News AMD Reportedly Eyes Samsung Chip Fab Deal in Dual-Source Strategy

[Admin]

AMD set to have three manufacturing partners: TSMC, GlobalFoundries, and Samsung Foundry.

AMD Reportedly Eyes Samsung Chip Fab Deal in Dual-Source Strategy : Read more

 

[-Fran-]

If AMD can strike a deal and source RAM modules so they have a good discount for GPUs and regular RAM, that would help them a lot, I'd say.

Specially on the HBM side for their Data Center stuff and, maybe, consumer 😀

One can dream 😛

Regards.

 

[TechieTwo]

It would be smart business practices especially with China's instability.

 

[ottonis]

I fully agree with the author. Diversification is mandatory in order to stay safe. Even if it's quite costly, collaborating with Samsung alongside TSMC is a smart and and anticipatory move by AMD not only in case something happens to TSMC but also in order to optimize future products from both sides of the production chain.

 

[TechieTwo]

Its USA itself that cause all this chaos and instability when they know they can't compete then they switch to blockage. Whole Asia is quite smart that we distance ourself from the West and time to put greenback behind also

The use of slave labor, child labor and the murder of protesters by the Chinese Communist Party are good reasons for the world to not conduct business with and to sanction the CCP.

 

[bit_user]

Its USA itself that cause all this chaos and instability when they know they can't compete then they switch to blockage.

This explanation doesn't make sense, when you consider that neither Taiwan nor South Korea are being sanctioned, yet both have more advanced semiconductor technology than the US, at present. Furthermore, China does not have the most advanced semiconductors.

So, perhaps you need to try a little harder to find the reason.

 

[Deleted member 1353997]

This explanation doesn't make sense, when you consider that neither Taiwan nor South Korea are being sanctioned, yet both have more advanced semiconductor technology than the US, at present. Furthermore, China does not have the most advanced semiconductors.

So, perhaps you need to try a little harder to find the reason.

That's a non-sequitur. Nobody claimed that China has the most advanced semiconductors.

The original claim was that one shouldn't conduct business with China due to its instability.
The second claim was that China's instability was caused by the USA and their sanctions.

So, perhaps you need to try a little harder to refute that claim.

 

[bit_user]

That's a non-sequitur. Nobody claimed that China has the most advanced semiconductors.

The phrase "can't compete" is open to interpretation.

The original claim was that one shouldn't conduct business with China due to its instability.
The second claim was that China's instability was caused by the USA and their sanctions.

So, perhaps you need to try a little harder to refute that claim.

My response wasn't addressed at claim 1, obviously. I was clearly questioning the rationale put forth as part of claim 2, which you omitted from your summary.

Moreover, I cited counter-examples where the US hasn't used sanctions against trading partners that are out-competing it. That explanation was clearly too simplistic, yet I'm the one you take issue with?

 

[Co BIY]

I fully agree with the author. Diversification is mandatory in order to stay safe. Even if it's quite costly, collaborating with Samsung alongside TSMC is a smart and and anticipatory move by AMD not only in case something happens to TSMC but also in order to optimize future products from both sides of the production chain.

Outside of the political, security and human rights issues - which cannot be discussed here without the comments sections getting shut down ...

Getting three bids for any contract is always a good idea. It follows that it benefits purchasers to ensure that there are at least three providers to deal with.

 

[bit_user]

Getting three bids for any contract is always a good idea. It follows that it benefits purchasers to ensure that there are at least three providers to deal with.

As the article mentions, the costs of migrating a design from one fab to another are simply massive. That's because each fab uses their own manufacturing process and has their own cell libraries tuned for it. Apparently, these cell libraries aren't just mix-n-match, plus you then have to redo layout. You might think of it like porting an operating system from one ISA (like x86) to another (like ARM). For that reason, chip designers typically chose a fab rather early in the design process and have to live with that decision.

Lately, there's a trend towards using AI to automate at least the layout portion of chip design. So, that could not only be a huge cost & time-saver, but might also make it easier and cheaper to switch fabs.

Cadence Cerebrus to Enable Chip Design with ML: PPA Optimization in Hours, not Months

www.anandtech.com

Best of all, AI-driven chip layout seems to offer nearly a full node worth of improvement, compared with human-driven layout. For all we know, AMD is in fact using this move to trial these techniques and technology.

 

[Deleted member 1353997]

The phrase "can't compete" is open to interpretation.

Hence the "non-sequitur". You jumped to a conclusion that didn't follow from the conversation.

My response wasn't addressed at claim 1, obviously. I was clearly questioning the rationale put forth as part of claim 2, which you omitted from your summary.

Sorry, but where in my post did I claim you were contesting the first claim rather than the second one? As you said, you obviously contested the second claim.

And how exactly did I "omit claim 2 from the summary", when it is literally the second point of my summary?

Moreover, I cited counter-examples where the US hasn't used sanctions against trading partners that are out-competing it. That explanation was clearly too simplistic, yet I'm the one you take issue with?

And how are those counter-examples relevant to the claim you're trying to contest? Just because the USA didn't sanction some countries, it doesn't mean that China's instability is not caused by US sanctions. That's why I called it a non-sequitur.

But if you so badly want someone to contest your own argument, regardless of whether it is relevant to the discussion or not: China is an economic superpower, like it or not. It is the second-largest one, making up 18.45% of the global economy with a growth of 8.1%, behind the USA's 23.93% and growth of 5.7%. South Korea is only the 10th largest economy and its growth is still behind the USA with 4%. Taiwan isn't even in the top 25. Why would the USA sanction trading partners that aren't even close to threatening its standing as global superpower, when China is "projected to surpass the American economy in the coming few years"?

And just so there is no misunderstanding between us: I am not claiming that the USA is sanctioning China to maintain its lead as an economic superpower. I am claiming that the USA has more to gain from sanctioning China than from sanctioning South Korea or Taiwan (or both). And that's even without mentioning Capitalism vs Communism.

If you want to contest that claim, you'll have to show a valid reason why the USA would prefer to sanction South Korea or Taiwan over China.

 

[bit_user]

Hence the "non-sequitur". You jumped to a conclusion that didn't follow from the conversation.

I was responding to a very specific claim that you omitted. I'll repeat it, so that you cannot pretend not to have seen it.

​

when they know they can't compete then they switch to blockage.

​

You didn't address this, at all. Your reply is therefore a non-sequitur. It's as if you were trying to derail the discussion from focusing on the most inflammatory and problematic allegation that I was trying to address.

But if you so badly want someone to contest your own argument, regardless of whether it is relevant to the discussion or not: <snip/>

You could say those same things about the EU.

These read like the arguments of a conspiracy theorist. For them, it's sufficient merely to show a motive. None of this amounts to evidence the sanctions were taken for anti-competitive reasons.

I am claiming that the USA has more to gain from sanctioning China than from sanctioning South Korea or Taiwan (or both).

A point no one contested. So, why? It doesn't advance the discussion.

 

[Deleted member 1353997]

I was responding to a very specific claim that you omitted. I'll repeat it, so that you cannot pretend not to have seen it.

when they know they can't compete then they switch to blockage.

You didn't address this, at all. Your reply is therefore a non-sequitur. It's as if you were trying to derail the discussion from focusing on the most inflammatory and problematic allegation that I was trying to address.

Ironically, your "very specific claim" can be split into 3 very different distinct claims. Since I still have no clue what you're referring to, I'm not going to assume any further:

1. The USA can't compete with China.
   
   That part, I actually did address, believe it or not:
   

   China is "projected to surpass the American economy in the coming few years"

   
   
2. The USA switched to blockade.
   
   Assuming "blockade" refers to sanctions, that actually did happen. I trust I don't need to provide sources for this specific claim, as there's plenty of articles on Tom's regarding that subject.
   
   
3. The USA sanctions countries for anti-competitive reasons.
   
   Admittedly, I didn't address that one. Why should I? This is just one person's opinion. They didn't provide any proof nor reasoning, so there's nothing to argue against. You, on the other hand, did provide a counter-argument, albeit a fallacious one. I called you out on it.

You could say those same things about the EU.

Maybe I could, so what? The EU isn't sanctioning China, so China's instability clearly isn't because of EU sanctions. And since the EU isn't sanctioning China, it clearly can't be doing so out of anti-competitive reasons (since it's not doing it, duh).

Besides, the EU and the USA are two completely different and independent entities. What one does has no bearing on the other. Just because the UK left the EU doesn't mean the USA is now left with only 49 states, for instance.

So yes, I could say the same about the EU, but why? It doesn't advance the discussion.

I am claiming that the USA has more to gain from sanctioning China than from sanctioning South Korea or Taiwan (or both).

A point no one contested. So, why? It doesn't advance the discussion.

It may not advance the discussion, but it helps reduce the risk of someone derailing it even further. There's a reason why I prefaced that paragraph with: "just so there is no misunderstanding between us".

 

[bit_user]

1. The USA can't compete with China.
   
   That part, I actually did address, believe it or not:

Actually, no. Simply being able to compete isn't equivalent with achieving or maintaining dominance. For instance, although South Korea is a smaller country, they've shown an ability to be competitive in sectors like semiconductors, industrial, electronics, and automotive.

2. The USA switched to blockade.

Assuming "blockade" refers to sanctions, that actually did happen.

No, the statement embeds an assumption that sanctions were adopted as an alternative to ordinary competition.

I trust I don't need to provide sources for this specific claim, as there's plenty of articles on Tom's regarding that subject.

The mere fact of sanctions aren't in question. It's the motive & objective that are at issue. You're obviously smart enough to understand that.

3. The USA sanctions countries for anti-competitive reasons.

Admittedly, I didn't address that one. Why should I? This is just one person's opinion. They didn't provide any proof nor reasoning, so there's nothing to argue against.

It's a statement of fact, which can be questioned and should be supported, if true. Given the gravity of the accusation, it warrants strong evidence, none of which was provided.

You, on the other hand, did provide a counter-argument, albeit a fallacious one.

No, my counter argument simply demonstrated instances in which the USA has been out-competed and did not resort to sanctions. A literal interpretation of the statement to which I replied would indicate that the USA always resorts to "blockage" when it can't compete. To show that's not literally true, only one counterexample is needed. In fact, I can't think of a single example that would support the accusation.

I called you out on it.

Again, why? You're bike shedding. Classic misdirection, to distract from the bigger issue.

Maybe I could, so what? The EU isn't sanctioning China,

No, the point I was trying to make is that the EU is a bigger economy than the USA, yet the US hasn't tried to sabotage its economy.

It may not advance the discussion, but it helps reduce the risk of someone derailing it even further.

I'm not sure why you're so intent on derailing discussion of the part I actually quoted and responded to. However, I'm now seriously doubting you're acting in good faith.

 

[Co BIY]

As the article mentions, the costs of migrating a design from one fab to another are simply massive. That's because each fab uses their own manufacturing process and has their own cell libraries tuned for it.

I think the success of the foundry model has allowed for the increasing success of standards in the ADT (Automated Design Tools) that might make the fab swapping easier.

The Tools may increasingly allow the same design to be ported to different processes pretty easily.

Intel did that for the Skylake processors. Designed for Intel 7 and back ported to 14nm.

 

[bit_user]

Intel did that for the Skylake processors. Designed for Intel 7 and back ported to 14nm.

Huh? No, I think you mean Rocket Lake was essentially Ice Lake, backported from 10 nm+ to 14 nm++++.

That was also Intel porting between their own nodes, which means their toolchains, libraries, and design flow were probably a heck of a lot more similar than you'd have between different foundries. And even then, it probably took Intel 18-24 months to bring it to market.

 

[Co BIY]

Huh? No, I think you mean Rocket Lake was essentially Ice Lake, backported from 10 nm+ to 14 nm++++.

That was also Intel porting between their own nodes, which means their toolchains, libraries, and design flow were probably a heck of a lot more similar than you'd have between different foundries. And even then, it probably took Intel 18-24 months to bring it to market.

You are right that it was Rocket Lake. But I think it shows what can be done especially given that those Intel CPUs are among the most complicated of products.

All the Foundries (Global, Samsung, Intel, TSMC) will be very motivated to make this happen because of the competition pressure. But I think you end up with more standardization and more flexibility for customers as a result.

 

[bit_user]

When it comes to China with Huawei dominance in tech areas. USA sees that China will sure one day surpass them in tech area one day thus now they resort to such distasteful tactics.

Again, look at the EU as a counterexample. Its economy is bigger than the US, but the US actually helped it quite a bit instead of trying to undermine it. Similarly, the US is now trying trying to help Africa, as well.

You're making very sweeping claims, without any evidence. You need to do much more than simply show a motive.

 

[bit_user]

You are right that it was Rocket Lake. But I think it shows what can be done especially given that those Intel CPUs are among the most complicated of products.

I'm sure the cost was easily in the hundreds of $M, like mentioned in the article. We're a long way from chip designers being able to accept late-stage bids from fabs.

All the Foundries (Global, Samsung, Intel, TSMC) will be very motivated to make this happen because of the competition pressure.

No, the lagging fabs are incentivized to make it easier to steal business from the leading ones. And GloFo is no longer in competition below 12 nm, as they dropped out of the 7 nm race and opted instead to focus on nodes for specialized products, like RF chips.

What we might see is Samsung and Intel try to make it easier for customers to port their designs from TSMC nodes. Intel has traditionally used their own in-house EDA tools, but I think they're moving to a more industry standard flow, for reasons like this.

 

[Co BIY]

I'm sure the cost was easily in the hundreds of $M, like mentioned in the article. We're a long way from chip designers being able to accept late-stage bids from fabs.


No, the lagging fabs are incentivized to make it easier to steal business from the leading ones. And GloFo is no longer in competition below 12 nm, as they dropped out of the 7 nm race and opted instead to focus on nodes for specialized products, like RF chips.

What we might see is Samsung and Intel try to make it easier for customers to port their designs from TSMC nodes. Intel has traditionally used their own in-house EDA tools, but I think they're moving to a more industry standard flow, for reasons like this.

The move to Chiplets may allow many parts of the formerly single-process integrated chips to move back "downstream" where there is more competition.

If the current leaders hit limits (and those are more likely to be economic than strictly physical) the current laggards could be back relatively quickly. Or well-heeled non-players could move in like Google , Apple , or Alibaba.