One-time AMD fab limb, GlobalFoundries, is laying the litigation down on IBM to escape the $2.5 billion damages Big Blue is trying to squeeze out of it just ahead of its initial public offering (IPO). That, it states, is the prime driver behind IBM's recent demands with the company only looking for "a quick payday."
Also, an extension to this.Not before the deal they didn't. Power8 was built on IBM's 22nm SOI process.
It however glosses over that P8/P9's nodes were not developed internally at GlobalFoundries. Rather they got completed versions when the transfer from IBM to GloFo was completed in 2015.32HP(I guess that was POWER7+?)
While that is true, it doesn't invalidate my statement.
If GlobalFoundries had for sure any amount of customers, we would have seen Mubadala dumping cash like they did for 20LPM/14XM.
I think overall it deals with AMD had TSMC as first-source and IBM had Samsung as first-source from the get-go.Honestly I still suspect AMD had the balls to just go for TSMC regardless of WSA exactly with the reasoning that GF in that case would have to cancel 7nm and hence freeing AMD from the WSA for 7nm.
The day the control was in the hands of the experts from Wadiya it was over.I think overall it deals with AMD had TSMC as first-source and IBM had Samsung as first-source from the get-go.
7LP and 7LP+ are both largely different from 7FF and 7LPP.
Most likely if GlobalFoundries wanted to win such designs they would have to do what they did for Jaguar -> Puma.
Support a transfer of TSMC28HP -> GF28A w/ minimal mask changes:
Kabini/Temash -> Bhavani/Mullins/Beema/Carrizo-L
However on a scale of 7LP and 7LP+. To support each version would probably be costly. While GlobalFoundries expected to be first-source/only-source to maximize design choices they made.
Overall, however it seems AMD/IBM just didn't want to be on GlobalFoundries at all. GlobalFoundries simply made too many mistakes and basically tripped on a spike. Which doomed every leading edge investment after it.
View attachment 45666
Scalable for the future till it runs out of money.
Meanwhile TSMC is here with 150k wpm with three modules...
View attachment 45667
Samsung's capacity is much higher with less constraints on investment of capacity. Hence, S3 & V1 both are doing 7LPP, which is more than GlobalFoundries minuscule capacity.
"They would need another $3B to get to 12 Kwpm capacity" - Breakfastbytes-Cadence.
A customer would have to be ultra confident willing to live and die with GlobalFoundries. While, they fiscally bled out all over the place.
TSMC 150k wpm target.
Samsung 150k wpm target.
GlobalFoundries "okay everyone, give us $3B-$4B so we can get 12k-15k wpm. Then, if you want more we will talk about 50k-60k wpm with a second module for >$10B."
I think it is the opposite. Management was to US-centric. When the market for GlobalFoundries was outside of the US.The day the control was in the hands of the experts from Wadiya it was over.
Their process advances got aladeen-ed.
GlobalFoundries already received their EUV insertion tools. => "two massive ASML NXE:3400B tools to support the company’s future production (GF has cleared space for four machines in total)."There are things we don't know about the situation as well. One possibility was when was the earliest they could actually receive EUV machines. If their spot in line was 3 years away, it wouldn't do them, AMD or IBM any good.
So there's still a possibility of them doing R&D on lower finfets provided the money came from the government?GlobalFoundries already received their EUV insertion tools. => "two massive ASML NXE:3400B tools to support the company’s future production (GF has cleared space for four machines in total)."
Which is the same machines TSMC use for their N7+ process(wikichip source) and Samsung use for their 7LPP process(anandtech).
It isn't so much the spot in the line for GlobalFoundries with EUV tools. Rather they only have one module worth of space for EUV tools.
Fab 8.1, Module 2 | Takes time to build and with money which GloFo didn't have. Rather investing on their own, GlobalFoundries is relying on potential funds from the CHIPS Act. To fund development for that fab module.
So, the fab option for GlobalFoundries increasing capacity of EUV would have taken enough time to finish. So, that ASML could even out EUV tool production before GlobalFoundries completes the new fab module for EUV nodes (7LP+/5LP/3LP).
They have a possibility of restarting 7LP -> 7LP+ -> 5LP -> 3LP production if they get funded. If US&SGs gives significant money to GlobalFoundries, it normally gets duplicated by Mubadala(in prior setups).So there's still a possibility of them doing R&D on lower finfets provided the money came from the government?
There are things we don't know about the situation as well. One possibility was when was the earliest they could actually receive EUV machines. If their spot in line was 3 years away, it wouldn't do them, AMD or IBM any good.
It was built for both.Was GF even going to use EUV for their 7nm node?
I wonder what the actual timeline was. In October 2014 IBM gave GloFo its fabs including its staff and even paid $1.5 billion on top as well as access to IP. Only in 2018, close to when 7nm was supposed to be launched, GloFo notified both AMD and IBM that it would no longer pursue 7nm at all.
GloFo converted IBM's East Fishkill to RF nodes. Then, transferred the SOI nodes to Malta which was done in 2017.In the meantime IBM didnt order anything at 14nm from GF, hence how could they jeopardize their dwidling financial position without any additional revenue from manufacturing IBM products ?.
I'm not sure about that. AMD is very conservative with its financial forecasts, by now notoriously so. AMD planned to have a steady growth of 20-30% per year. Unless AMD changed its plans to that at some point before Zen's launch I can't see how GloFo would expect significantly higher sales than AMD itself, and even in that case GloFo would have to have known better at an earlier point than 2018.If you look at the timeline it is obvious that GF banked on AMD s Zen sales ( Zen was released in march 2017) to take off and get the necessary cash to get 7nm ready, but as everyone noticed sales were unexpectedly low and in march 2018 GF needed 2-4 billions to complete the process industrialisation.
Yea IBMs next gen design was based on GFs 7nm. When GF dropped their 7nm it put IBM at least a full year behind on its next power chip.
So is Ryzen being baked at TSMC then?
GloFo converted IBM's East Fishkill to RF nodes. Then, transferred the SOI nodes to Malta which was done in 2017.
POWER9
I'm not sure about that. AMD is very conservative with its financial forecasts, by now notoriously so. AMD planned to have a steady growth of 20-30% per year. Unless AMD changed its plans to that at some point before Zen's launch I can't see how GloFo would expect significantly higher sales than AMD itself, and even in that case GloFo would have to have known better at an earlier point than 2018.
Malta did SOI for a time.SOI process was always made in their Dresden plant...
But how much money did this provide.?.
likely very merginal revenues.
Before Zen in the public eye AMD was walking the fine line of bankruptcy (and stock below $2 in Feb. 2016), and regarding that no kind of WSA would have given more security. If GloFo really planned for more sales back then even before Zen launched it must have seemed like a pipe dream to everybody.AMD didnt change their plans but one has to admit that they didnt get as much revenue as expected during 2017, and as said it would have required 2-4bn to get 7nm fully industrialised, to compare to AMD s sales in 2017 wich were 5bn at most.
Unless AMD renegociated the WSA agreement in a favourable direction to GF there was no chance for the latter to take such a risk, and as you pointed it since AMD were conservative in their forecasts they didnt take the risk to sign for more waffers.
I'm not sure about that. AMD is very conservative with its financial forecasts, by now notoriously so. AMD planned to have a steady growth of 20-30% per year. Unless AMD changed its plans to that at some point before Zen's launch I can't see how GloFo would expect significantly higher sales than AMD itself, and even in that case GloFo would have to have known better at an earlier point than 2018.
That would make sense. But it's also the opposite of what has been stated before, GloFo supposedly citing lacking sales for not wanting to pursue 7nm. Can't be both at once, can it? When it announced the "strategy shift" GloFo mentioned none of these two reasons, though both a valid ways to save money of which GloFo was bleeding a lot before.GF first announced that they could not make enough 7nm CPUs for AMD before they canceled 7nm. It seems they did not build out enough in time and because of that, could not justify the loss for the investment.