This isn't a surprise is it? Kind of expect AMD is going to continue some form of business dealings with Global Foundries just it won't be as exclusive as it was in the first few years after going fabless.
This particular website aspires to journalistic integrity and attention to detail. "till" spread liberally around the text is amateur. Childish even. If that's the new aspiration, fine. Integrity takes a back seat.
This post was written, as breaking news, by an editor whose first language isn't English, while the other Senior Editors are out of action. Give a bit of slack.
When I do my live blogs, even as a native English speaker, there are spelling mistakes abound based on typing fast. That's the trade off with live and breaking news. This one word, is the grand scheme of things, is so obstinately minute compared to the mistakes I make on live news reporting, or any of our other titles, do you really think that one word is causing the downgrading of our reporting? Really? I mean, really really? If one cheerio comes out of the box malformed, do you swear off cheerios forever?
I would accept the first argument. Non-native English. OK, slack given. Teach him the correct word since we do not intend to actually till up the years. But then you make silly excuses for it... It's not a live blog, is it? It's a piece written and then published, not malformed out of the box. Do you believe in attention to detail or not? If you genuinely don't care... Fine.
woggs, cut them some slack. Mistakes happen. That's life. Anandtech has always been one of the best at minimizing them, especially considering the technical nature of their articles.
Is this what AnandTech has devolved to since Anand left? So-called "senior" editors lashing out at users for pointing out mistakes while making excuses for said mistakes, instead of... I dunno... fixing those mistakes... or even better, not making them at all? How fucking difficult is it to spend an extra 5 minutes proofreading?
The fact that "GlobalFoudries" in the title is misspelled is just icing on this shit cake.
There's nothing wrong with pointing out spelling mistakes but there is no need to be an ass about it. Did the misspellings in some way alter the message? No, it didn't so how does that affect the integrity. Should Anton strive to get better? Sure, but it takes time to become proficient in another language. Unless any of you all speak another language with perfect fluidity, fuck off. Ian is right to defend Anton.
It's a valid complaint, though. You sound like a kid rationalizing to their teacher when confronted with issues in a paper. It used to be that Anandtech was a high-quality, unbiased, well-presented site with deep insight into new developments in technology. There has been a serious decline in the overall quality and maturity of the content of this site since Anand's departure.
"Till" seems to be valid, even if in the US it is less liked and apparently considered less formal. There's some disagreement, but based on some wrong info.
EVER opened a dictionary and looked for those two words? They are the same. The only difference is that the older till is used less often except in phrases (till death...). There are indeed articles where ranting about language or typos makes sense, but this example isn't.
Shouldn't be... including the no penalty for going to others for 7nm. GF was in material breach of the previous wafer agreement, since AMD & GF were working on implementing 7nm together. When GF canceled 7nm that pretty much kiboshed the wafer agreement.
Yeah, I've been wondering if they could leverage FD-SOI for some low power APUs or something... once upon a time in a Fab far far way (in Germany) AMD fabbed all their processors on SOI. :) and maybe even mix in some eMRAM; haven't heard the status on that for FDX-12 yet.
GF statement should be: “AMD remains an important (cow to milk money from) as we reshape our portfolio to intensify investment in the technologies that provide the most value to customers, (since we don't have the money to invest in newer nodes) . We are proud to be a critical (suction point) of AMD’s current-generation 14nm and 12nm products, and we look forward to continuing to play a key role in (lowering) their next-generation products revenues.
Pretty much sums up the whole thing accurately. First AMD was forced to stay with GF or incur large fines, and now that they finally broken free of that constraint, it comes with a levy that AMD is obligated to continue using GF for something. This is the only reason AMD is using GF for the Zen2 and beyond interconnect (I/O) and now upcoming Ryzen+ mobility APUs (12nm).
In 2020, AMD is likely to only have I/O orders from GF (either on 14nm or 12nm).
That's going to make things interesting for laptops (and semi-custom) for Zen 2: with Uncore now locked to 12nm for the next few years, that's going to leave them at a power budget deficit compared to Intel, who are aggressively moving even their chipsets over to leading edge processes to keep power down. With Raven Ridge and Picasso AMD went for monolithic dies for CPU & GPU (rather than discrete dies for each), if they keep that up for Zen 2 moving to 7nm that's going to leave them short in the wafer agreement on their highest volume parts.
No. Intel effectively can't maintain production volume using 14nm, so they had to resort to 22nm to fill up some of the blanks and keep up with the demand (which seems like a step backwards).
AMD made a deal with GlobalFoundries that they need to keep ordering something from them in order to make a 'clean break'. This is just GF playing dirty and milking AMD for more revenue... but at least the royalty fees have been removed. So, AMD will only be using GlobalFoundries this year to produce 12nm Ryzen+ mobile stuff (APU's) along with the Interconnect for Zen 2. And next year (2020), they will probably only have the Interconnect on the back-order from GF for Zen 3.
That's kinda part of the deal for AMD to get away from GF. It shouldn't hamper them in the long run (hopefully).
For Zen 2, AMD is only using the GF to get the Interconnect... everything else is done at TSMC. This deal will basically result in Ryzen+ (12nm) APU's coming it this year (3xxx series) along with the Interconnect. And next year (2020), AMD will probably only have I/O connect done at GF (on either 14 or 12nm) for Zen 3.
Nah... the chiplets and most of the hardware for Zen 2 and Zen 3 will be done by TSMC (Zen 3 might use other fabs, but we're not there yet, so we don't know). Only the Interconnect will be done on 14nm for Zen 2, along with the 3xxx series 12nm AMD mobile products (Ryzen+ APU's) this year. Next year, we are likely only going to see the Interconnect for Zen 3 from GF... but other than that, I don't think AMD will be ordering anything else from GF.
Its a bit of a letdown that GF have locked AMD into making them buy stuff from them until 2021, but its a price for a complete break (and good riddance too)... but I'm hoping this won't hamper things too much for AMD (or at all). I do think the I/O can be scaled down to 7nm (contrary to what Lisa Su said - I think she may have said it originally because she was being diplomatic and hid the premise that one of the stipulations from moving away from GF was that they continue purchasing stuff from them for a certain time). Now that the royalty fees have been eliminated... its possible this stipulation for using GF until 2021 was part of the deal. Its better than AMD having to stay with GF.
I think what Lisa Su said was there was no need to move the I/O to 7nm not that it couldn't be. I thought I read somewhere about AMD getting back into the chipset business. This could be another thing that they put off on GF.
Hopefully the Zen2 line will be so successful that they will need all of GF capacity just to make the I/O chiplet.
Question: Is GloFo actively developing/optimizing it's 14 and 12 nm tech for lower power consumption, or have those efforts been put on hold also? Not to say that what Intel did with tinkering with its 14 nm FinFet tech (while running into a wall at 10 nm) is all great, but they did manage to improve the performance/power at least somewhat. So, how much air is left in GloFo's 14 and 12 nm process?
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
37 Comments
Back to Article
just4U - Tuesday, January 29, 2019 - link
This isn't a surprise is it? Kind of expect AMD is going to continue some form of business dealings with Global Foundries just it won't be as exclusive as it was in the first few years after going fabless.woggs - Tuesday, January 29, 2019 - link
"Till 2021" Really? Aren't you folks journalists? Is "Until" really too hard?Duckeenie - Tuesday, January 29, 2019 - link
If you look closely even 'til was too hard.Opencg - Tuesday, January 29, 2019 - link
I'm more of a logic and meaning guy than a grammar nazi. I'm sure you guys can find plenty of websites with good grammar if that's yo thingwoggs - Tuesday, January 29, 2019 - link
This particular website aspires to journalistic integrity and attention to detail. "till" spread liberally around the text is amateur. Childish even. If that's the new aspiration, fine. Integrity takes a back seat.Ian Cutress - Tuesday, January 29, 2019 - link
This post was written, as breaking news, by an editor whose first language isn't English, while the other Senior Editors are out of action. Give a bit of slack.When I do my live blogs, even as a native English speaker, there are spelling mistakes abound based on typing fast. That's the trade off with live and breaking news. This one word, is the grand scheme of things, is so obstinately minute compared to the mistakes I make on live news reporting, or any of our other titles, do you really think that one word is causing the downgrading of our reporting? Really? I mean, really really? If one cheerio comes out of the box malformed, do you swear off cheerios forever?
woggs - Tuesday, January 29, 2019 - link
I would accept the first argument. Non-native English. OK, slack given. Teach him the correct word since we do not intend to actually till up the years. But then you make silly excuses for it... It's not a live blog, is it? It's a piece written and then published, not malformed out of the box. Do you believe in attention to detail or not? If you genuinely don't care... Fine.Cooe - Wednesday, January 30, 2019 - link
Holy total dick, Batman! Get lost woggs, and preferably never come back. Your kind of jerk isn't wanted around here.woggs - Tuesday, January 29, 2019 - link
And I do like cheerios.sleepeeg3 - Wednesday, January 30, 2019 - link
woggs, cut them some slack. Mistakes happen. That's life. Anandtech has always been one of the best at minimizing them, especially considering the technical nature of their articles.The_Assimilator - Wednesday, January 30, 2019 - link
Is this what AnandTech has devolved to since Anand left? So-called "senior" editors lashing out at users for pointing out mistakes while making excuses for said mistakes, instead of... I dunno... fixing those mistakes... or even better, not making them at all? How fucking difficult is it to spend an extra 5 minutes proofreading?The fact that "GlobalFoudries" in the title is misspelled is just icing on this shit cake.
beginner99 - Wednesday, January 30, 2019 - link
It's like Spellcheckers haven't been invented yet.levizx - Wednesday, January 30, 2019 - link
Except spell checkers won't correct GlobalFoundries correctly. In fact my spell checker is telling me I spelt it wrong.GreenReaper - Wednesday, January 30, 2019 - link
Right-click, add to dictionary, you're done.Manch - Wednesday, January 30, 2019 - link
There's nothing wrong with pointing out spelling mistakes but there is no need to be an ass about it. Did the misspellings in some way alter the message? No, it didn't so how does that affect the integrity. Should Anton strive to get better? Sure, but it takes time to become proficient in another language. Unless any of you all speak another language with perfect fluidity, fuck off. Ian is right to defend Anton.NeBlackCat - Wednesday, January 30, 2019 - link
"This one word, is the grand scheme of things, is so obstinately minute"Oh dear. How exactly is it "obstinately minute"? Has it refused to grow despite being surrounded by manure?
Daeros - Wednesday, January 30, 2019 - link
It's a valid complaint, though. You sound like a kid rationalizing to their teacher when confronted with issues in a paper. It used to be that Anandtech was a high-quality, unbiased, well-presented site with deep insight into new developments in technology. There has been a serious decline in the overall quality and maturity of the content of this site since Anand's departure.Zizy - Wednesday, January 30, 2019 - link
If one chocolate comes with bugs in it, will you buy that brand again? I haven't, even though I used to buy 10-20 kg/year (not all for me obviously).sheh - Wednesday, January 30, 2019 - link
"Till" seems to be valid, even if in the US it is less liked and apparently considered less formal.There's some disagreement, but based on some wrong info.
https://www.merriam-webster.com/words-at-play/shou...
https://dictionary.cambridge.org/grammar/british-g... (wrongly said to be "until" shortened)
https://en.oxforddictionaries.com/definition/till
https://www.grammarly.com/blog/until-till-til/
https://grammarist.com/usage/until-till-til/
Zizy - Wednesday, January 30, 2019 - link
EVER opened a dictionary and looked for those two words? They are the same. The only difference is that the older till is used less often except in phrases (till death...). There are indeed articles where ranting about language or typos makes sense, but this example isn't.DeepLearner - Monday, February 4, 2019 - link
I cannot understand how anyone could be so upset by this. I didn't even notice. What is the matter with you?Freeb!rd - Tuesday, January 29, 2019 - link
Shouldn't be... including the no penalty for going to others for 7nm. GF was in material breach of the previous wafer agreement, since AMD & GF were working on implementing 7nm together. When GF canceled 7nm that pretty much kiboshed the wafer agreement.iwod - Tuesday, January 29, 2019 - link
Does the agreement includes the usage of their FD-SOI?I have been wondering if the I/O Die will switch to GF's FDX after 12nm.
Freeb!rd - Tuesday, January 29, 2019 - link
Yeah, I've been wondering if they could leverage FD-SOI for some low power APUs or something... once upon a time in a Fab far far way (in Germany) AMD fabbed all their processors on SOI. :) and maybe even mix in some eMRAM; haven't heard the status on that for FDX-12 yet.https://www.globalfoundries.com/news-events/press-...
yeeeeman - Wednesday, January 30, 2019 - link
GF statement should be:“AMD remains an important (cow to milk money from) as we reshape our portfolio to intensify investment in the technologies that provide the most value to customers, (since we don't have the money to invest in newer nodes) . We are proud to be a critical (suction point) of AMD’s current-generation 14nm and 12nm products, and we look forward to continuing to play a key role in (lowering) their next-generation products revenues.
deksman2 - Wednesday, January 30, 2019 - link
Pretty much sums up the whole thing accurately.First AMD was forced to stay with GF or incur large fines, and now that they finally broken free of that constraint, it comes with a levy that AMD is obligated to continue using GF for something.
This is the only reason AMD is using GF for the Zen2 and beyond interconnect (I/O) and now upcoming Ryzen+ mobility APUs (12nm).
In 2020, AMD is likely to only have I/O orders from GF (either on 14nm or 12nm).
edzieba - Wednesday, January 30, 2019 - link
That's going to make things interesting for laptops (and semi-custom) for Zen 2: with Uncore now locked to 12nm for the next few years, that's going to leave them at a power budget deficit compared to Intel, who are aggressively moving even their chipsets over to leading edge processes to keep power down. With Raven Ridge and Picasso AMD went for monolithic dies for CPU & GPU (rather than discrete dies for each), if they keep that up for Zen 2 moving to 7nm that's going to leave them short in the wafer agreement on their highest volume parts.Daeros - Wednesday, January 30, 2019 - link
Is this the same Intel that has had to move production of chipsets to older (22nnm) fabs to keep up with demand?deksman2 - Wednesday, January 30, 2019 - link
No. Intel effectively can't maintain production volume using 14nm, so they had to resort to 22nm to fill up some of the blanks and keep up with the demand (which seems like a step backwards).AMD made a deal with GlobalFoundries that they need to keep ordering something from them in order to make a 'clean break'.
This is just GF playing dirty and milking AMD for more revenue... but at least the royalty fees have been removed.
So, AMD will only be using GlobalFoundries this year to produce 12nm Ryzen+ mobile stuff (APU's) along with the Interconnect for Zen 2.
And next year (2020), they will probably only have the Interconnect on the back-order from GF for Zen 3.
That's kinda part of the deal for AMD to get away from GF. It shouldn't hamper them in the long run (hopefully).
deksman2 - Wednesday, January 30, 2019 - link
For Zen 2, AMD is only using the GF to get the Interconnect... everything else is done at TSMC.This deal will basically result in Ryzen+ (12nm) APU's coming it this year (3xxx series) along with the Interconnect.
And next year (2020), AMD will probably only have I/O connect done at GF (on either 14 or 12nm) for Zen 3.
johnNDenver - Wednesday, January 30, 2019 - link
I thought I read somewhere about AMD getting back into the chipset business. This could be another thing that they put off on GF.deksman2 - Thursday, January 31, 2019 - link
Possibly. At the moment there is limited information at best, so I'm just relaying what I know to date.:-)
eva02langley - Wednesday, January 30, 2019 - link
I am seeing potential console silicon on 12nm.Honestly, if they can just make IO die out of this deal, that would be awesome.
AMD needs to find an exit door from that contract.
deksman2 - Wednesday, January 30, 2019 - link
Nah... the chiplets and most of the hardware for Zen 2 and Zen 3 will be done by TSMC (Zen 3 might use other fabs, but we're not there yet, so we don't know).Only the Interconnect will be done on 14nm for Zen 2, along with the 3xxx series 12nm AMD mobile products (Ryzen+ APU's) this year.
Next year, we are likely only going to see the Interconnect for Zen 3 from GF... but other than that, I don't think AMD will be ordering anything else from GF.
Its a bit of a letdown that GF have locked AMD into making them buy stuff from them until 2021, but its a price for a complete break (and good riddance too)... but I'm hoping this won't hamper things too much for AMD (or at all).
I do think the I/O can be scaled down to 7nm (contrary to what Lisa Su said - I think she may have said it originally because she was being diplomatic and hid the premise that one of the stipulations from moving away from GF was that they continue purchasing stuff from them for a certain time).
Now that the royalty fees have been eliminated... its possible this stipulation for using GF until 2021 was part of the deal.
Its better than AMD having to stay with GF.
johnNDenver - Wednesday, January 30, 2019 - link
I think what Lisa Su said was there was no need to move the I/O to 7nm not that it couldn't be.I thought I read somewhere about AMD getting back into the chipset business. This could be another thing that they put off on GF.
Hopefully the Zen2 line will be so successful that they will need all of GF capacity just to make the I/O chiplet.
eastcoast_pete - Thursday, January 31, 2019 - link
Question: Is GloFo actively developing/optimizing it's 14 and 12 nm tech for lower power consumption, or have those efforts been put on hold also? Not to say that what Intel did with tinkering with its 14 nm FinFet tech (while running into a wall at 10 nm) is all great, but they did manage to improve the performance/power at least somewhat. So, how much air is left in GloFo's 14 and 12 nm process?zamroni - Thursday, January 31, 2019 - link
Amd should ditch global foundry. Gf is AMD's Achilles heel