AMD to fab all 7nm chips at TSMC

Published by

Click here to post a comment for AMD to fab all 7nm chips at TSMC on our message forum
https://forums.guru3d.com/data/avatars/m/243/243702.jpg
In other words, AMD was checking both GloFo and TSMC and GloFo was not good enough. That's why GloFo stated that most of their clients prefer 14/12nm. Likely anyone who checked their 7nm was not impressed by improvement over 14/12nm once price was factored in. I have feeling that reputational impact for GloFo is on its way. Stock Holders, Where are you? Gone?
https://forums.guru3d.com/data/avatars/m/103/103120.jpg
There were so many trolling that Intel got stuck at 10/7 generation. But eventually all of them got stuck in there. And eventually it's Intel and TSMC who are producing 10/7 chips already. At least in limited quantities. GloFo got stuck with 7 generation for longer. They had to license Samsung 7LP to catch up, but that wasn't also ready enough for 2018 so it lead to Zen 2 slip to 2019 effectively opening a way to Zen+ made on 14LP+, which was renamed to 12LP. Samsung eventually cancelled it in favor of EUV 7LP+, but it won't go production earlier of 2H 2019. Now, if TSMC production capacity wouldn't be enough there will be another Zen++ refresh next year and Zen 2 will not come out until 2020. Or, at least, they can make it on 14LP+ (12LP). Meanwhile 10+ Ice Lake-U are expected in december 2018 and Ice Lake-S in august next year.
https://forums.guru3d.com/data/avatars/m/243/243702.jpg
coth:

There were so many trolling that Intel got stuck at 10/7 generation. But eventually all of them got stuck in there. And eventually it's Intel and TSMC who are producing 10/7 chips already. At least in limited quantities. GloFo got stuck with 7 generation for longer. They had to license Samsung 7LP to catch up, but that wasn't also ready enough for 2018 so it lead to Zen 2 slip to 2019 effectively opening a way to Zen+ made on 14LP+, which was renamed to 12LP. Samsung eventually cancelled it in favor of EUV 7LP+, but it won't go production earlier of 2H 2019. Now, if TSMC production capacity wouldn't be enough there will be another Zen++ refresh next year and Zen 2 will not come out until 2020. Or, at least, they can make it on 14LP+ (12LP). Meanwhile 10+ Ice Lake-U are expected in december 2018 and Ice Lake-S in august next year.
Qualcomm believes that their 7nm made by TSMC will be here this year. Then idea that AMD's already completed revision of Zen 2 would have to wait for 7nm... They would just rearange it for 14/12nm. 2020 is year for Zen 3. And as AMD stated, they taped out MULTIPLE 7nm chip designs with TSMC. This means 3 or more products. And it means AMD knew about GloFo stopping their 7nm for quite some time. I really wonder... There is Zen 2 core for all those EPYC CPUs. There is 7nm Vega. What is 3rd or more? Maybe TR core? Maybe there is more than 1 Vega on 7nm?
https://forums.guru3d.com/data/avatars/m/258/258664.jpg
At least that sounds like Zen2 is still on track for 7nm. Good thing! I'm still waiting to see if I upgrade to Zen2 or anything Intel (less preferred) next year...
data/avatar/default/avatar28.webp
While I doubt TSMC can mass-produce all these products on 7nm, it's good news for consumers. TSMC is considered superior to GloFo in terms of performance and density. my issue is if they can scale 7nm to bigger chips.
https://forums.guru3d.com/data/avatars/m/103/103120.jpg
Fox2232:

Qualcomm believes that their 7nm made by TSMC will be here this year. Then idea that AMD's already completed revision of Zen 2 would have to wait for 7nm... They would just rearange it for 14/12nm. 2020 is year for Zen 3. And as AMD stated, they taped out MULTIPLE 7nm chip designs with TSMC. This means 3 or more products. And it means AMD knew about GloFo stopping their 7nm for quite some time. I really wonder... There is Zen 2 core for all those EPYC CPUs. There is 7nm Vega. What is 3rd or more? Maybe TR core? Maybe there is more than 1 Vega on 7nm?
3 is significantly less than 30 million. Again, as said - production capacity is still very limited. And there is a huge queue for 7FF. Aside of Qualcomm and AMD there are also Nvidia, Apple, MediaTek, Huawei etc etc. Even now even probably IBM.
data/avatar/default/avatar02.webp
HardwareCaps:

While I doubt TSMC can mass-produce all these products on 7nm, it's good news for consumers. TSMC is considered superior to GloFo in terms of performance and density. my issue is if they can scale 7nm to bigger chips.
Those are not good news for consumers in a long therm. In short therm, yeah probably 7nm Vega and Zen2 will rock, due to known cooperation between TSMC and NVIDIA, as we know what TSMC can do with their fabrication. But in a long therm ? Let's go back in time, when GloFo had started. TSMC acted like monopol back then. Higher prices, lower invesments in new research, roadmaps for the new products were wider and wider. Then GloFo came with their 14nm and TSMC had to fight for their place in the market again. So now 7nm is basically done, so it won't affect the end market that much. But what next? What in 3-5-7 yrs? Competition in any market is good for the customers, and saying that GloFo being in deep shit is good for us is a bold statement. So to be a little sarcastic. Enjoy your TSMC's Vega and Zen2, how it's better than GloFo and then cry in next 4 year that we got overpriced products from AMD too.
https://forums.guru3d.com/data/avatars/m/105/105985.jpg
I think they got what they wanted on both sides ...consoles low power is np for them I bet but forget bigger chips. plus the cost to make them is not worth it at this time just a guess
https://forums.guru3d.com/data/avatars/m/243/243702.jpg
coth:

3 is significantly less than 30 million. Again, as said - production capacity is still very limited. And there is a huge queue for 7FF. Aside of Qualcomm and AMD there are also Nvidia, Apple, MediaTek, Huawei etc etc. Even now even probably IBM.
Tapeout has different meaning than full production, it means that they finalized those designs into exact form they'll have in those following 30M+ chips which AMD's clients will be able to purchase. While it does not say that TSMC is ready for mass production, it means that AMD has chips which work way they are happy about. And that confirms that GloFo's decision to stop 7nm does not cause delay for AMD. From this we can presume that AMD is not behind their schedule. And there is no reason to make some rumors and catastrophic scenarios someone here may want to make.
https://forums.guru3d.com/data/avatars/m/243/243702.jpg
Jespi:

Those are not good news for consumers in a long therm. In short therm, yeah probably 7nm Vega and Zen2 will rock, due to known cooperation between TSMC and NVIDIA, as we know what TSMC can do with their fabrication. But in a long therm ? Let's go back in time, when GloFo had started. TSMC acted like monopol back then. Higher prices, lower invesments in new research, roadmaps for the new products were wider and wider. Then GloFo came with their 14nm and TSMC had to fight for their place in the market again. So now 7nm is basically done, so it won't affect the end market that much. But what next? What in 3-5-7 yrs? Competition in any market is good for the customers, and saying that GloFo being in deep crap is good for us is a bold statement. So to be a little sarcastic. Enjoy your TSMC's Vega and Zen2, how it's better than GloFo and then cry in next 4 year that we got overpriced products from AMD too.
Long term is not problem. Moment Samsung of someone who manages to get 7nm working in way GloFo considers it profitable enough, they'll jump on it. So now there is TSMC 7nm. From AMD's historical talk, they expected to be on 7nm, 7nm+, 7nm++, ... for quite some time. This means that even with 2 year delay on GloFo's side there's not going to be big pricing problem as Price gouging of 7nm+ vs 7nm is not something clients of forge will accept.
https://forums.guru3d.com/data/avatars/m/246/246171.jpg
With 1 less competitor, I wonder if this will cause TSMC to raise their prices. They have enough customers where if one of them doesn't like the price point and drops out, TSMC can be like "fine whatever, I still have more than enough work to do".
https://forums.guru3d.com/data/avatars/m/270/270008.jpg
I hope TSMC's process is good to go for AMD's larger chips like Vega and Zen2 and also will hit the frequencies needed. Making chips hit higher frequencies at 7nm / 10nm has been freaking hard ask Intel since all they have out so far is low power/frequency chips. If the process is good to go AMD is going to have a nice lead on Intel for a year or so.
https://forums.guru3d.com/data/avatars/m/243/243702.jpg
schmidtbag:

With 1 less competitor, I wonder if this will cause TSMC to raise their prices. They have enough customers where if one of them doesn't like the price point and drops out, TSMC can be like "fine whatever, I still have more than enough work to do".
There was always enough customers for all foundries. That's why those who wanted more business just had to make more foundries. I wonder if GloFo 14/12nm will remain competitive. That's main question. Price-Clock-PowerEfficiency. Those 3 have to be in somewhat competitive ratio, otherwise GloFo gets only customers which were not able to make deal somewhere else. This is not going to be problem in 2019 as 7nm production will not be that big. And TSMC which is going to invest more into 7nm is likely investing less into 14/12nm. But 2020 when there may be 7nm+... At least we'll see limits of each node.
https://forums.guru3d.com/data/avatars/m/271/271560.jpg
coth:

3 is significantly less than 30 million. Again, as said - production capacity is still very limited. And there is a huge queue for 7FF. Aside of Qualcomm and AMD there are also Nvidia, Apple, MediaTek, Huawei etc etc. Even now even probably IBM.
not quite. AMD and TSMC have a production contract, as does Apple and Qualcomm. everyone else will have to wait until their (TSMC) new factories are completed (with the help of the Taiwanese government). and Qualcomm is gunning for the Intel i-5. their new ARM (7nm) processors are designed to outperform Intel in laptop, tablet, and notebook with ALL DAY performance off of a single charge.
https://forums.guru3d.com/data/avatars/m/271/271560.jpg
schmidtbag:

With 1 less competitor, I wonder if this will cause TSMC to raise their prices. They have enough customers where if one of them doesn't like the price point and drops out, TSMC can be like "fine whatever, I still have more than enough work to do".
good question, but i doubt it. Samsung has targeted the foundry business for expansion, once they have their (similar) process finished they can apply economies of scale that TSMC doesn't possess (yet...it's what TSMC is going for).
https://forums.guru3d.com/data/avatars/m/243/243702.jpg
tunejunky:

not quite. AMD and TSMC have a production contract, as does Apple and Qualcomm. everyone else will have to wait until their (TSMC) new factories are completed (with the help of the Taiwanese government). and Qualcomm is gunning for the Intel i-5. their new ARM (7nm) processors are designed to outperform Intel in laptop, tablet, and notebook with ALL DAY performance off of a single charge.
Power efficiency wise... YES! But one needs usability of applications too. ARM Windows => x86 seems to be supported. x86_64 does not look like it has support (for now). For office work, it will be just fine... till you throw some big excel document on it. Productivity like video encoding will likely sux for x86 applications. Likely apps made for ARM will work much better. Games will be pretty weak. If given device can run Android applications in "sandboxed" mode, it would be much more attractive. But I expect that if anything. it will have dual boot.