need more power, add a cheap discrete GPU for $30-40. It always can take a quad-core i5 as well, something Llano cannot do. Don't get me wrong, Llano is OK, but not amazing by any means.

 

[Arachnotronic]

Ladies and gents, brace yourselves for the "end of an era" tomorrow. If AMD goes all-in into smartphone/tablets on us, the AMD as "we know it" will be no more

I just hope Rory understands the monumental mountain in front of him of taking on not just Intel but the Qualcomms, Samsungs, TIs of this world, and well ultimately a competitor even more vicious than Intel -> Apple!

He does. They're not going to kill off successful parts of their business. If it makes money and if it's broadly applicable to their whole product lineup (*cough* EFFICIENT GPUS *cough*), then they will make it.

Rory Read is probably not stupid. He ran Lenovo, a PC maker, and he ran it just fine in the face of tablet/smartphones. I am confident enough, in fact, of this that I've piled on some more AMD shares pre-Project WIN.

Relax.

 

[notty22]

AMD dropped hints this past Summer of a possible gpu strategy shift.
AMD Considers Dropping Low-end Discrete Graphics

 

[RussianSensation]

He does. They're not going to kill off successful parts of their business. If it makes money and if it's broadly applicable to their whole product lineup (*cough* EFFICIENT GPUS *cough*), then they will make it.

Rory Read is probably not stupid. He ran Lenovo, a PC maker, and he ran it just fine in the face of tablet/smartphones. I am confident enough, in fact, of this that I've piled on some more AMD shares pre-Project WIN.

Relax.

hhehe I am relaxed. I will be very happy to be wrong about AMD leaving the high-end GPU and CPU business. ^_^ I don't really care if AMD goes into smartphones or tablets as long as they continue to make good CPUs/GPUs for desktops/laptops. Personally, as long as AMD doesn't take $ away from CPUs and GPUs, I could care less if they make the best smartphone/tablet SoC since I have little interest in that market segment.

Anand provided an update, explaining that Project WIN is just an internal plan: http://www.anandtech.com/show/5079/amds-project-win-a-misunderstanding

 

[grimpr]

Come on now, AMD ditching the entire GPU lineup with strong cores, strong ATI engineers and talent, the painfully slow buildup of the recently succesfull FirePro line and the companys heritage and ties to the console industry? This is pure madness!

Personally i can forsee a withdraw from the PC Enthusiast crowd which always had been an Intel stronghold.

 

[Arachnotronic]

The "PC Enthusiast" crowd has many a definition. If you mean the guys who drop >$600 on a CPU, >$300 on a mobo, and dual/triple/quad high end GPUs? Then yes, they will be out of the CPU component of that race. But that's so incredibly a minority that it doesn't matter.

How many of you have a i7 970 or better?

 

[Vesku]

CPU performance = Intel
GPU performance = AMD
Power consumption = Intel

How exactly does Llano 'blow away' anything from Intel? Its a solid product, but saying anything more is really ignorant. If you just need a HTPC, its good, but still worse than a cheaper CPU + cheap discrete. if you game, Llano is still too slow. Its power-hungry and CPU performance is lagging for the price.

Edit: A cheap MB + SB-celeron is plenty capable. if you need more power, add a cheap discrete GPU for $30-40. It always can take a quad-core i5 as well, something Llano cannot do. Don't get me wrong, Llano is OK, but not amazing by any means.

http://techreport.com/articles.x/21902/8

AMD is now competitive in the notebook space. Sales seem to indicate an appreciation for a good IGP.

 

[Olikan]

Edit: A cheap MB + SB-celeron is plenty capable. if you need more power, add a cheap discrete GPU for $30-40. It always can take a quad-core i5 as well, something Llano cannot do. Don't get me wrong, Llano is OK, but not amazing by any means.

llano sucks for desktops, more than bulldozer imho.

but llano is very strong in laptops,
nothing match it's batery time while gaming, and is the best perf/$ in it's price tag.
Actually it will always be the best perf/$ until intel catch up with it's igps.

 

[SparkyJJO]

Hate it when I misread the poll and voted the wrong way /facepalm

Anyway, these doomsday threads get annoying.

 

[cbn]

TBH, there isn't much need to add another one. CUDA will slowly die off, should AMP become accepted across platforms, which is likely to happen, IMO. While nVidia certainly took advantage of having something nobody else could run on GPUs, CUDA is significantly more capable than anything but AMP, right now, and AMP is, of course, brand new.

What about MS AMP for SuperComputers? (eg, Department of Energy "Titan" here)

Is there any chance we could see that happen? Or would AMD still need to develop their own CUDA equivalent for this?

 

[Cerb]

What about MS AMP for SuperComputers? (eg, Department of Energy "Titan" here)

Is there any chance we could see that happen? Or would AMD still need to develop their own CUDA equivalent for this?

Right now, AMD GPUs suited for such work are still in the future. Tesla has ECC. Some RAS we take for granted (storage data ECC, filesystem CRCs, CRCs on most IO, etc.). Many RAS features useful for mainframes and minicomputers of old are easy to design around. ECC RAM and caches, and MCA, however, are best done by the hardware*.

Beyond that, we really won't know until AMP gets used out in the wild for awhile on Windows, whether it can be suitable for that kind of work or not. Then, after that, it will still need plenty of software support to be built up for *n*x environments, which will be no small task. AMP will be anything but a quick fix, even assuming it does get adopted widely outside of Windows. The thing going for it is that, IMO, and obviously others' opinions, too, it is a huge step in the right direction.

* with an abstraction layer above the processor that sufficiently decouples the processor from the code it is going to execute, MCA could theoretically be implemented without the programmer having to worry about it, beyond reporting of uncorrectable RAM errors. I doubt this has escaped NVidia's attention.

 

[nickb64]

I think they'll probably be in about the same position as they have been the last few years, relevant to the very price conscious who want to play games. I don't think Intel will allow them to fail, because they need AMD to at least appear to matter to avoid the Feds coming in and slapping them with antitrust stuff.

As others have said, not everything revolves around games.

 

[cbn]

http://www.anandtech.com/show/4455/amds-graphics-core-next-preview-amd-architects-for-compute/7

I was very happy to read this:

Finally, while we’ve talked about the GCN architecture in great detail we haven’t talked about how to program it. Of course there’s OpenCL, but with GCN there’s going to be so much more. Next week we will be taking a look at AMD’s Fusion System Architecture, a high-level abstraction layer that will make GPU programming even more CPU-like, an advancement necessary to bring forth the kind of heterogeneous computing AMD is shooting for. We will also be taking a look at Microsoft’s C++ Accelerated Massive Parallelism (AMP), a C++ extension to bridge the gap between current and future architectures by allowing developers to program for GPUs in C++ even if the GPU doesn’t fully support the C++ feature set.