AMD Fusion may slip into the next decade

Intel has had a graphics department for years and far more money for R&D, did you really expect them to fall behind in this?

perhaps AMD made that delay because they are developing something cool which they havent still announced, like a socketed GPU. I am a Fusion fanboy and i read some interviews with Phil and others. Some say late 2008/early 2009 while some say late 2009\early 2010

Intel has had a graphics department for years and far more money for R&D, did you really expect them to fall behind in this?

intel will just glue a gpu to a cpu if they fall behind....

If only money is the determining factor, how can a G965 fall behind RS480 and Crush-51? :wink:

intel's mistake in the G965 was focusing on shader ops other than pixel fillrate. IMO the reason for delay is that AMD couldnt start serious Fusion design until 2007, it would take two years to design, and one year to prototype, test and debug.

If intel gets into the bpu/GPU chip like AMD does right now it doesn't look like they will stand a chance. Intel has there own grphic department but they are not in the same league as Nvidia and AMD/ATI. So with the way things look right now AMD fusion chip would be the clear winner for this. but this is 2 to 3 years down the road mind you

Intel is working on a new graphics chip. Project name is rumored :wink: to be Larabee.

I think the only two things that hold Intel back from pushing a lot harder into the GPU market is antitrust crap and profit margins. Intel just makes a lot more money for CPU's than they do chipsets, and card makers have even lower margins in most cases (except possibly at the very high end).

Intel has had a graphics department for years and far more money for R&D, did you really expect them to fall behind in this?

intel will just glue a gpu to a cpu if they fall behind....


ROFLMAO

You mean like AMD "glued" 2 mobos together to come up with a quad core aka QFX??? :tongue:

Intel has had a graphics department for years and far more money for R&D, did you really expect them to fall behind in this?

intel will just glue a gpu to a cpu if they fall behind....


That's a good one, what do you call "Quad Garbage FX?"

Anyway, I never took the AMD's Fusion seriously.

Intel has had a graphics department for years and far more money for R&D, did you really expect them to fall behind in this?

intel will just glue a gpu to a cpu if they fall behind....

that made me LOL xD

Intel has had a graphics department for years and far more money for R&D, did you really expect them to fall behind in this?

intel will just glue a gpu to a cpu if they fall behind....

Fusion will be exactly the same thing.

Everything I've heard about Fusion has been "heterogenious multi core", did I miss something important? I suppose they could just put them both on one chip for the first generation or something, but it doesn't seem like their style. They don't have a lot of production capacity to mess around with glueing things together I thought.

dammit why does amd keep slipping behind in my eyes? i want to see them come out with something amazing (not that i'm a fanboy or anything), but in this intel-dominated market where c2d is lord, jeeze, amd needs a major silver bullet.

i'm hoping that they pull something out of their asses, something not on their "roadmap" or w/e, they're just like, "here, have this $50 16-core proc."

unfortunately i don't really think that'll happen.

It seems like a much better idea than an 80core CPU to me.


I haven't heard anything about nVidia making CPUs. I would think VIA is probably closer to being able to do this than nVidia. Both companies make IGPs, but Via also makes CPUs.

I hope that Fusion does come though, and doesn't sux. If I'm going to be forced to buy multi-core I want my extra cores to actually do something!

Eh, if you can manage to ignore the intel fanboys long enough you can see that there are still a few niches left to AMD. They are few and slim though   Lack of choice makes me uncomfortable also.

yeah, i was thinking of going 939 b/c its so frikkin cheap, and i'm a bottom-feeder of tech, but i am worried about amd's flagship business 'n stuff. if their top line fails, then eventually their bottom line/old stuff will too. i mean i'm not real worried, i just wanna see that silver bullet....!

Actually I believe Larabee is the company name. There was a story about it on The Reg right when the DOJ subpoenaed nVidia and ATi.

Vara is with Intel R&D.

It would be rather myopic to believe that in a multi-core world, all cores would always be carbon copies of one another. There could be cores that rock at integer math, cores for floating point, cores for streaming instructions, cores for graphics processing. If you really think about it, the only reason GPU's ever came into being was because general purpose CPU's couldn't handle the graphics processing workload. I could easily imagine a day when the GPU is just a subcomponent of the CPU, not something you have to go buy separately.

Meh, who cares really if it's Sis, S3 or Matrox, or *gasp* Transmeta.
Its not like you guys are gonna step away from discrete graphics solutions, or are you?

I just can't picture this: How do you fuse a GPU and a CPU and hit the right spot?
There's people who need high-powered CPUs but do fine with an x300 kinda card because they don't game. There's the ones who need both, and the ones that just surf the net. And how big is the product variety going to be?
"Yeah..I'll have a C8D F9800 and radeon z2500xtx combo please"...

LOL

Do they come in SLi?

Having the CPU next to the GPU also helps increasing bandwidth. This increase in badnwidth can help developers make more realistic games.

http://www.graphicshardware.org/presentations/pharr-key...

this tells you why CPU GPU integration is necessary.

AMD has been falling behind on most promises, not surprising here. The process technology needs to produce a high performance Fusion chip and keep yields high require 45 nm or even 32 nm circuitry due to the monstrous size of such a chip. Let AMD work 65 nm   and then let them talk about Fusion. We are in 2007 and AMD 65 nm is paper lunch still. You think AMD has the bandwidth to make fusion happen before 2010? I am not talking prototype, I am talking volume production.

Fusion can happen only if:
1. IBM produces a working 45 nm by 2008. IBM 65 nm is AWOL. We know AMD process R&D is a joke. AMD is on its path to become fabless
2. Immersion lithography is much improved and production worthy at high yields
3. AMD makes money on 65 nm investment, again where is 65 nm Hector?
4. Forget all the above, AMD will work with a top foundry to make Fusion. Even then, Fusion will be delegated to entry level desktop and notebook PC's.

AMD process appears to be less advanced than IBM's process. 65nm Cell can reach 5 and more GHz while AMD is stuck at 3GHz.
Based on IBM's problems with Cell yields one can assume that IBM process is more advanced delivering higher performance transistors while AMD process is focusing on higher yields in expanse of performance.
Unless, this is only a smoke and mirrors and K8L will clock at 5GHz 8)
and totally kill Intel.

is AMD K10 still under developement? OR are they going with an APU next?

I don't know I'd have hazarded a guess that K8L is a stopgap until Fusion comes along; for me it would make sense to have an across the board new architecture to complement Fusion and Torrenza. If its going to be 2010 until we see this, I think AMD has to do something new inbetween - they said there would be future IPC improvements next year, but that sounds more like K8Lv2 rather than a whole new architecture.

actually Intel is doing very well in every segment they are competing.

key points being which markets they have chosen to compete in, Intel likely doesen't compete in the middle to high end segments for costs reasons, while the profits are fat and abundant atm being divided by ATI and Nvidia if Intel were to arrive it would unbalance the segment entirely their would be a clear loser requiring aggressive pricing tactics which would kill the class for all involved.

that's not to say the consumer wouldn't be the benefactor but why bother if the biggest accomplishment from entering the class would be a price war ruining the reasons you decided to chase after.

as a side note with Vista's 3d hungry interface one would have to assume Intel has something in the works lest they sacrifice their graphics division entirely.... my guess is they have something in the works.

That is interesting. Why isn't intel talking about it then? I got bombarded with news from IDF not so long ago, no mention of heterogeneous multi-core CPUs... Intel has been talking quietly about doing system-on-chip which is kinda similar but no real details have been given other than "it's a system on a chip" (and by "chip" they seem to mean: very small PCB).

I didn't say "AMD's heterogeneous cores are interesting but Intel's are not". Heterogeneous cores are interesting. AMD is the one that has them on their roadmap. Intel is merely rumored to have one in R&D and has some whitepapers that talk about the concept but hasn't announced any plans to actually do it? Did I just miss it? I took a few minutes to search just now but didn't find...



So sometime after 2007 but before 2012 to 2015 we "might" see heterogeneous quad-core processesors from intel. This puts them on about the same track as AMD (assuming this includes GPU cores, I don't see why it wouldn't but again: they didn't say). But no dates given, nothing on the roadmaps, and this is side-note comment in an after-the-fact interview concerning the announcement of 80core processors that was a big announcement at IDF. Is it supposed to be a secret? Other than your investment portfolio is there something that makes you think they are actually planning on doing this in the same timeframe as AMD? Because, if they haven't announced plans to do anything with it, that pretty much answers your question as to why AMD's "Fusion" is more interesting then Intels "_______" but my point really was that any heterogeneous multicore processor was interesting so if you have info plz do share  

mmmm, frosted PCB *drool*

as i suspected the info was wrong, it has been changed

http://www.xbitlabs.com/news/cpu/display/20070116234458...

thank goodness.