GlobalFoundries Increases AMD 32nm Shipments by 80%
GlobalFoundries has been chastised by AMD's new CEO Rory Read for the production problems that limited the availability of 32 nm processors.
It appears that the foundry is now making progress as it stated in a press release that it has increased AMD 32 nm processor shipments by 80 percent between Q3 and Q4 of last year. According to GlobalFoundries, 32 nm chips now represent a third of AMD's product mix.
After the hiccup, the foundry is apparently quite interested in further expanding its business. The company said that it is planning on spending about $3 billion to fund an expansion of its facilities in Singapore, Germany and New York. Among the production successes are the manufacturing of chips for IBM at its Fab 8 in New York as well as the production of 28 nm ARM Cortex-A9 chips that run at up to 2.5 GHz.
Last week, rumors emerged that GlobalFoundries may be in talks of acquiring DRAM maker Promos, which would give the company production capability in Taiwan and the ability to cater to the Chinese market.
Discrete graphics. Nuff' said.
"You have quality? We have the quantity!"
Discrete graphics. Nuff' said.
lets say a 100% used wafer is 100% (realistically parts of the wafer cant be used, but lets assume that 100% of it is usable for easier math)
lets also assume that you can not go passed 100%
for the ability for them to claim 80% better, means that the yields on chips prior to that was at the very best 55%
i wish i knew what the average yield per wafer was so i can figure out how bad gf was doing prior to this, so lets get some assumptions.
if normal yeild is = what it was prior to 80% boost at best
95 = 52%
90 = 50%
85 = 47%
80 = 44%
75 = 41%
Actually, you're wrong in a major way, and in a minor way. Intel does not do pure shrinks anymore, and has not for a while. You will see some minor performance improvements. You also left out USB 3.0, which now does not need extra chips added.
So, we should get better clock speed, lower power, PCIE 3.0, USB 3.0, better graphics, all while being cheaper for Intel to make, and cheaper for motherboard makers in the not so long term.
It's not earth shattering, but it's a solid improvement from a position of already staggering superiority.
average yields on a new process is ~25%. Considering AMD was reporting low yields on llano, llano would have been below 20% yield. Bulldozer would probably be around 15-20% yields.
80% better would put yield up to above 30% probably which is still not optimal.
The key point about IB is the better power efficiency which is driving Intel's strategy far into the future as the number of transistors increase, heat must be managed to achieve the performance increases we all want.