EETimes: TSMC estimates they hold 90% of the world's pending 28nm tape-outs

[Idontcare]

Design starts triple for TSMC at 28-nm

"The smartphone and tablet is the new killer application," said Maria Marced, president of TSMC Europe. "We are seeing a design explosion at 28-nm. We have 89 tape-outs in the pipeline," Marced added. She said that by TSMC's estimate the company currently holds 90 percent of the world's pending tape-outs at 28-nm.

And the company is already shipping 28-nm silicon for some customers, Marced said. "The mobile Internet is demanding the same performance at much lower power." TSMC (Hsinchu, Taiwan) is offering both high-k metal gate (HKMG) and conventional polysilicon processes at 28-nm and has 20-nm production slated for volume production in the second-half of 2012.

Source: http://www.eetimes.com/electronics-news/4215986/Design-starts-triple-for-TSMC-at-28-nm

90% of 28nm tapeouts!?

If this is true then this is nothing short of a total landslide defeat for GloFo at 28nm

 

[Arkadrel]

Holy... wow.. that sounds promising.

I know Intel is abit ahead in the fab bussiness, but I cant help but think that development is getting harder with time as proccesses shrink.

Is it possible at some point TSMC or GF could catch up?

That alliance thingy where they split research costs, has to be easier on them all, than Intel carrying the intire burden themselfs.

 

[mnewsham]

Wow I could see 60-70% maybe, but 90%? GloFo dropped the ball, anyone know how thier 28nm is coming along (if at all)

 

[Tuna-Fish]

90% of 28nm tapeouts!?

If this is true then this is nothing short of a total landslide defeat for GloFo at 28nm

Well, tapeouts still only measure designs, not volume. A single arm soc that's a hit for smartphones would likely be able to take all of GloFo:s capacity for the time being. Still, why would so few order chips from them?

I know Intel is abit ahead in the fab bussiness, but I cant help but think that development is getting harder with time as proccesses shrink.

Is it possible at some point TSMC or GF could catch up?

That alliance thingy where they split research costs, has to be easier on them all, than Intel carrying the intire burden themselfs.

For now, Intel is actually gaining more ground. I don't expect that to change any time soon.

 

[Phynaz]

Read this earlier, and my first thought was "What happened to UMC?".

 

[exar333]

Wow...that is pretty amazing. Does this mean GloFo just gets the scraps leftover?

 

[Idontcare]

Well, tapeouts still only measure designs, not volume. A single arm soc that's a hit for smartphones would likely be able to take all of GloFo:s capacity for the time being. Still, why would so few order chips from them?

You miss the point of why the stat is relevant to track.

It speaks to the confidence, or lack thereof, across the global customer base that GloFo's 28nm foundry model is viable, which has the knock-on effects of future business on the nodes to come.

GloFo needed 28nm to be a mindshare victory. If they are not gaining traction in the foundry space then potential would-be customers are going to rightly start to question just how long they can be expected to survive as a foundry.

Understand that the entire premise of the foundry business, from the perspective of the fabless company, is to outsource the risk associated with production.

Trading the risk associated with production timelines for the risk of not being confident the foundry will deliver or has the endurance to stay in the industry (are they a flash in the pan or a foundry you can count on for 10+ yrs?) is not an acceptable trade-off.

As they say in IT: "no one ever got fired for buying Intel", so is true for fabless companies "no one ever got fired for going TSMC".

GloFo is unproven and an unknown (not counting the legacy accounts they acquired with Chartered and AMD)...that is not an equation that speaks to risk-reduction for a fabless company.

 

[jimbo75]

90% of 28nm tapeouts!?

If this is true then this is nothing short of a total landslide defeat for GloFo at 28nm

Not quite. GF has 60k wpm at 28nm and will sell every wafer.

What it means is AMD get's a ton of extra wafers for 3-4 tapouts, 2 apu's and 2 gpu's. Nvidia is in a heap of trouble and bobcat v2 will have a whole year with an enormous advantage until intel gets atom at 22nm out in 2013.

 

[jimbo75]

You miss the point of why the stat is relevant to track.

It speaks to the confidence, or lack thereof, across the global customer base that GloFo's 28nm foundry model is viable, which has the knock-on effects of future business on the nodes to come.

No it speaks volumes for companies sticking to what they know against the logical decision to move to a better process.

The real deal is, GF's 28nm will spank TSMC's so hard it will look like it's a whole node ahead. The players who moved to GF will have a huge advantage in the market, especially AMD's graphics which will utterly crush Nvidia's (even more I mean) at 28nm.

 

[Martimus]

Near monopolies like this always depress me. I don't like that TSMC is so dominant, and I hope that another foundry is able to compete with them in the high performance market. If GF doesn't succeed in gaining traction there, I can only hope that Intel enters the open foundry business to compete.

IDC, that was a very good post, and something I didn't consider. It could just be a short term problem as potential customers get more comfortable with GF's long term viability. At the same time, there must be some other causes of trepidation if TSMC really has at least NINE TIMES as many customers lined up for the 28nm process. From what I have seen, both are planning on production in this node around the same time. (Although I think TSMC is going gatelast HKMG versus gatefirst HKMG for GF, which would make TSMC more desireable for performance in non-SOI production.)

The other thing that gets me is that I had thought GF had made a big push to sign up all the ARM producers, but the article makes it sound like nearly all ARM production at 28nm will be done at TSMC.

 

[Idontcare]

Read this earlier, and my first thought was "What happened to UMC?".

http://www.eetimes.com/electronics-news/4212366/Samsung-lags-in-foundry-rankings-

UMC is to TSMC as AMD used to be to Intel.

Almost, theirs was a 1:6 ratio whereas UMC and TSMC is more of a 1:4 ratio.

 

[Idontcare]

<- personally worked with TSMC, UMC, Chartered, and SMIC, as well as having many contacts (past co-workers) who work at both GloFo and IBM...

Not quite. GF has 60k wpm at 28nm and will sell every wafer.

What it means is AMD get's a ton of extra wafers for 3-4 tapouts, 2 apu's and 2 gpu's. Nvidia is in a heap of trouble and bobcat v2 will have a whole year with an enormous advantage until intel gets atom at 22nm out in 2013.

No it speaks volumes for companies sticking to what they know against the logical decision to move to a better process.

The real deal is, GF's 28nm will spank TSMC's so hard it will look like it's a whole node ahead. The players who moved to GF will have a huge advantage in the market, especially AMD's graphics which will utterly crush Nvidia's (even more I mean) at 28nm.

^ gobs of meaningless gibberish that has no basis in reality, but its fun to dream I suppose.

 

[Idontcare]

IDC, that was a very good post, and something I didn't consider. It could just be a short term problem as potential customers get more comfortable with GF's long term viability. At the same time, there must be some other causes of trepidation if TSMC really has at least NINE TIMES as many customers lined up for the 28nm process. From what I have seen, both are planning on production in this node around the same time. (Although I think TSMC is going gatelast HKMG versus gatefirst HKMG for GF, which would make TSMC more desireable for performance in non-SOI production.)

Similar happened with IBM when they threw their hat into the foundry market as well. Everyone sat on the fence waiting to see if avoiding IBM was less risky than going with them.

The few fabless guys that went with IBM (Cyrix, Xilinx, ATI) quickly and publicly regretted/lamented the decision. They lost revenue, elevated risk, and suffered for it.

GloFo may well have the world's best 28nm process to grace the planet, but being fabless isn't about being super-dependent on your foundry, it is the opposite. You have already decided your business model is about differentiating your products at the design/feature level and not at the process-tech level.

Foundries compete with each other on the basis of production cost and timeline to volume capacity for your product. It is not really about being 10&#37; faster in the xtor dept or having 0.2 lower k-value in the BEOL dielectric.

The vast majority of fabless companies are not interested in relying on the foundry to provide the market differentiation of their product at the consumer level. Companies that do tend to be companies that keep their fabs so they have control over this for all the right reasons.

That is what outsourcing is all about, lowering cost and risk. If you don't manage your foundry then you haven't lowered risk and that is just as bad as not managing the cost.

 

[jimbo75]

<- personally worked with TSMC, UMC, Chartered, and SMIC, as well as having many contacts (past co-workers) who work at both GloFo and IBM...

And yet you actually believe GF has only 8 28nm tapeouts? Curious indeed.

^ gobs of meaningless gibberish that has no basis in reality, but its fun to dream I suppose.

You actually think GF won't sell every wafer? Or maybe you think those tapeouts in AMD's last CC were at TSMC instead of GF? I'd like you to elaborate on exactly how this is "gibberish", I really would.

 

[OCGuy]

No it speaks volumes for companies sticking to what they know against the logical decision to move to a better process.

The real deal is, GF's 28nm will spank TSMC's so hard it will look like it's a whole node ahead. The players who moved to GF will have a huge advantage in the market, especially AMD's graphics which will utterly crush Nvidia's (even more I mean) at 28nm.

Not quite. GF has 60k wpm at 28nm and will sell every wafer.
What it means is AMD get's a ton of extra wafers for 3-4 tapouts, 2 apu's and 2 gpu's. Nvidia is in a heap of trouble and bobcat v2 will have a whole year with an enormous advantage until intel gets atom at 22nm out in 2013.

Why do you have to spout AMD fanboi gibberish in every post? Sheesh....we get it...



I thought I had heard that even nV may have been looking at GF due to TSMC's 32nm debacle. Maybe this is proving that rumour to be false.

Either way, this is impressive, and shows industry confidence in TSMC.


Edit: Actually I guess this could just be showing lack of confidence in the other players, not necessarily a boat load of confidence in TSMC.

 

[Martimus]

You actually think GF won't sell every wafer? Or maybe you think those tapeouts in AMD's last CC were at TSMC instead of GF? I'd like you to elaborate on exactly how this is "gibberish", I really would.

What 28nm tapeouts did AMD show at their last conference call from Global Foundries? Did they show the 28nm Bobcat? Maybe the next Radeon?

I would be surprised if AMD swapped from TSMC to GF for 28nm on the Bobcat shrink, but I could see them hedging their graphics by producing some dies at both foundries.

Do you realize that Llano and Bulldozer are both taped out on the 32nm SOI process which is not what this article is talking about? This article is about the 28nm bulk process (inclusive of both HKMG high performance nodes, and the more conventional non-HKMG node).

 

[jimbo75]

Why do you have to spout AMD fanboi gibberish in every post? Sheesh....we get it...

Because somebody has to counter the intel fanboy crap that is so prevalent in this forum, obviously.

I thought I had heard that even nV may have been looking at GF due to TSMC's 32nm debacle. Maybe this is proving that rumour to be false.

Either way, this is impressive, and shows industry confidence in TSMC.

Why would ANYBODY have confidence in TSMC? Take a look at older roadmaps when 28nm was already out 6 months ago according to them. TSMC is full of it.

Edit: Actually I guess this could just be showing lack of confidence in the other players, not necessarily a boat load of confidence in TSMC.

Nobody in the entire industry has any confidence in TSMC. However, GF needs to PROVE they can do it first and IDC knows this, which is why this entire thread is just more intel fanboy crap.

 

[busydude]

<Gibberish>

Hey man.. can you please stop posting? When did posting about TSMC = Inter fanboy crap? Please, this thread had some useful information that people, like me, can learn and understand.. stop your nonsense.

 

[jimbo75]

What 28nm tapeouts did AMD show at their last conference call from Global Foundries? Did they show the 28nm Bobcat? Maybe the next Radeon?

Definitely the 28nm bobcat and probably the next radeon, but I don't know which end of the market.

I would be surprised if AMD swapped from TSMC to GF for 28nm on the Bobcat shrink, but I could see them hedging their graphics by producing some dies at both foundries.

Do you realize that Llano and Bulldozer are both taped out on the 32nm SOI process which is not what this article is talking about? This article is about the 28nm bulk process (inclusive of both HKMG high performance nodes, and the more conventional non-HKMG node).

Of course, Bobcat 2 is a shrink from 40nm to 28nm, but it's at GF not TSMC.

Also, Nvidia has not taped out anything on 28nm yet (at least not according to their last CC http://seekingalpha.com/article/269...sses-q1-2012-results-earnings-call-transcript).

 

[jimbo75]

Hey man.. can you please stop posting? When did posting about TSMC = Inter fanboy crap? Please, this thread had some useful information that people, like me, can learn and understand.. stop your nonsense.

Can you please stop posting? If you believe any of this nonsense your head must zip up the back.

I'm amazed how fast you are all scrambling to suddenly believe TSMC is on track with amazing 28nm performance. Have you forgotten their 40nm fiasco already?

 

[Martimus]

Nobody in the entire industry has any confidence in TSMC. However, GF needs to PROVE they can do it first and IDC knows this, which is why this entire thread is just more intel fanboy crap.

Whoa! This thread is about TSMC. I don't see how you can construe any of it to be about "Intel Fanboy crap". What are you talking about? How does Intel have anything to do with this thread? Also what does nVidia have anything to do with this thread either? You really lost me there.

Definitely the 28nm bobcat and probably the next radeon, but I don't know which end of the market.

Can you by any chance show me a link to where AMD says they are moving the 28nm Bobcat variant to GF? I don't think that is out of the realm of possibility, but I thought I read a quote from AMD that they were going to produce the 28nm variant at TSMC. I don't have that handy, but I am hoping that you have your quote handy so you can set my thoughts on the matter straight.

 

[piesquared]

Can you please stop posting? If you believe any of this nonsense your head must zip up the back.

I'm amazed how fast you are all scrambling to suddenly believe TSMC is on track with amazing 28nm performance. Have you forgotten their 40nm fiasco already?

Don't forget how they screwed customers over by cancelling 32nm mid stream. Surely that generated huge confidence to customers. No doubt certain members thought of that, just decided not to mention it...

 

[busydude]

Don't forget how they screwed customers over by cancelling 32nm mid stream.

Do you realize that 32nm/28nm are basically the same? They are just marketing it differently this time.

 

[jimbo75]

Whoa! This thread is about TSMC. I don't see how you can construe any of it to be about "Intel Fanboy crap". What are you talking about? How does Intel have anything to do with this thread? Also what does nVidia have anything to do with this thread either? You really lost me there.

I think it's fairly obvious that certain posters have an interest in downplaying GF and AMD at every opportunity. I'm sure it's more than just me who can figure that out as well.

I merely pointed out the counter to IDC's claim that this was, and I quote

If this is true then this is nothing short of a total landslide defeat for GloFo at 28nm

and

It speaks to the confidence, or lack thereof, across the global customer base that GloFo's 28nm foundry model is viable

Both of which are nonsense. The fact is, who would risk their business by switching to GF until they have proven they are capable?

BAD though TSMC is, it is still "better the devil you know". IDC knows this, yet would have us believe that this is all down to GF doing it wrong somehow.

Can you by any chance show me a link to where AMD says they are moving the 28nm Bobcat variant to GF? I don't think that is out of the realm of possibility, but I thought I read a quote from AMD that they were going to produce the 28nm variant at TSMC. I don't have that handy, but I am hoping that you have your quote handy so you can set my thoughts on the matter straight.

AMD stated already that they are dual sourcing at 28nm. Charlie let it out the bag (intentionally or not I'm not sure) that Wichita taped out at GF. You can believe that or not, with his track record I choose to believe it.

 

[OCGuy]

Don't forget how they screwed customers over by cancelling 32nm mid stream. Surely that generated huge confidence to customers. No doubt certain members thought of that, just decided not to mention it...

Yes, it's all part of a huge conspiracy that nobody mentioned it in this thread...wait a second:

I thought I had heard that even nV may have been looking at GF due to TSMC's 32nm debacle

Why is it that the people with the low post counts are the ones derailing CPU threads...