bit-tech.net

Global Foundries GTC 2010

Comments 1 to 25 of 31

Reply
Snips 20th October 2010, 16:09 Quote
Why did AMD sell off Global Foundries when it clearly shows it can stand on it's own two feet. Probably better than AMD can in the present climate.

This article really does cast a few further clouds on AMD's horizon even with the positive vibes about Llano.
Xir 20th October 2010, 16:17 Quote
Quote:
it's started building Fab 8 in Saratoga County, New York...and when New York Fab 8 comes online next year
That's the former AMD "East Fishkill" Fab right? Initiated (decided, land bought and bonus from new york state granted and officially started) 2006, it'd better become online by 2012, sheesh!
Quote:
why hasn't it used the opportunity to trial 450mm equipment, which would achieve the capacity ramp that much better? Especially if they are buying new equipment to fill that new fab
Because the Equipment is non-existant...Semicon showed 2-3 Handlers/Chucks for 450mm wafers... Bah.

Very nice GloFo timeline... Timeline...especially since 2000-2006 (7ish) was AMD only :D
Quote:
Originally Posted by Snips
Why did AMD sell off Global Foundries when it clearly shows it can stand on it's own two feet.
Because they really, really needed the money :D And it is a trend.
The Jerry Sanders days of "Real men own real fabs" are over but for intel.
Maybe intel knows better and Sanders was right ;) Time will tell.
Bindibadgi 20th October 2010, 16:24 Quote
The Taiwan market has far more 450mm products. I don't know about the *whole* cycle - like handlers/machinery etc, but there are at least several companies here at one small innovation trade show (not even a semi conductor tradeshow!) ready for 450 in some form. It's specialist tech, there has to be demand for support companies to provide.
Xir 20th October 2010, 16:27 Quote
As long as AMAT and KLA aren't in....we'll see about the Taiwanese.
What did you see, FOUP's? (Waferpods) and storage devices?
Lazy_Amp 20th October 2010, 17:39 Quote
Quote:
Originally Posted by Snips
Why did AMD sell off Global Foundries when it clearly shows it can stand on it's own two feet. Probably better than AMD can in the present climate.

This article really does cast a few further clouds on AMD's horizon even with the positive vibes about Llano.

AHAHAHAHAHA

Man, if you people only knew...

Yes, this presentation from GF makes it absolutely evident that any problems with bringing Llano (and 32nm) to market is AMD's fault alone (Hint: Bad decisions on both sides).

Anyway sarcasm aside, it is good for AMD if GF expands to other companies, so that it can increase it's number of customers, which hopefully generates more revenue, which hopefully allows it to expand and bring out new tech faster.

And don't forget that AMD owns 44% of GF. If GF loses money, then AMD incurs losses as well.
Krayzie_B.o.n.e. 20th October 2010, 18:06 Quote
yep AMD sold off GF because they needed the cash, they almost sold off Opteron division to Oracle but it seems Oracle wanted the whole thing where as with GF AMD wanted to retain a large controlling share so no deal.

Seems AMD has some bright tech ideas but lack the business savvy of Intel to get it's chips into everything.
Intel owning their own private Fab ensures quality from point A to Z without interruption.

Whoever is running AMD gpu division needs to clone them self so they can run the CPU division too.
Snips 20th October 2010, 21:02 Quote
Quote:
Originally Posted by Lazy_Amp
Quote:
Originally Posted by Snips
Why did AMD sell off Global Foundries when it clearly shows it can stand on it's own two feet. Probably better than AMD can in the present climate.

This article really does cast a few further clouds on AMD's horizon even with the positive vibes about Llano.

AHAHAHAHAHA

Man, if you people only knew...

Yes, this presentation from GF makes it absolutely evident that any problems with bringing Llano (and 32nm) to market is AMD's fault alone (Hint: Bad decisions on both sides).

Anyway sarcasm aside, it is good for AMD if GF expands to other companies, so that it can increase it's number of customers, which hopefully generates more revenue, which hopefully allows it to expand and bring out new tech faster.

And don't forget that AMD owns 44% of GF. If GF loses money, then AMD incurs losses as well.

Unless, like in the past any GF profits were swallowed up by AMD's huge losses. I can't really see the funny side that you clearly seem to be enjoying.
matt... 20th October 2010, 22:35 Quote
Quote:
Originally Posted by Snips
Why did AMD sell off Global Foundries when it clearly shows it can stand on it's own two feet. Probably better than AMD can in the present climate.

This article really does cast a few further clouds on AMD's horizon even with the positive vibes about Llano.

It's all about returning shareholder value, and with AMD holding a large stake in GF it can maintain quality throughout the development cycle quite easily, while also freeing up GF's skills to service rival chip designers
needs which wouldn't have used GF if it was still part of AMD.

The money gained by spinning out 56 per cent of GF onto the stock exchange was probably more easily gained for a chip fab company than if a competitively challenged AMD had gone to the market cap in hand so it could raise funds to better compete against Intel...
Mentai 21st October 2010, 00:47 Quote
That was a great read. Nice work Richard.
Bindibadgi 21st October 2010, 03:19 Quote
Quote:
Originally Posted by Mentai
That was a great read. Nice work Richard.

You're most welcome :)
DiegoAAC 21st October 2010, 05:59 Quote
193 nm is Far UV, EUV is still far from ready for volume production.
450mm wafers will only be needed when the 'value' *PUs have over 10*10^9 transistors. Besides, didn't even the new Intel Oregon fab will be initially producing with 450 mm wafers.

If you are aimlessly requesting new process-features, at least request something crazy like volume electron-beam lithography or 3D circuits with multi-layer graphene heat spreaders.
Adnoctum 21st October 2010, 07:59 Quote
Quote:
Originally Posted by Article
AMD's own slides are slightly confusing because it claims both SOI (silicon on insulator) and High-k MG technologies are in use.
The two aren't incompatible. You can implement Hi-K MG on SOI.
Quote:
Originally Posted by Article
Whereas TSMC's 32nm node is exclusively SOI, with High-k MG launching only on 28nm; apparently sometime around now, although we think TSMC's roadmaps are optimistic at best - Global Foundries on the other hand, is firing ahead with High-k metal gate at both 32nm and 28nm. .
1. TSMC cancelled their 32nm node, which is why the AMD HD6xxx series is stuck at 40nm.
2. The cancelled node was bulk, and not SOI.
Quote:
Originally Posted by Article
...but the difference between both companies is that where Global Foundries is putting its 'High Performance' stamp on 32nm, TSMC is pushing the same thing at 28nm.
The GF 32nm process is SOI for CPUs, the TSMC 28nm process is bulk silicon for GPUs. The processes are for different needs and requirements. GF will have 28nm bulk process, so this should have been compared.
Quote:
Originally Posted by Snips
Why did AMD sell off Global Foundries when it clearly shows it can stand on it's own two feet. Probably better than AMD can in the present climate.
AMD didn't directly sell off their fabs. They partially sold a percentage of their fabs, and reduced their ownership further by inviting investment cash from Dubai, which is in turn enabling GF to offer these new processes, which AMD is now profiting (in the non-financial sense) from.
It isn't necessarily a bad thing that AMD has put their foundry at arms reach, for several reasons:
1. AMD can no longer afford to keep up with Intel's huge investment in fabs and process shrinks,
2. Sometimes AMDs fabs had run with excess capacity that no one else (say an ARM licensee) could use,
3. Sometimes AMDs fabs had no excess capacity (such as during the K8 era) and AMD had to go to Chartered (now a part of GF) to make more,
4. Outside customers means that the costs of process shrinks and fab upgrades can be spread over many others,
5. While AMD had direct control of the fabs, their competition (esp. nVidia) wasn't about to let AMD make their products which would compete with AMD. A similar thing happened to Asus, when their OEM customers forced them to split into Asus (consumer products) and Pegatron (contract manufacturer).

It should be pointed out that much of the losses that AMD has been suffering from have come from GF write-offs, and not from their core business (pun intended!). AMD should be better of going forward, but that remains to be seen.
Snips 21st October 2010, 08:07 Quote
I'm sorry but you cannot blame AMD losses on GF write-offs. An extremely poor showing on CPU performances, which in turn put potential buyers off, which in turn forces AMD to price drop, which in turn increase losses.
Fizzban 21st October 2010, 08:12 Quote
Am I the only one who wants a 450mm polished wafer as a mirror? :D
Adnoctum 21st October 2010, 08:38 Quote
Quote:
Originally Posted by Snips
I'm sorry but you cannot blame AMD losses on GF write-offs. An extremely poor showing on CPU performances, which in turn put potential buyers off, which in turn forces AMD to price drop, which in turn increase losses.

You shouldn't be authoritatively commenting on something you don't know about without putting "In my opinion..." first.
As a public company AMDs financial reports are published, and it has been widely reported that the losses are mostly down to GF rather than the CPU division.

"AMD makes a profit, until you account for GlobalFoundries":http://channel.hexus.net/content/item.php?item=26956
Bindibadgi 21st October 2010, 09:12 Quote
Quote:
Originally Posted by Adnoctum

1. TSMC cancelled their 32nm node, which is why the AMD HD6xxx series is stuck at 40nm.
2. The cancelled node was bulk, and not SOI.

I forgot to change that bit. I was reading older PRs from TSMC when I wrote it but spoke to AMD directly this week that confirmed 6000 was meant to be 32nm. Before it was unconfirmed.
Quote:

The GF 32nm process is SOI for CPUs, the TSMC 28nm process is bulk silicon for GPUs. The processes are for different needs and requirements. GF will have 28nm bulk process, so this should have been compared.


Duly noted!

DiegoAAC - depends on your definition of extreme vs far. Some people call 154nm extreme while others 193.
Xir 21st October 2010, 09:22 Quote
Quote:
Originally Posted by Adnoctum
3. Sometimes AMDs fabs had no excess capacity (such as during the K8 era) and AMD had to go to Chartered (now a part of GF) to make more.
And then after not even 6 months they pulled out of Chartered (as they couldn't handle the process) and rethought their decision to not build FAB36 in Dresden (which they then did)
:D
Quote:
Originally Posted by Fizzban
Am I the only one who wants a 450mm polished wafer as a mirror? :D
Question of price... Ask any of the Waferproducers for a sample, because samples are beeing manufactured (but expect to pay heftily, there are no production lines yet) :D
Adnoctum 21st October 2010, 10:23 Quote
Quote:
Originally Posted by Bindibadgi
Duly noted!

I'm not surprised, who can keep up with the ins-and-outs of the various processes at the foundries?
I saw a slide a few months ago showing GF's roadmap of processes. If I remember, there were three (3!!) 40nm bulk silicon processes: High Performance, Low Power and Super Low Power. At 28nm there will be two, HP (graphics cards) and SLP (mobile devices).

28nm bulk at both GF and TSMC at the same time. It will be interesting to see where AMD will go for their graphics cards. From what I've read, the core needs to be designed at an early stage for a given process, so no swapping between the two for production.

I wouldn't think they would go all one or the other, but split the HD7xxx line between the two, especially as GF is likely be somewhat capacity constrained. Fab 1 - Module 2 in Dresden is set to be fitted for 28nm making 25,000 wpm, and Fab 2 in NY won't be ready until 2012. While TSMC has two fabs capable of 100,000 wpm (although not all at 28nm), and building a third for a 2012 start.
Will the HD7xxx high end go to GF or TSMC? AMD has no doubt already made the decision.
Xir 21st October 2010, 12:15 Quote
I've started to slip a bit.

GloFo Fab 1 Module 1 used to be AMD FAb36
GloFo Fab 1 Module 2 used to be AMD Fab38 (which in turn was the revamped FAB30)

Correct?
Snips 21st October 2010, 12:26 Quote
Quote:
Originally Posted by Adnoctum
Quote:
Originally Posted by Snips
I'm sorry but you cannot blame AMD losses on GF write-offs. An extremely poor showing on CPU performances, which in turn put potential buyers off, which in turn forces AMD to price drop, which in turn increase losses.

You shouldn't be authoritatively commenting on something you don't know about without putting "In my opinion..." first.
As a public company AMDs financial reports are published, and it has been widely reported that the losses are mostly down to GF rather than the CPU division.

"AMD makes a profit, until you account for GlobalFoundries":http://channel.hexus.net/content/item.php?item=26956

So I go back to my original comment. How can GF be claiming they are making a profit now but according to AMD they are also not? Either way, you have to conceed that if AMD put up a better showing they would have showed a profit sometime in the last year.
Xir 21st October 2010, 12:34 Quote
I stand corrected

Luther Forest GloFo Fab...uhhh 8? Is not in East Fishkill , but somewhat to the north of that...although I can't find an adress.
Adnoctum 21st October 2010, 13:37 Quote
Interestingly the old numbers indicated the year since AMD was created: Fab 25 (now owned by Spansion) in Austin, Texas was built 25 years after AMD's creation, Fab 30 - 30 years, and so on. The new GF system screws it up a bit.
Adnoctum 21st October 2010, 14:22 Quote
Quote:
Originally Posted by Snips
So I go back to my original comment. How can GF be claiming they are making a profit now but according to AMD they are also not?

I don't pretend to be an accountant, nor very knowledgeable on such financial matters, but I think it goes like this.

The losses related to GF AREN'T cash losses, as in GF spent $100 but only earned $85, making a $15 loss.
These are reported as being non-cash losses, meaning the value of AMD's stake in GF is dropping in value.
Thus, I can only infer from this that the ATIC huge money injections are diluting AMD's stake, which then have to be included in AMD's results. Truthfully, I don't really know. Perhaps someone can find a link for us? I have looked.
Quote:
Originally Posted by Snips
Either way, you have to concede that if AMD put up a better showing they would have showed a profit sometime in the last year.

So if they had put out better products, they would have made more money?
Not that that helped AMD during the K8 era. I suppose there is only so much a free market can achieve in the face of corporate corruption.
As I said, the GF losses were paper losses, not actual money losses. It just means that AMD's stake in GF is worth less than it was last quarter. Revenue was actually up, and they made a PROFIT on their chip operations.

Anyway, this is off topic. The article was on GF and current and future processes and speculation on what this means.
Snips 21st October 2010, 15:10 Quote
Yes they would have made more money and their chip profits were more than likely GPU rather than CPU.
WildThing 21st October 2010, 15:44 Quote
Quote:
Originally Posted by Mentai
That was a great read. Nice work Richard.

+ 1, love these articles. :p
Log in

You are not logged in, please login with your forum account below. If you don't already have an account please register to start contributing.



Discuss in the forums