AMD Fusion Info

[Rhoxed]

http://www.tgdaily.com/html_tm...nt-view-38703-135.html

 

[SlowSpyder]

http://www.tgdaily.com/html_tm...nt-view-38703-135.html

Is that the link?

 

[Rhoxed]

that would be it hah

 

[Arkaign]

Fail.

 

[Acanthus]

Not really fantastic news overall...

TSMC is generally behind both Intel and AMD in process technology.

Its still a long way off, and is only dual core...

By 2010 we should be looking at lots of multithreaded software, similar to the dual core vs quad core situation now... but its going to be the "n cores vs n*2 cores" argument ad nauseum.

 

[Phynaz]

Does TSMC have an x86 license?

This would also imply Phenom has been redesigned to be fabbed on bulk silicon, something I find difficult to beleive.

 

[Lonyo]

Since it says dual core Phenom, you can be fairly sure that if the report is true, it's been at least marginally redesigned.
Also they are probably aiming more for power efficiency than performance, so again it would make sense that there would be an element of chip redesign in there. Optimising for power efficiency, dropping down to dual core and making it work on a 40nm process all seem like sensible steps, and if you're doing two, why not go for #3 as well?

Also I think that AMD have the right to outsource some of its x86 chips to 3rd party foundries, which would mean that TSMC can fab these chips based on that, as long as AMD doesn't make too many.

 

[bunnyfubbles]

Originally posted by: Acanthus
Not really fantastic news overall...

TSMC is generally behind both Intel and AMD in process technology.

Its still a long way off, and is only dual core...

By 2010 we should be looking at lots of multithreaded software, similar to the dual core vs quad core situation now... but its going to be the "n cores vs n*2 cores" argument ad nauseum.

I don't think Fusion was ever supposed to be an enthusiast targeted product, yet some reason that's how too many have been treating it. It really just sounds like a holy grail for general computing as it truly is something of a 'system on a chip'. Most users really don't need massive processing power, and dual cores (perhaps with some GPU assistance) should be more than enough for the masses, even in 2010.

 

[quadomatic]

Maybe it'll be good for laptop gaming though. If it is, hopefully it's not too expensive, and I'll then be able to get a laptop that uses it so I can do some decent gaming.

 

[Lonyo]

Originally posted by: bunnyfubbles

Originally posted by: Acanthus
Not really fantastic news overall...

TSMC is generally behind both Intel and AMD in process technology.

Its still a long way off, and is only dual core...

By 2010 we should be looking at lots of multithreaded software, similar to the dual core vs quad core situation now... but its going to be the "n cores vs n*2 cores" argument ad nauseum.

I don't think Fusion was ever supposed to be an enthusiast targeted product, yet some reason that's how too many have been treating it. It really just sounds like a holy grail for general computing as it truly is something of a 'system on a chip'. Most users really don't need massive processing power, and dual cores (perhaps with some GPU assistance) should be more than enough for the masses, even in 2010.

Indeed.
You can get a laptop with a 30w CPU + northbridge/GPU at another 20w, or get a single chip (Fusion) which is 40w, and also costs less as a single chip rather than 2 individual ones (both in chip costs and motherboard costs), and also potentially allows more compact motherboards.
Cheaper, lower power machine with good enough performance and potentially more layout flexibility.
(Numbers are random and pulled from nowhere, but serve to show the general theme of the product).

 

[Zstream]

I will bet the GPU will equal somewhere around the 3850 in terms of speed.

 

[taltamir]

everyone here is saying that 2010 is too late... the article said that fusion will arrive @ 40nm "soon" and will be shrunk to 32nm during 2010.

 

[CTho9305]

Originally posted by: Acanthus
Its still a long way off, and is only dual core...
By 2010 we should be looking at lts of multithreaded software, similar to the dual core vs quad core situation now... but its going to be the "n cores vs n*2 cores" argument ad nauseum.

What makes you think we'll have lots of multithreaded apps in 1.5 yrs? I think we already have seen many of the easy things threaded, and a lot of the remainder is either inherently single threaded or is very difficult to thread well. 1.5 years isn't all that long for larger projects (look at the rate of change in Firefox or MS Windows).

 

[quadomatic]

I read that the Fusion was supposed to hit in 2nd half of 2009, not 2010...

 

[soccerballtux]

SOI on 32nm? Poor amd