Feels like a pretty big deal. Maybe some short term disruption to AMD, but surely that's a good move overall? The WSA seems certainly more limiting to AMD - if they need high performance products it seems hard to say that they've always been better off on GF.
It'll be interesting to see what happens to GCN moving back to TSMC as well. It certainly wont account for the entire performance difference, but the TSMC finfet processes have all been measurably better than the 14nm process that Glofo licensed, right?
IMO hard to say if the difference of TSMC 16nm vs GF14nm was "measurably better" because my understanding is there was not a single product made on both of them by the same company making the product using a specific chip "tooled" for both of them (example...GTX 1050 Ti at TSMC AND GF to see the real difference)
The only way to know for sure for sure would be to get 2 more or less equal produces (as far as spec were to be concerned) and compared them directly as far as the speed they could achieve, power they use, heat they push out and so forth.
way back in the day they did this with vehicle engines same spec at different plants and you WANTED from plant X not plant Y because it was that much better...
I dream of the day we are able to "unhook" a portion of the chip (like a lego block) so when you absolutely do not need the power, unhook it completely, or, add a block on...like HBM but more substantial, kind of surprised they have not done this, easy to add performance or remove performance while knowing exactly what that block would be capable of (for the maker of it)
I know internally Intel, Ngreedia and the like do this (by enabling or disabling features by "clipping" the die) would be kind of nice in a way though cause they could sell a "base core" and the end user who has extra coin can make it a "beast core" ^.^
Letting you "add" to such a highly integrated circuit doesn't make sense technically or financially. It's far cheaper to build the full chip and deactivate the defective units (or even fully functioning ones) than it is to build dozens of completely different SKUs that can later be combined somehow.
So I'm reasonably sure that when it comes to chips you'll get less flexibility in the future, not more. If we're lucky they won't move exclusively to HBM and have to upgrade the CPU for more RAM. :)
Gen-Z has some potential in that area, or even Infinity Fabric. Something like the old school idea of 2+ CPU systems, or co-processors being in the system. The problem these days is that CPU performance is tied very closely to how quickly different components within a CPU can talk to each other, so going to another physical "chip" would cause a huge performance hit. We can't even get video cards where we can upgrade the video memory by there being sockets/slots on the card at this point, which would free video card makers from being caught between higher memory prices and a MSRP.
This announcement cleverly states nothing while making people who weren't paying attention think that AMD knew about GloFo never producing 7nm all along... you know, like when Cutress ran around GloFo's fab in that bunny suit not too long ago.
AMD was *always* going to produce its expensive high-end parts at TSMC and this has been known for a long time.
What was also known was that desktop RyZen 2 and the APUs that actually make up the bulk of AMD's sales were supposed to be made at GloFo because GloFo was supposed to be *cheaper*.
Well guess what: The cheaper option is out the door, and unlike GloFo, TSMC is not an incompetently run charity that exists to give a third-tier player like AMD steep discounts when the big boys will pay real money for 7nm capacity.
Mark my words, there *will* be an impact on consumer chips over this even if some low-clocked server parts launch next June.
I don't get it when AMD said this has no impact on their deliveries. GPU and EPYC were targeting TSMC's 7nm and thus they are fine. But what about Zen2 based consumer CPU/APU? It depends on when AMD was informed of this change, we might not see those CPU/APU in 2019.
Has there been any indication that Zen2 CPUs won't be based on the same dice as Zen2 Epyc (and thus fabbed by the same company)? If so, that would be a noteworthy change. I wouldn't be surprised if 3000-series APUs are delayed or otherwise changed, though.
It was actually planned that both TSMC and Global Foundries would BOTH be making Ryzen third gen processors, because the 7nm equipment used by both were similar enough to allow it(for the first time in ages). I suspect that poor management at Global is to blame for their problems while TSMC was able to get things working.
All products were to switch to 7nm by 2020, so there's no longer any need for 14nm/12nm. That means they'll have to make a huge payment to get out of the WSA.
Two things 1) As GF abandoned 7nm would AMD still be liable to penalty payments under the WSA? 2) Chipsets do not need 7nm and can still be produced at GF
I hope this contract contained paragraphs for this kind of thins also like sunken dev costs. I mean a chip that was to tape out soon already cost double digits in millions. Will GF pay that back? And AFAIK back in the day it was clear the WSA was only valid as long as GF is offering competitive performance / nodes. So maybe this could be a hug ewin for AMD getting rid of the WSA. Also how IBM will react? IBM is huge and has huge funds so that will probably be a long and dirty legal battle. After all power cpus really heavily on high clocks and they had a deal that GF develops a process just fro IBM...Can they just toss that in the trash without serious legal or monetary consequences? doubtful.
This could be both good and bad news for AMD. It could allow them to launch Ryzen with Zen 2 sooner but the big question is how it clocks. They need 5GHz+ not 4GHz, can they achieve that at TSMC?
What makes you believe that AMD needs to hit 5Ghz on Zen 2 parts when they have the IPC advantage over INTEL right now? INTEL over advantages is frequency, they can't really gain more IPC out of their refined stagnate architecture.
I think they are finding out the same troubles that Intel found out with 10nm - but keep in mind in Intel calculations there 10nm is same as other 7nm. Keep in mind also it does not just the frequency or the number cores that matter, but what is under the hood. I believe with later Intel CPU's they found they can get more performance in same process.
I try to remain objective externally, but internally I'm an AMD guy, pulling for the underdog.
That being said, at best AMD matches Intel's IPC, they don't beat it.
Intel has a few differences in their integer setup and AVX processing that everyone is optimized for, runs right past AMD.
Now if AMD were optimized for it has a more powerful integer setup, though not necessarily faster.
I don't think it is the lithography that is limiting the clock speed. The FX and Sandy Bridge could hit 5GHz on 32nm. I've got an FX that can hit 5GHz any time I feel like it, chip can handle it all day long, though the motherboard can't.
Regardless, lets say that AMD has the IPC edge by 5%, Intel can put out a 5GHz CPU *today* and beat that. Intel actually could make a few changes to their "stagnate" architecture and push more out of it. It isn't stagnant, if you look back at the architecture this one begins with the Pentium Pro and has been refined and added to ever since. Honestly, Nehalem was a big difference from the Core 2, Sandy Bridge was a big departure from Nehalem, Haswell was a rather big jump from Sandy Bridge, though they haven't made huge changes since. Sandy Bridge had enough changes made it could have been viewed as a different architecture and if you look at the entire line Intel has made huge changes in the Core ix lineup and kept the name. Previously architecture changes of that magnitude have warranted a new name. The PIII to Banias to Dothan realistically were more similar than the Nehalem to Sandy Bridge to Haswell.
This is where third generation Ryzen comes into play. First generation Ryzen to second fixed some memory/cache related performance limitations, but not a lot of big design changes. The rumors are that third generation may provide a 10-15 percent IPC boost to Ryzen. If the jump to 7nm gets AMD to 5GHz, plus that IPC improvement, then AMD may very well have an IPC lead, and not much slower than what Intel chips can hit for clock speed. That is why the jump to 7nm is so important, because it has the potential to put AMD in the lead in single-threaded, as well as multi-threaded workloads.
The perception by many people is that clock speed is the one area that allows Intel to dominate in certain areas, such as gaming, and those single threaded benchmarks that don't really matter. If AMD is able to hit 5GHz, any advantage that Intel had in terms of performance would be gone, with the exception where extra memory channels come into play(since Ryzen branded products only have a dual-channel memory controller due to socket AM4).
A lot of this doesn't make sense. AMD claims that its 7 nm chips will be fabbed by TSMC; on the other hand, quite reliable industry sources and observers state pretty definitively that at least 75% of TSMC's 7 nm capacity will go to Apple, at least in 2018, and likely forward, with the remainder essentially all for mobile SoCs from Huawei, Qualcomm, plus some cryptomining ASICs. For example, this https://www.eetimes.com/document.asp?doc_id=133361... in the eetimes. So, either TSMC can ramp up its 7 nm fab A LOT, or AMD will only ship a small number of some EPYCs on 7 nm so they can say they shipped 7 nm chips, or they are just giving us story right now. I believe AMD's CPUs and GPUs will be on 12 nm for another two years or more, unless they can make a deal with Samsung, and Samsung gets its 7 nm process geared up for large transistor counts. Right now, I'd need to see an official TSMC announcement of a large volume deal with AMD for 2018/19 before I buy AMD's story, and I am not the only one.
TSMC doesn't have a lot of X86 chips in the making and especially none of them are on 7nm. It was a clear indicative that they are interested because of their lack of experience in the field.
I don't see a huge delay happening anytime soon. Like stated before, AMD was aware of this for a long time and they had plan their root way before EPYC ROME at TSMC.
X86 or ARM doesn't matter, both use same wafers to be fabbed. In fact the situation is quite worse for x86 because desktop CPUs have much bigger die size u usually.
So yeah, AMD will definitely have problems getting enough fab capacity for their 7nm products especially with needing it for both CPUs and GPUs.
I don't think Intel would manufacture AMD's chips. However, I wouldn't be surprised if Intel swoops in, takes the tech, the IP, the equipment (mucho dinero DUV ASML kit!) and key personnel of GloFo's hands for cheap, and gets IBM's blessings (any buyer would need that, IBM owns key IP) in return for manufacturing IBM's mainframe CPUs and other silicon in 7 nm. An Intel acquisition of the tech (or the entire foundry?) would also take care of the Trusted Foundry problem.
When AMD had their lawsuit with Intel, AMD should have forced Intel to split off their fabs as a part of the settlement. That would have allowed AMD to use the Intel fab process without it being a part of Intel.
GloFo's announcement was pretty clear that they will, at least for now, seize all development of 7 nm and lower nodes, so that includes 5 nm. I believe that this "we'll make our chips on 7 nm at TSMC" announcement by AMD is a mix of whistling in the dark, trying keep to some analysts happy (for how long?), and might even be somewhat of an attempt to keep AMD's personnel calm. Let's face it: the GloFo announcement was a surprise for AMD (and most everybody else), and AMD really got screwed by GloFo's sudden exit from 7 nm. GloFo was supposed to be their 7 nm volume supplier, but apparently the volume wasn't enough for GloFo to keep plowing billions of $$$ into getting fab 8 up-and-running on 7 nm DUV.
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.
28 Comments
Back to Article
Drumsticks - Monday, August 27, 2018 - link
Feels like a pretty big deal. Maybe some short term disruption to AMD, but surely that's a good move overall? The WSA seems certainly more limiting to AMD - if they need high performance products it seems hard to say that they've always been better off on GF.It'll be interesting to see what happens to GCN moving back to TSMC as well. It certainly wont account for the entire performance difference, but the TSMC finfet processes have all been measurably better than the 14nm process that Glofo licensed, right?
Dragonstongue - Monday, August 27, 2018 - link
IMO hard to say if the difference of TSMC 16nm vs GF14nm was "measurably better" because my understanding is there was not a single product made on both of them by the same company making the product using a specific chip "tooled" for both of them (example...GTX 1050 Ti at TSMC AND GF to see the real difference)The only way to know for sure for sure would be to get 2 more or less equal produces (as far as spec were to be concerned) and compared them directly as far as the speed they could achieve, power they use, heat they push out and so forth.
way back in the day they did this with vehicle engines same spec at different plants and you WANTED from plant X not plant Y because it was that much better...
I dream of the day we are able to "unhook" a portion of the chip (like a lego block) so when you absolutely do not need the power, unhook it completely, or, add a block on...like HBM but more substantial, kind of surprised they have not done this, easy to add performance or remove performance while knowing exactly what that block would be capable of (for the maker of it)
I know internally Intel, Ngreedia and the like do this (by enabling or disabling features by "clipping" the die) would be kind of nice in a way though cause they could sell a "base core" and the end user who has extra coin can make it a "beast core" ^.^
close - Tuesday, August 28, 2018 - link
Letting you "add" to such a highly integrated circuit doesn't make sense technically or financially. It's far cheaper to build the full chip and deactivate the defective units (or even fully functioning ones) than it is to build dozens of completely different SKUs that can later be combined somehow.So I'm reasonably sure that when it comes to chips you'll get less flexibility in the future, not more. If we're lucky they won't move exclusively to HBM and have to upgrade the CPU for more RAM. :)
Targon - Wednesday, August 29, 2018 - link
Gen-Z has some potential in that area, or even Infinity Fabric. Something like the old school idea of 2+ CPU systems, or co-processors being in the system. The problem these days is that CPU performance is tied very closely to how quickly different components within a CPU can talk to each other, so going to another physical "chip" would cause a huge performance hit. We can't even get video cards where we can upgrade the video memory by there being sockets/slots on the card at this point, which would free video card makers from being caught between higher memory prices and a MSRP.CajunArson - Monday, August 27, 2018 - link
This announcement cleverly states nothing while making people who weren't paying attention think that AMD knew about GloFo never producing 7nm all along... you know, like when Cutress ran around GloFo's fab in that bunny suit not too long ago.AMD was *always* going to produce its expensive high-end parts at TSMC and this has been known for a long time.
What was also known was that desktop RyZen 2 and the APUs that actually make up the bulk of AMD's sales were supposed to be made at GloFo because GloFo was supposed to be *cheaper*.
Well guess what: The cheaper option is out the door, and unlike GloFo, TSMC is not an incompetently run charity that exists to give a third-tier player like AMD steep discounts when the big boys will pay real money for 7nm capacity.
Mark my words, there *will* be an impact on consumer chips over this even if some low-clocked server parts launch next June.
enzotiger - Monday, August 27, 2018 - link
I don't get it when AMD said this has no impact on their deliveries. GPU and EPYC were targeting TSMC's 7nm and thus they are fine. But what about Zen2 based consumer CPU/APU? It depends on when AMD was informed of this change, we might not see those CPU/APU in 2019.Valantar - Tuesday, August 28, 2018 - link
Has there been any indication that Zen2 CPUs won't be based on the same dice as Zen2 Epyc (and thus fabbed by the same company)? If so, that would be a noteworthy change. I wouldn't be surprised if 3000-series APUs are delayed or otherwise changed, though.Targon - Wednesday, August 29, 2018 - link
It was actually planned that both TSMC and Global Foundries would BOTH be making Ryzen third gen processors, because the 7nm equipment used by both were similar enough to allow it(for the first time in ages). I suspect that poor management at Global is to blame for their problems while TSMC was able to get things working.ghostbit - Monday, August 27, 2018 - link
Meh, GF PDKs aren't that good anyways.lefty2 - Tuesday, August 28, 2018 - link
All products were to switch to 7nm by 2020, so there's no longer any need for 14nm/12nm. That means they'll have to make a huge payment to get out of the WSA.Duncan Macdonald - Tuesday, August 28, 2018 - link
Two things 1) As GF abandoned 7nm would AMD still be liable to penalty payments under the WSA?2) Chipsets do not need 7nm and can still be produced at GF
beginner99 - Tuesday, August 28, 2018 - link
I hope this contract contained paragraphs for this kind of thins also like sunken dev costs. I mean a chip that was to tape out soon already cost double digits in millions. Will GF pay that back?And AFAIK back in the day it was clear the WSA was only valid as long as GF is offering competitive performance / nodes. So maybe this could be a hug ewin for AMD getting rid of the WSA.
Also how IBM will react? IBM is huge and has huge funds so that will probably be a long and dirty legal battle. After all power cpus really heavily on high clocks and they had a deal that GF develops a process just fro IBM...Can they just toss that in the trash without serious legal or monetary consequences? doubtful.
jjj - Tuesday, August 28, 2018 - link
This could be both good and bad news for AMD.It could allow them to launch Ryzen with Zen 2 sooner but the big question is how it clocks. They need 5GHz+ not 4GHz, can they achieve that at TSMC?
NeuralNexus - Tuesday, August 28, 2018 - link
What makes you believe that AMD needs to hit 5Ghz on Zen 2 parts when they have the IPC advantage over INTEL right now? INTEL over advantages is frequency, they can't really gain more IPC out of their refined stagnate architecture.HStewart - Tuesday, August 28, 2018 - link
I think they are finding out the same troubles that Intel found out with 10nm - but keep in mind in Intel calculations there 10nm is same as other 7nm. Keep in mind also it does not just the frequency or the number cores that matter, but what is under the hood. I believe with later Intel CPU's they found they can get more performance in same process.enzotiger - Tuesday, August 28, 2018 - link
Any data to support your claim that AMD has better IPC?0ldman79 - Wednesday, August 29, 2018 - link
I try to remain objective externally, but internally I'm an AMD guy, pulling for the underdog.That being said, at best AMD matches Intel's IPC, they don't beat it.
Intel has a few differences in their integer setup and AVX processing that everyone is optimized for, runs right past AMD.
Now if AMD were optimized for it has a more powerful integer setup, though not necessarily faster.
I don't think it is the lithography that is limiting the clock speed. The FX and Sandy Bridge could hit 5GHz on 32nm. I've got an FX that can hit 5GHz any time I feel like it, chip can handle it all day long, though the motherboard can't.
Regardless, lets say that AMD has the IPC edge by 5%, Intel can put out a 5GHz CPU *today* and beat that. Intel actually could make a few changes to their "stagnate" architecture and push more out of it. It isn't stagnant, if you look back at the architecture this one begins with the Pentium Pro and has been refined and added to ever since. Honestly, Nehalem was a big difference from the Core 2, Sandy Bridge was a big departure from Nehalem, Haswell was a rather big jump from Sandy Bridge, though they haven't made huge changes since.
Sandy Bridge had enough changes made it could have been viewed as a different architecture and if you look at the entire line Intel has made huge changes in the Core ix lineup and kept the name. Previously architecture changes of that magnitude have warranted a new name. The PIII to Banias to Dothan realistically were more similar than the Nehalem to Sandy Bridge to Haswell.
Targon - Wednesday, August 29, 2018 - link
This is where third generation Ryzen comes into play. First generation Ryzen to second fixed some memory/cache related performance limitations, but not a lot of big design changes. The rumors are that third generation may provide a 10-15 percent IPC boost to Ryzen. If the jump to 7nm gets AMD to 5GHz, plus that IPC improvement, then AMD may very well have an IPC lead, and not much slower than what Intel chips can hit for clock speed. That is why the jump to 7nm is so important, because it has the potential to put AMD in the lead in single-threaded, as well as multi-threaded workloads.Targon - Wednesday, August 29, 2018 - link
The perception by many people is that clock speed is the one area that allows Intel to dominate in certain areas, such as gaming, and those single threaded benchmarks that don't really matter. If AMD is able to hit 5GHz, any advantage that Intel had in terms of performance would be gone, with the exception where extra memory channels come into play(since Ryzen branded products only have a dual-channel memory controller due to socket AM4).beginner99 - Tuesday, August 28, 2018 - link
Yeah. This killed the dream of a 5ghz zen2 on IBMs HPC process.eastcoast_pete - Tuesday, August 28, 2018 - link
A lot of this doesn't make sense. AMD claims that its 7 nm chips will be fabbed by TSMC; on the other hand, quite reliable industry sources and observers state pretty definitively that at least 75% of TSMC's 7 nm capacity will go to Apple, at least in 2018, and likely forward, with the remainder essentially all for mobile SoCs from Huawei, Qualcomm, plus some cryptomining ASICs. For example, this https://www.eetimes.com/document.asp?doc_id=133361... in the eetimes. So, either TSMC can ramp up its 7 nm fab A LOT, or AMD will only ship a small number of some EPYCs on 7 nm so they can say they shipped 7 nm chips, or they are just giving us story right now. I believe AMD's CPUs and GPUs will be on 12 nm for another two years or more, unless they can make a deal with Samsung, and Samsung gets its 7 nm process geared up for large transistor counts. Right now, I'd need to see an official TSMC announcement of a large volume deal with AMD for 2018/19 before I buy AMD's story, and I am not the only one.vladx - Wednesday, August 29, 2018 - link
Lmao good luck competing with Qualcomm and Huawei, two much richer companies, for fab capacity.eva02langley - Wednesday, August 29, 2018 - link
TSMC doesn't have a lot of X86 chips in the making and especially none of them are on 7nm. It was a clear indicative that they are interested because of their lack of experience in the field.I don't see a huge delay happening anytime soon. Like stated before, AMD was aware of this for a long time and they had plan their root way before EPYC ROME at TSMC.
vladx - Thursday, August 30, 2018 - link
X86 or ARM doesn't matter, both use same wafers to be fabbed. In fact the situation is quite worse for x86 because desktop CPUs have much bigger die size u usually.So yeah, AMD will definitely have problems getting enough fab capacity for their 7nm products especially with needing it for both CPUs and GPUs.
0ldman79 - Wednesday, August 29, 2018 - link
So AMD might need a foundry in the near future and GloFo sounds like it has hit a snag with 7nm, which it's only going to get harder from here on.I thought Global Foundries were already working with IBM on 5nm?
Wouldn't it be crazy if AMD wound up contracting Intel to manufacture 7nm AMD CPUs?
eastcoast_pete - Wednesday, August 29, 2018 - link
I don't think Intel would manufacture AMD's chips. However, I wouldn't be surprised if Intel swoops in, takes the tech, the IP, the equipment (mucho dinero DUV ASML kit!) and key personnel of GloFo's hands for cheap, and gets IBM's blessings (any buyer would need that, IBM owns key IP) in return for manufacturing IBM's mainframe CPUs and other silicon in 7 nm. An Intel acquisition of the tech (or the entire foundry?) would also take care of the Trusted Foundry problem.Targon - Thursday, August 30, 2018 - link
When AMD had their lawsuit with Intel, AMD should have forced Intel to split off their fabs as a part of the settlement. That would have allowed AMD to use the Intel fab process without it being a part of Intel.eastcoast_pete - Wednesday, August 29, 2018 - link
GloFo's announcement was pretty clear that they will, at least for now, seize all development of 7 nm and lower nodes, so that includes 5 nm. I believe that this "we'll make our chips on 7 nm at TSMC" announcement by AMD is a mix of whistling in the dark, trying keep to some analysts happy (for how long?), and might even be somewhat of an attempt to keep AMD's personnel calm. Let's face it: the GloFo announcement was a surprise for AMD (and most everybody else), and AMD really got screwed by GloFo's sudden exit from 7 nm. GloFo was supposed to be their 7 nm volume supplier, but apparently the volume wasn't enough for GloFo to keep plowing billions of $$$ into getting fab 8 up-and-running on 7 nm DUV.