While on the topic of AMD - although probably just knee jerk reaction to AMD moving some key people from the GPU side over to the CPU side, I hear from a previously reliable source that AMD is looking to sell off the GPU business for some quick cash (this is probably just rumour) but I'm keeping my ear to the ground on this one.
I assume this is the post you are talking about:
http://hardforum.com/showpost.php?p=1037999552&postcount=157
He even made the following statement too:
"BD doesn't have mainframe RAS features or the same scalability which is expected for the market Itanium primarily now targets as a competitor to IBM's POWER and Oracle's SPARC. And Intel doesn't need BD for the commodity server market since it already has Xeon, which is both a smaller chip and also has higher performance (plus it lacks GF's manufacturing fleas... brace for January when GF's pricing for AMD goes back to wafers instead of "good die" on a poor 32nm process)."
He talks about scalability and yet ORNL is using 38000+ Interlagos CPUs in Titan:
http://www.extremetech.com/extreme/...400-processor-20-petaflop-successor-to-jaguar
It is going to be the most powerful supercomputer in the world. What's his idea of scaleability?? 50000 CPUs?? 100000 CPUs??
I have seen claims by that chap on forums. There are many people like him who make a million claims,of which a few will be true;however there are loads of claims they make which are nonsense and are not accurate. Will they EVER talk about these?? NO.
Loads of people knew that the fab business was loosing AMD loads of money. However,the X86 agreement with Intel meant they could not become fabless. With Intel releasing them from this requirement it was no wonder they went fabless.
That chap is making it sound he was the only chap on the internetz who said selling the fabs made sense
financially.
I even said the same thing at around the same time on another forum!! +1 internetz to me too!!
The APUs use the same GPU architectures as their discrete cousins. If you sell off the GPU division then what happens to GPU architecture R and D for the CPUs?? A decent amount of these costs are met by consumer and professional GPU sales.
So basically what that chap is saying is one of the following:
1.)AMD only makes X86 server CPUs. However,according to that chap they won't scale so its a fail already!! On top of this they are going to become MEGA EXPENSIVE in January!!
2.)AMD ditches the lower end and midrange X86 market. They will have no hope trying to compete with Intel in this market with no IGPs. OEMs want lower costs. All of the ARM CPUs running Windows 8 will also have IGPs. Fail already!! Of course remember that all the FX CPUs from January are going to become MEGA EXPENSIVE on top of this!!
3.)They ditch X86 altogether for non-server CPUs and go ARM. They end up being a licensee. However with no GPU people left this would fail too. Fail already!!
4.)AMD just licenses a whole design including GPU and just rebrands the SOC. So AMD becomes a marketing and support company! Riiiight!
5.)They use their existing cores and license a GPU from another company. Nvidia maybe??
Of course it should be VERY easy and cheap for them too. A very reliable strategy!
6.)AMD sells the graphics division and licenses its former designs??Another company buys the graphics division and continues R and D. However,wouldn't that actually make things more complicated??
AMD needs a GPU division if it is to compete with any of the current CPU companies.
Its the GPU business which is keeping AMD competitive and if anything it will help them even more for mobile products. Also,don't you think that if AMD is going to focus more on their fusion
CPUs they would be having more GPU people working on them?? If they have no GPU people any more than how are they going to make competitive products?? Intel is investing more in IGPs and so are the various companies designing new ARM based GPUs. These types of CPUs are going to make up the bulk of all CPUs sold to consumers. CPUs without GPUs might be only the preserve of the higher market for the time-being.
AMD is also working on GCN so it can have a better chance to break into the HPC market too. If anything AMD might end up putting less focus on the
higher end consumer GPU market and moving towards the midrange and below. This is where the bulk of sales will be anyway.