> So we will have to see just what AMD will use the increasingly dated 12/14nm process for over the coming years.
IIRC, all of AMD’s Chromebook APUs (both Ryzen and Athlon brands) are built on 12nm or 14nm. Maybe they’ll just ship a metric ton of those, as Windows 10X dies a quiet death.
The volume demand for Chromebooks may eat a little bit into that $1.6 billion, but how cheap are Chromebook APUs to fabricate? $15-20/die?
Will not help. People who have reviewed the Ryzen 3x based chromebooks have found that it is not performing in the level say the equivalent 12nm+ did against the Intel 14nm + competition in the Windows world. Looking at how Snapdragon 7c is faltering in Chrome OS vs Mediatek P70, I am inclined to think that Google is more to blame for not optimizing for AMD x64 architecture over Intel's.
<Quote> Regular benchmarks probably don’t tell the whole story here since AMD ships with a bit more onboard GPU ability than a comparable Intel chip (until Tiger Lake comes along, anyway), but I felt a bit of slowdown here and there with my standard workflow. Keeping 10-15 windows open across multiple Virtual Desk on dual displays, 10th-gen Intel devices don’t tend to show any signs of slowdown, but I saw it with this AMD 3500C. : When it comes down to it, you really have to decide if a few things matter to you. Do you like solidly built, feature-rich devices or do you just need a Chromebook that performs well? If performance, battery and screen are all you are after, something like the Acer Chromebook Spin 713 will be faster, have a better screen, and longer battery life for at least a few hundred dollars less. <EndQuote>
If you are paying big bux for a chrome book you are doing it wrong. They are not feature rich and are a way to offload old tech and still make some money for hardware companies. They are a commodity at this point.
Benchmarking chromebooks is a bit silly - "Keeping 10-15 windows open across multiple Virtual Desk on dual displays, 10th-gen Intel devices don’t tend to show any signs of slowdown, but I saw it with this AMD 3500C." I understand wanting to know what they can do but its so out of context with what other devices can do that it really doesn't tell you a ton.
Also, Chrome Unboxed is not a wonderful reference. I've subbed to them for awhile and it is essentially a chromebook marketing channel at this point.
Light duty they are great at. I have one that has lasted several years. I love the thing but only use it for specific tasks - access to remote/virtual desktop, a sound input for my stereo, tablet mode when i want to lounge.
I had the ASUS Eee PC in college which was linux and had similar subpar hardware for the time. It worked good enough. The truth is that chromebooks are wayyyy better than the 200 i paid for that laptop, but that is mostly due to better software and web services.
I have three chromebooks. All three are refurbished units The first one is a 2018 Acer celeron 15 inch non touch, then Acer Pentium N4200 touch screen convertible and Asus Pentium N4200 touch non convertible. The latter 2 I got in early 2020 due to Virtual Zoom classes for my kids for Covid. The highest price I paid was 230$ before taxes. Both my kids are comfortable working with them I will not pay > 400 for a chromebook. Acer one is larger and heavier, but is far better than the Asus whose only advantage is that it has 64GB internal storage compared to 32 GB for Acer.
These Ryzen powered chromebooks are a joke as far as price is concerned. The Core i3/i5 ones based even on 10th gen Core processors are cheaper and faster than those. If the manufacturers price them below 400$ then definitely they would be good for the money. Tiger lake i3 pummels the 10th gen i5 in benchmarks. Power is no contest. Even in Windows the last generation Surface used a custom processor with the 37xx series and it got totally wholloped by the Core i5 in all but some games due to the Graphics which Intel was always behind. Every person who reviewed that said that AMD Surface 3 makes no sense in any kind of usage. With the new Surface 4 being based on AMD 48xx series it has been a significant improvement in both power and performance.
Like AMD bought a knife to a gunfight with the 28nm chromebook APUs like 9220c, which got beat by Pentium 4200 in price/power/performance this time also it seems to continue to become irrelevant in the chromebook market.
I wanted to see a flood of older GPUs in this GPU dry market. AMD could have been pumping out cards through 2020-2021 had they kept up with GloFo which produced the chips for their Polaris and GCN cards.
^^^ This, now would be a decent time to spin up those RX 580 wafers again. Sadly, they could sell for RX 480 launch day prices ($250ish) and both make a killing selling 5 year old tech while also doing gamers a huge solid.
Newegg has people selling GT 1030 in that price range right now.
Maybe someone could organize a group buy where 1,000,000 people pledge to buy one of the above chips in partnership with the card manufacturers. Maybe a big name like LinusTechTips or something could be the advertiser and that way they help AMD to meet their wafer agreement.
Polaris might make sense for AMD to produce, but Vega really doesn't - the cost of packaging the HBM probably can't be made back vs. the card's likely selling price, even in the current distorted market.
I am a little surprised they haven't leaned on that more given the current shortage.
Those GPUs would be some seriously desperate measures.
I would like AMD to use these 12nm/14nm parts in their new chiplet design GPUs. New tech on old process. Hopefully such a contraption will end up being faster than VEGA, while having higher yields
So you are saying that *if* AMD got the entire Chromebook market of 12 million devices and *if* the APU cost was actually at the top of your range ($20) and *if* Chromebook sales persisted at COVID-19 hysteria pre-vaccination levels, AMD would be able to fill 240M/year our of their 1.6B quota with GF.
Some big ifs here and still quite a way to go to get to those 1.6B ...
I think the $1.6 billion going to global foundries. Is for when AMD absorbs Xilinx. I bet alot of their FPGAs can be produced on the 14 & 12nm nodes. Not to mention new builds of the older consoles. Since there's still a market for the older consoles.
What do people want older consoles for? I have read that the older games play fine on new ones. Even if that's false, there should be enough used HW out there for gamers, right?
The FPGA emulation scene is getting hotter and hotter.
And with 14nm or 12nm GloFlo process nodes, they can easily make earlier generation Consoles into a SoC and be done with it.
There's still a scene for retro consoles. Imagine how Retro Consoles of much earlier generations could perform if it was one SoC instead of all those different IC's linked together on the MoBo.
AMD is still producing the memory chiplet on GF. Every Ryzen or Eypc sold includes a chiplet produced by GF. $1.6Billion is no problem for them to meet if they keep doing this. But as others noted with Xilinx and their custom business they probably have no problem hitting that dollar figure.
We also don't know if that $1.6B includes services other than wafer starts. That could very well include testing, mounting and other services.
$1.6B over the next years is small change for AMD and they still need a lot of 14/12nm wafers for Rome, Milan & Ryzen 4000/5000 series IO dies, MB chipsets, etc. Newer generations have all just launched in the last 6 months and will still be in the market for that time frame (2023-2024).
I guess you also don't realize that this new deal means that AMD won't have to pay penalties for Xillinx products that are on TSMC's older nodes. Under the old WSA AMD could not freely use TSMC modes older than 7nm.
This is a much needed improvement on the old WSA and only people that are not focusing on the whole picture see it differently.
This supply agreement is dragging down by each iteration.
Die size is not AMDs Problem, it is power consumption and perf/watt. In both metrics they are trailing the leader (apple) and to be stuck in 2024 with 12nm, when you have to supply PCI6 (super high clocks -> massive power hog), DDR5 and the leader is 4 nodes more advanced (TSMC 2nm) is brutal.
I think it's a given that part of the reason for this move is to allow AMD to make 7nm IODs for the next generation of products.
But for the current products, they still need to make the dies, and these products will be available for a while yet, even after the next generation comes out in 2022.
And then there's low-end Chromebook fodder - Zen+ APUs and the like, that could be used for a long time yet.
And how dumb of you to assume 4 nodes of difference is big enough for IO? Remember 4 nodes ≠ 4 generations. For IO, 12LP+ will still be a full generation of advancement on 12LP. And AMD won't even ship 2nm by until Q42024.
I wasn't talking about future EPYC designs. Milan and even Rome are still ramping all we be at least for the next 2 years considering they are the last designs using DDR4 memory so the GF wafers in the WSA are put to good use going in the highest margin products. Milan is better than anything Intel has with DDR4 and next gen DDR5 Genoa will use 7nm IO dies when going against DDR5 Sapphire Rapids. Milan is better in performance and most importantly in yields and availability. Intel has 10nm high core count Ice Lakes but those are golden sample dies with the lowest possible yields.
Comparing server CPUs with Apple SoCs is comparing apples and oranges. You should compare Ryzen 5800U on TSMC 7nm with the M1 at 25W (which is what the M1 uses at full load) and see that AMD is highly competitive even being 1 node behind and using outdated Vega graphics instead of RDNA2. If the 5800U was using RDNA2 and TSMC 5nm like the A1 it would outperform it in every way. Apple has a node advantage and they are only able to use it because of their walled garden OS which means they are limited in market share growth. Just like on mobile.
Talking about AMD being stuck with 12nm in 2024 compared to Apple using TSMC 2nm (on top of comparing server CPUs with mobile SoCs) shows just how ignorant you are by not knowing that AMD has been shipping 7nm SoCs for the last 2 yeas and its roadmaps already show 6nm this year and 5nm SoCs next year.
Right now they could probably order a bunch more RX 580s and 590s and be pretty confident of selling them - though perhaps not a good use of their available graphics memory
AMD now should rebuy fabs in Dresden and New York, buy equpiment for make chips, and buy process from Samsung or TSMC or Intel, and they may make some optimization for their needs. This way AMD can quickly resolve shortages, and make their products more competitive.
I do agree but tsmc is going to be their main chip supplier the only thing I could think of them making silicon other than with tsmc is old 580s or chipsets or different chips for motherboards if they decide to make networking gear I suppose they would use it for that but 1.6 billion in 12-14nm is ALOT of freaking silicon.
Non-leading edge nodes are supposed to become more cost-efficient with newer lithography techniques, especially if the fab does not have to pay the leading-edge costs of developing the lines first. The current set of "non-leading edge" nodes is still very small and efficient for almost every application, anyway. Both AMD and Intel are going to order a large amount of non-leading edge silicon for IO and other non-core functions. GloFo's strategy is not a bad one considering the extreme cost and difficulty of developing higher transistor densities beyong 10nm.
I'm a fan of the idea of Glo Fo being used to make more RX 560's, 570's, 580's, the questions are, do they even have the capacity available currently, and can they ramp and start shipping them in time to cash in on the the current climate.
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.
Facebook
Twitter
RSS
41 Comments
Back to Article
ikjadoon - Thursday, May 13, 2021 - link
> So we will have to see just what AMD will use the increasingly dated 12/14nm process for over the coming years.IIRC, all of AMD’s Chromebook APUs (both Ryzen and Athlon brands) are built on 12nm or 14nm. Maybe they’ll just ship a metric ton of those, as Windows 10X dies a quiet death.
The volume demand for Chromebooks may eat a little bit into that $1.6 billion, but how cheap are Chromebook APUs to fabricate? $15-20/die?
All 12nm or 14nm, https://www.amd.com/en/processors/chromebook
Chromebooks are up 275% YoY, with shipments up from 3+ million to 12 million: https://chromeunboxed.com/chromebook-sales-q1-2021...
rocketbuddha - Thursday, May 13, 2021 - link
Will not help. People who have reviewed the Ryzen 3x based chromebooks have found that it is not performing in the level say the equivalent 12nm+ did against the Intel 14nm + competition in the Windows world.Looking at how Snapdragon 7c is faltering in Chrome OS vs Mediatek P70, I am inclined to think that Google is more to blame for not optimizing for AMD x64 architecture over Intel's.
https://chromeunboxed.com/lenovo-thinkpad-c13-yoga...
<Quote>
Regular benchmarks probably don’t tell the whole story here since AMD ships with a bit more onboard GPU ability than a comparable Intel chip (until Tiger Lake comes along, anyway), but I felt a bit of slowdown here and there with my standard workflow. Keeping 10-15 windows open across multiple Virtual Desk on dual displays, 10th-gen Intel devices don’t tend to show any signs of slowdown, but I saw it with this AMD 3500C.
:
When it comes down to it, you really have to decide if a few things matter to you. Do you like solidly built, feature-rich devices or do you just need a Chromebook that performs well? If performance, battery and screen are all you are after, something like the Acer Chromebook Spin 713 will be faster, have a better screen, and longer battery life for at least a few hundred dollars less.
<EndQuote>
bpurkapi - Friday, May 14, 2021 - link
If you are paying big bux for a chrome book you are doing it wrong. They are not feature rich and are a way to offload old tech and still make some money for hardware companies. They are a commodity at this point.Benchmarking chromebooks is a bit silly - "Keeping 10-15 windows open across multiple Virtual Desk on dual displays, 10th-gen Intel devices don’t tend to show any signs of slowdown, but I saw it with this AMD 3500C." I understand wanting to know what they can do but its so out of context with what other devices can do that it really doesn't tell you a ton.
Also, Chrome Unboxed is not a wonderful reference. I've subbed to them for awhile and it is essentially a chromebook marketing channel at this point.
Light duty they are great at. I have one that has lasted several years. I love the thing but only use it for specific tasks - access to remote/virtual desktop, a sound input for my stereo, tablet mode when i want to lounge.
I had the ASUS Eee PC in college which was linux and had similar subpar hardware for the time. It worked good enough. The truth is that chromebooks are wayyyy better than the 200 i paid for that laptop, but that is mostly due to better software and web services.
rocketbuddha - Saturday, May 15, 2021 - link
I have three chromebooks. All three are refurbished unitsThe first one is a 2018 Acer celeron 15 inch non touch, then Acer Pentium N4200 touch screen convertible and Asus Pentium N4200 touch non convertible. The latter 2 I got in early 2020 due to Virtual Zoom classes for my kids for Covid.
The highest price I paid was 230$ before taxes.
Both my kids are comfortable working with them
I will not pay > 400 for a chromebook.
Acer one is larger and heavier, but is far better than the Asus whose only advantage is that it has 64GB internal storage compared to 32 GB for Acer.
These Ryzen powered chromebooks are a joke as far as price is concerned. The Core i3/i5 ones based even on 10th gen Core processors are cheaper and faster than those. If the manufacturers price them below 400$ then definitely they would be good for the money. Tiger lake i3 pummels the 10th gen i5 in benchmarks. Power is no contest. Even in Windows the last generation Surface used a custom processor with the 37xx series and it got totally wholloped by the Core i5 in all but some games due to the Graphics which Intel was always behind. Every person who reviewed that said that AMD Surface 3 makes no sense in any kind of usage. With the new Surface 4 being based on AMD 48xx series it has been a significant improvement in both power and performance.
Like AMD bought a knife to a gunfight with the 28nm chromebook APUs like 9220c, which got beat by Pentium 4200 in price/power/performance this time also it seems to continue to become irrelevant in the chromebook market.
clsmithj - Friday, May 14, 2021 - link
I wanted to see a flood of older GPUs in this GPU dry market. AMD could have been pumping out cards through 2020-2021 had they kept up with GloFo which produced the chips for their Polaris and GCN cards.RX 590 (12nm)
VEGA 64 (14nm)
VEGA 56 (14nm)
RX 580 (14nm)
RX 570 (14nm)
RX 560 (14nm)
RX 550 (14nm)
gijames1225 - Friday, May 14, 2021 - link
^^^ This, now would be a decent time to spin up those RX 580 wafers again. Sadly, they could sell for RX 480 launch day prices ($250ish) and both make a killing selling 5 year old tech while also doing gamers a huge solid.Newegg has people selling GT 1030 in that price range right now.
Linustechtips12#6900xt - Friday, May 14, 2021 - link
would be a great idea for around 250$ like a 590 boosted edition or a 480+++ bassicallyHifihedgehog - Monday, May 17, 2021 - link
Maybe someone could organize a group buy where 1,000,000 people pledge to buy one of the above chips in partnership with the card manufacturers. Maybe a big name like LinusTechTips or something could be the advertiser and that way they help AMD to meet their wafer agreement.Spunjji - Friday, May 14, 2021 - link
Polaris might make sense for AMD to produce, but Vega really doesn't - the cost of packaging the HBM probably can't be made back vs. the card's likely selling price, even in the current distorted market.I am a little surprised they haven't leaned on that more given the current shortage.
psychobriggsy - Friday, May 14, 2021 - link
The issue isn't so much silicon as the substrates and packaging material, which would probably affect 12nm silicon as well. But otherwise, I agree.haukionkannel - Friday, May 14, 2021 - link
Yep... Is it better to make old 500 series again or use same material to produce new 6000 series... hmm... ;)meacupla - Friday, May 14, 2021 - link
Those GPUs would be some seriously desperate measures.I would like AMD to use these 12nm/14nm parts in their new chiplet design GPUs.
New tech on old process.
Hopefully such a contraption will end up being faster than VEGA, while having higher yields
Linustechtips12#6900xt - Friday, May 14, 2021 - link
Most likely it will go into chipsets and maybe some networking gear if they decide to pursue that pathArnulf - Friday, May 14, 2021 - link
So you are saying that *if* AMD got the entire Chromebook market of 12 million devices and *if* the APU cost was actually at the top of your range ($20) and *if* Chromebook sales persisted at COVID-19 hysteria pre-vaccination levels, AMD would be able to fill 240M/year our of their 1.6B quota with GF.Some big ifs here and still quite a way to go to get to those 1.6B ...
AMDrulZ - Thursday, May 13, 2021 - link
I think the $1.6 billion going to global foundries. Is for when AMD absorbs Xilinx. I bet alot of their FPGAs can be produced on the 14 & 12nm nodes. Not to mention new builds of the older consoles. Since there's still a market for the older consoles.ballsystemlord - Thursday, May 13, 2021 - link
What do people want older consoles for? I have read that the older games play fine on new ones. Even if that's false, there should be enough used HW out there for gamers, right?Kamen Rider Blade - Friday, May 14, 2021 - link
The FPGA emulation scene is getting hotter and hotter.And with 14nm or 12nm GloFlo process nodes, they can easily make earlier generation Consoles into a SoC and be done with it.
There's still a scene for retro consoles. Imagine how Retro Consoles of much earlier generations could perform if it was one SoC instead of all those different IC's linked together on the MoBo.
Spunjji - Friday, May 14, 2021 - link
They're cheaper than the new ones. That's basically it.rahvin - Wednesday, May 19, 2021 - link
They are cheaper than the new console. This allows MS and Sony to sell the older consoles into more price sensitive markets.rahvin - Wednesday, May 19, 2021 - link
AMD is still producing the memory chiplet on GF. Every Ryzen or Eypc sold includes a chiplet produced by GF. $1.6Billion is no problem for them to meet if they keep doing this. But as others noted with Xilinx and their custom business they probably have no problem hitting that dollar figure.We also don't know if that $1.6B includes services other than wafer starts. That could very well include testing, mounting and other services.
Blastdoor - Thursday, May 13, 2021 - link
This deal is getting worse all the timehttps://www.youtube.com/watch?v=3D8TEJtQRhw
sgeocla - Friday, May 14, 2021 - link
$1.6B over the next years is small change for AMD and they still need a lot of 14/12nm wafers for Rome, Milan & Ryzen 4000/5000 series IO dies, MB chipsets, etc. Newer generations have all just launched in the last 6 months and will still be in the market for that time frame (2023-2024).I guess you also don't realize that this new deal means that AMD won't have to pay penalties for Xillinx products that are on TSMC's older nodes. Under the old WSA AMD could not freely use TSMC modes older than 7nm.
This is a much needed improvement on the old WSA and only people that are not focusing on the whole picture see it differently.
node55 - Friday, May 14, 2021 - link
You might want to rethink the IO die strategy.It already consumes 46% of the power of Milan and is the reason Milan is power constraint and is underwhelming.
https://www.anandtech.com/show/16529/amd-epyc-mila...
This supply agreement is dragging down by each iteration.
Die size is not AMDs Problem, it is power consumption and perf/watt. In both metrics they are trailing the leader (apple) and to be stuck in 2024 with 12nm, when you have to supply PCI6 (super high clocks -> massive power hog), DDR5 and the leader is 4 nodes more advanced (TSMC 2nm) is brutal.
psychobriggsy - Friday, May 14, 2021 - link
I think it's a given that part of the reason for this move is to allow AMD to make 7nm IODs for the next generation of products.But for the current products, they still need to make the dies, and these products will be available for a while yet, even after the next generation comes out in 2022.
And then there's low-end Chromebook fodder - Zen+ APUs and the like, that could be used for a long time yet.
dotjaz - Saturday, May 15, 2021 - link
And how dumb of you to assume 4 nodes of difference is big enough for IO? Remember 4 nodes ≠ 4 generations. For IO, 12LP+ will still be a full generation of advancement on 12LP. And AMD won't even ship 2nm by until Q42024.sgeocla - Monday, May 17, 2021 - link
I wasn't talking about future EPYC designs. Milan and even Rome are still ramping all we be at least for the next 2 years considering they are the last designs using DDR4 memory so the GF wafers in the WSA are put to good use going in the highest margin products.Milan is better than anything Intel has with DDR4 and next gen DDR5 Genoa will use 7nm IO dies when going against DDR5 Sapphire Rapids. Milan is better in performance and most importantly in yields and availability. Intel has 10nm high core count Ice Lakes but those are golden sample dies with the lowest possible yields.
Comparing server CPUs with Apple SoCs is comparing apples and oranges. You should compare Ryzen 5800U on TSMC 7nm with the M1 at 25W (which is what the M1 uses at full load) and see that AMD is highly competitive even being 1 node behind and using outdated Vega graphics instead of RDNA2. If the 5800U was using RDNA2 and TSMC 5nm like the A1 it would outperform it in every way. Apple has a node advantage and they are only able to use it because of their walled garden OS which means they are limited in market share growth. Just like on mobile.
Talking about AMD being stuck with 12nm in 2024 compared to Apple using TSMC 2nm (on top of comparing server CPUs with mobile SoCs) shows just how ignorant you are by not knowing that AMD has been shipping 7nm SoCs for the last 2 yeas and its roadmaps already show 6nm this year and 5nm SoCs next year.
Kamen Rider Blade - Friday, May 14, 2021 - link
AMD's X570 Chipset was built on GloFlo 14nm.AMD's 300/400 series Chipset was built on GloFlo 55nm.
So I can forsee room for more 14nm Chipsets in the future for the full X#70 / B#50 / A#20 Chipsets all on 14nm.
GloFlo still has 12nm LP & LP+ Process Nodes.
So that should satisfy all their AM4 / AM5 ChipSet needs moving forward for the next few years.
Flynny123 - Friday, May 14, 2021 - link
Right now they could probably order a bunch more RX 580s and 590s and be pretty confident of selling them - though perhaps not a good use of their available graphics memoryManuelDiego - Friday, May 14, 2021 - link
"...US fab GlobalFoundries...". The United States of Abu Dhabi?dotjaz - Saturday, May 15, 2021 - link
Are you dumb? GloFo's advanced nodes are all coming from their New York fab.TristanSDX - Friday, May 14, 2021 - link
AMD now should rebuy fabs in Dresden and New York, buy equpiment for make chips, and buy process from Samsung or TSMC or Intel, and they may make some optimization for their needs. This way AMD can quickly resolve shortages, and make their products more competitive.Linustechtips12#6900xt - Friday, May 14, 2021 - link
I do agree but tsmc is going to be their main chip supplier the only thing I could think of them making silicon other than with tsmc is old 580s or chipsets or different chips for motherboards if they decide to make networking gear I suppose they would use it for that but 1.6 billion in 12-14nm is ALOT of freaking silicon.Kamen Rider Blade - Monday, May 17, 2021 - link
That's going backwards, there's a reason why AMD sold off Global Foundaries.There's no point in buying back GloFlo.
del42sa - Friday, May 14, 2021 - link
so now they will supply IO die using 12nm LP+ which is an equivalent of 10nm and later ?msroadkill612 - Friday, May 14, 2021 - link
Light duty old tech Intel SOC, vs same from AMD APU???One has a more competent CPU & crap IGP,.
AMD have competent CPU, much better graphics & (AM4) platform IMO.
Of these ingredients, the one which would be most missed... the biggest dealbreaker for most folks, would be poor graphics.
For Intel to compete, they need dgpu (cost & power issues), or a new design w/ decent IGP (defeats the point of cheap old tech?).
msroadkill612 - Friday, May 14, 2021 - link
Re fundamentals, AMD's Infinity Fabric Bus happily hosts GloFo processors alongside TSM 7nm chips.They are well positioned to extend the economic life of older tech in subsidiary roles.
Composite - Friday, May 14, 2021 - link
Not sure if merger with Xilinx can help with this situation. Xilinx FPGA could consume some of those 14/12 nm process?Railander - Saturday, May 15, 2021 - link
the reflection on the wafer on the right looks like a masked PogChamp lolMachinus - Saturday, May 15, 2021 - link
Non-leading edge nodes are supposed to become more cost-efficient with newer lithography techniques, especially if the fab does not have to pay the leading-edge costs of developing the lines first. The current set of "non-leading edge" nodes is still very small and efficient for almost every application, anyway. Both AMD and Intel are going to order a large amount of non-leading edge silicon for IO and other non-core functions. GloFo's strategy is not a bad one considering the extreme cost and difficulty of developing higher transistor densities beyong 10nm.zaza - Monday, May 17, 2021 - link
I am guessing that will be using this for future chipsets and IO die for their processors.artk2219 - Tuesday, May 18, 2021 - link
I'm a fan of the idea of Glo Fo being used to make more RX 560's, 570's, 580's, the questions are, do they even have the capacity available currently, and can they ramp and start shipping them in time to cash in on the the current climate.