Globalfoundries files plans for second mammoth plant in Malta (450 mm)

[SiliconWars]

http://www.timesunion.com/business/article/Fab-plan-filed-not-set-4243843.php

At 475,000 square feet, the clean room alone will be the size of nearly 10 football fields, more than 50 percent larger than the clean room in the existing factory, or "fab,"

Plans for Fab 8.2 come just weeks after GlobalFoundries announced that it would also be building a $2 billion technology and development center on the Malta site.

In other GF news, a Youtube video from DAC 2013 has some interesting stuff - http://www.youtube.com/watch?v=0cv8xxMxROM

Note the board around 1 minute 40 seconds. "High performance products likely to skip 20nm", which I would guess basically means AMD won't be releasing Excavator on 20nm.

 

[ShintaiDK]

Rather smallish isnt it? Each D1X module is 1.1mio sqf.

Whats the cash grant on this one? They got 1.4B$ for the previous one.

New fabs are in dire need since most old fabs cant be retrofitted for 450mm.

 

[SiliconWars]

Rather smallish isnt it? Each D1X module is 1.1mio sqf.

Yeah I guess that's why they are sitting half empty.

Whats the cash grant on this one? They got 1.4B$ for the previous one.

Large grants for projects like these are the norm. I'm sure Intel usually claims 20%+ for theirs.

 

[Ajay]

Note the board around 1 minute 40 seconds. "High performance products likely to skip 20nm", which I would guess basically means AMD won't be releasing Excavator on 20nm.

That's a bummer. I thought we might actually have a decent option again with XV @ 20nm - so much for that theory.

 

[SiliconWars]

That's a bummer. I thought we might actually have a decent option again with XV @ 20nm - so much for that theory.

It looks like GF have gone all-in for a low power 20nm whereas TSMC have gone for something in the middle. It makes sense from GF's perspective as this could give them a big performance win over TSMC (assuming it's not too late, which it probably will be).

 

[jpiniero]

Makes sense to make the fab 450 mm compatible, even though it's several years before it's really feasible.

 

[ShintaiDK]

Makes sense to make the fab 450 mm compatible, even though it's several years before it's really feasible.

Its about the height if I understand it correctly. Older fabs are simply unable to host 450mm equipment due to the ceiling height.

 

[Arachnotronic]

Yeah I guess that's why they are sitting half empty.

Such ignorance...let me educate your clearly anti-Intel self

In terms of your question of the levers that we have, I think that’s a good example of how we can be responsive in a very tactical time horizon to changes in demand. And as I look forward across the year, the prediction right now is we’re going to run at a healthy rate of utilization. In fact, we’re starting the quarter at that utilization rate. I think inventories will continue to be healthy.

Straight for Intel's most recent earnings call (http://seekingalpha.com/article/134...-results-earnings-call-transcript?part=single)

 

[SiliconWars]

Healthy huh, is that what they're calling it now.

http://www.itworld.com/data-center/352199/idle-intel-fabs-foretell-pc-decline

But idle they are. Intel's fabs may be at their lowest rates of utilization ever seen, according to Jim McGregor, president of Tirias Research, a semiconductor analyst based in Phoenix, Arizona

Intel's four fabrication plants in the Arizona area are all reportedly running below capacity, with one not even operational. Fab 12, which makes processors, has been below capacity; Fab 22, which makes chipsets, is also running below capacity, as is Fab 32. Fab 42, the newest, isn't even running

That doesn't sound so healthy to me.

McGregor says his contacts within Intel manufacturing put the company's global utilization capacity at around 60%. Normally, Intel's fabs run at 95% and only come down when the equipment is being upgraded.

Nor that. And it continues...

Quite a few of the fabs were taking 50% as many orders in Q4 as they did in Q3

Btw I'm not anti-Intel, I'm anti Intel-fanboy BS. Intel's huge double-sized fabs are sitting half empty while GF can't get enough capacity to fulfil demand.

 

[mrmt]

Healthy huh, is that what they're calling it now.

http://www.itworld.com/data-center/352199/idle-intel-fabs-foretell-pc-decline

Just FYI, the story was written before the quarterly call where the Intel exec mentioned healthy utilisation rates, so the exec probably was using Q4 and/or early Q1 data.

Between an executive in a quarterly call and the opinion of a research firm, I'd go with the former without much thinking. Not because the exec is more credible, but because the exec has *everything* to lose if he lie in a call.

 

[SiliconWars]

And what exactly is "healthy"? Did they give numbers?

 

[mrmt]

And what exactly is "healthy"? Did they give numbers?

I'm sure you know that this information is not open, but I'd interpret "healthy" as a rate that is not as good as it could be, but not offering risks company's annual guidance at risk, even if it's at the bottom of the interval. I'd also interpret that "healthy" means bigger than last quarter's, when Intel had to shut down fabs due to lack of demand.

 

[cytg111]

more room to mess more up in i suppose

 

[Abwx]

They are shipping the same number of CPUs while part
of their facilities were switched from 32 to 22nm with
smaller CPU dies so it makes no doubt that the available
waffers output should be well below max capacity , surely
in the 60% range at most.

 

[SiliconWars]

I'm sure you know that this information is not open, but I'd interpret "healthy" as a rate that is not as good as it could be, but not offering risks company's annual guidance at risk, even if it's at the bottom of the interval. I'd also interpret that "healthy" means bigger than last quarter's, when Intel had to shut down fabs due to lack of demand.

Did Intel describe their fab situation as "unhealthy" in Q4 last year? Because unless they did, we have no way of knowing if fab utilization is up or not.

"healthy" to me sounds like it could be anything so long as Intel isn't actually losing money, which will probably only happen at ~30% capacity. That still doesn't mean their fabs aren't sitting half-empty when most indicators clearly point to that. Could things have improved recently? Well I'd hope so given the Haswell launch but you never know.

By the sounds of things this situation is here to stay and Intel has too much capacity. Funnily enough, Otellini did say that there was oversupply trouble ahead 2 years ago. It appears he got the wrong victims though - http://www.eetimes.com/document.asp?doc_id=1258702

 

[mrmt]

Did Intel describe their fab situation as "unhealthy" in Q4 last year? Because unless they did, we have no way of knowing if fab utilization is up or

They didn't need to, did they? A stopped fab is incurred operational and capital costs for no reward. I don't think someone needs to label such an obviously bad situation as unhealthy.

Healthy by any standards you measure is not a stopped fab like in Q4, meaning that the situation improved in Q1/Q2.

 

[podspi]

That's a bummer. I thought we might actually have a decent option again with XV @ 20nm - so much for that theory.

Read it again, it says 'likely'. Excavator @ 20nm CONFIRMED

What is AMD going to do? They're going to have a heck of a time competing against Intel's 14nm on TSMC/GF 28nm...

 

[SiliconWars]

Read it again, it says 'likely'. Excavator @ 20nm CONFIRMED

What is AMD going to do? They're going to have a heck of a time competing against Intel's 14nm on TSMC/GF 28nm...

Likely to be skipped, Podspi.

In regards what they'll do, they'll have 20nm TSMC mid 2014 and 20nm GF later for mobile parts. They may have plans for Excavator or perhaps a new core for 14XM, or they maybe just no longer care.

 

[bryanW1995]

Yeah I guess that's why they are sitting half empty.



Large grants for projects like these are the norm. I'm sure Intel usually claims 20%+ for theirs.

Healthy huh, is that what they're calling it now.

http://www.itworld.com/data-center/352199/idle-intel-fabs-foretell-pc-decline

That doesn't sound so healthy to me.

Nor that. And it continues...

Btw I'm not anti-Intel, I'm anti Intel-fanboy BS. Intel's huge double-sized fabs are sitting half empty while GF can't get enough capacity to fulfil demand.

And what exactly is "healthy"? Did they give numbers?

Did Intel describe their fab situation as "unhealthy" in Q4 last year? Because unless they did, we have no way of knowing if fab utilization is up or not.

"healthy" to me sounds like it could be anything so long as Intel isn't actually losing money, which will probably only happen at ~30% capacity. That still doesn't mean their fabs aren't sitting half-empty when most indicators clearly point to that. Could things have improved recently? Well I'd hope so given the Haswell launch but you never know.

By the sounds of things this situation is here to stay and Intel has too much capacity. Funnily enough, Otellini did say that there was oversupply trouble ahead 2 years ago. It appears he got the wrong victims though - http://www.eetimes.com/document.asp?doc_id=1258702

I'm so glad that you're here to keep the riffraff in line. Too many blind intel fanboys lurking around taking up our precious cpu cycles. /sarcasm

Looks like GF is doing a whole lot better than most of us expected when AMD spun them off. That's good to see, if for no other reason than it is likely to help keep the AMD gpu division competitive.

 

[thilanliyan]

That's good to see, if for no other reason than it is likely to help keep the AMD gpu division competitive.

Doesn't AMD fab their GPUs at TSMC?

 

[podspi]

Likely to be skipped, Podspi.

In regards what they'll do, they'll have 20nm TSMC mid 2014 and 20nm GF later for mobile parts. They may have plans for Excavator or perhaps a new core for 14XM, or they maybe just no longer care.

(That was the joke). Can AMD even use TSMC 20nm for non-Jaguar chips? They have the WSA to fulfill. I am shocked that contract doesn't have any requirements for relative performance compared to other contract fabs...

 

[SiliconWars]

You would assume so but you never know. AMD is already "penalising" GF back by ditching SOI and reducing R&D payments to GF by going bulk.

Any clause that exists would presumably give AMD first refusal otherwise it must be the most one-sided "agreement" ever made. If AMD is contractually obliged to buy wafers from GF long-term, they'll be well served in demanding the latest tech even if it's only used as a bargaining chip.

I'd sure love to get my hands on the full details.

 

[Abwx]

Since Llano is branded HKMG by Glofo without mention if it is built
on SOI waffer one can only wonder if their 32nm is actualy SOI.

So either they re already bulk or they can use waffers of both kind ,
wich is possible since the insulator is just a thin film scatered
over the silicium , so it s quite possible that AMD benefit an
individualzed process flow.

http://www.globalfoundries.com/media/hkmg_videos.aspx

http://www.globalfoundries.com/technology/32-28nm.aspx

Edit : seems that it s still SOI as they are SOITECs
main customer.

 

[Ajay]

You would assume so but you never know. AMD is already "penalising" GF back by ditching SOI and reducing R&D payments to GF by going bulk.

Any clause that exists would presumably give AMD first refusal otherwise it must be the most one-sided "agreement" ever made. If AMD is contractually obliged to buy wafers from GF long-term, they'll be well served in demanding the latest tech even if it's only used as a bargaining chip.

I'd sure love to get my hands on the full details.

Use the search feature. It is one of the the most one sided agreements ever, but there are reasons for this. There have been several threads covering what details have been able to be dug up.

 

[NostaSeronx]

They have the WSA to fulfill.

How they got around that is by making a massive 550 mm² 32-nm CPU for 2014 called Warsaw.