The realization of AMD FUSION......Bulldozer

[drescherjm]

The problem with on-chip, on-die, or on-board graphics is memory bandwidth. While they can probably get things playable for the 1280x1024 crowd, they are going to have to put some decent amounts of GDDR3 or GDDR5 on the board to get real performance. System memory just does not have the necessary bandwidth for gaming in its current form. Either that, or come up with a GDDR3 or GDDR5 slot so the user could choose their graphics memory.

Bulldozer will be quad channel DDR3. Although even that will not be as much bandwidth as high end GPUs and it will have to share that bandwith with the cores and the rest of the system.

 

[jimmyb]

The problem with on-chip, on-die, or on-board graphics is memory bandwidth.

This isn't so much a problem with the graphics processors being on-chip, so much as it is the products aren't targeted to the high end market, and subsequently aren't typically paired with dedicated high-bandwidth memory.

From a design perspective, on-die graphics processing offers potentially better latencies and bandwidth.

 

[Digital Viper-X-]

I don't think AMD fusion is as simple as Intels implementation, at least not the vision of where fusion is headed. The purpose of the GPU isn't for gaming / display

currently, the cpu is comprised of integer and floating point units, so the depending on the layout, the cpu might be integer strong(intel) or floating point strong, or both. The idea is to leverage the GPU part of the equation, to handle the FPU loads better, currently programs don't do that, but with opencl / direct compute, eventually software will catch up, and you'll be able to have a strong Int performance, without sacrificing Floating point performance by offloading FP work to the GPU.

I'd say it will take 3 years or so for this to fully catch on(in this implementation, middle ware like physx would be come obsolete)


http://www.anandtech.com/show/2872

Doubling the integer resources but not the FP resources works even better when you look at AMD’s whole motivation behind Fusion. Much heavy FP work is expected to be moved to the GPU anyway, there’s little sense in duplicating FP hardware on the Bulldozer core when it will eventually have a fully capable GPU sitting on the same piece of silicon. While the first incarnation of Bulldozer, the Zambezi CPU, won't have an on-die GPU, presumably future APUs will use the new core. In those designs the Bulldozer cores and the GPU will most likely even share the L3 cache. It’s really a very elegant design and the basis for what AMD, Intel and NVIDIA have been talking about for years now. The CPU will do what it does best while the GPU does what it is good at.

another good article.
http://www.anandtech.com/show/2881

 

[W.Feather]

i agree, the Fusion(Bulldozer) from AMD is undoubtedly more complicated, which is why I think there is a chance it will keep pace with the next generation Intel chips

 

[Digital Viper-X-]

i agree, the Fusion(Bulldozer) from AMD is undoubtedly more complicated, which is why I think there is a chance it will keep pace with the next generation Intel chips

it sure is going to be exciting, this is the reason that bulldozer is FP weak and int heavy, even though BD will first come out without fusion(the first fusion cpu will use a phenom 2 core)