[Digitimes] Kaveri delayed

[frozentundra123456]

Pretty sure that's nothing like what I said.



What's the point in a company that sits on 5.5 billion while the other companies in competition with it race ahead? That's what Nvidia is doing, that's why they can't compete with the other ARM guys.



AMD is a technology company not a company that exists to make shareholders money. The lack of dividend payments should be enough to keep that kind of investor away, so if anyone was expecting anything more then they should have done their homework a bit better.

AMD wouldnt have been sitting on 5 billion, because they did not have it. They went deeply into debt to buy ATI, putting them in desperate financial straights. That is probably why they negotiated such an onerous contract with GF. You never get a good deal when you negotiate from a desperate situation.

And even after all that they did not exactly "race ahead". Llano was late, and the cpu portion was very mediocre. I remember looking forward very eagerly to Llano, and when it came out, I said "What, just the same old cpu with a graphics chip stuck on that is still very mediocre for gaming." Intel was actually first to market with an igp, although admittedly not up to the level of Llano, but good enough for most users to dispense with a discrete card.

 

[Vesku]

That's how the business works. If AMD had booked a years wafers from TSMC, then cancelled them for any reason they'd still have to pay whatever get-out clause was in the contract. You can't expect GF to set aside the wafers for AMD then foot the bill when AMD decides they don't want them.

AMD could have met their wafer obligations and dropped prices of the chips instead, but they opted to save what, $50 million or so instead? And with the drop in wafers they've been able to keep prices of APU's stable.

I just don't get why GF is being made out as the bad guys here when all they've done is stick to the contract and shown a lot of willingness to change it to AMD's benefit.

Those running AMD at the time are just as responsible but those financially involved on the GF side are in as deep, having a significant investment in AMD itself and appointing Ruiz Chairman of the spin off. WSA negotiated between a company and the spin off that had the ex AMD CEO as Chairman, not even revealed until it started costing AMD quite a bit of money. The companies concealed from the public the full entanglements of the WSA as long as possible.

And even after all that they did not exactly "race ahead". Llano was late, and the cpu portion was very mediocre. I remember looking forward very eagerly to Llano, and when it came out, I said "What, just the same old cpu with a graphics chip stuck on that is still very mediocre for gaming." Intel was actually first to market with an igp, although admittedly not up to the level of Llano, but good enough for most users to dispense with a discrete card.

Llano was actually a big improvement in mobile for AMD. Unfortunately companies just put them in the same designs they had for the hotter and more power hungry previous Turion and Phenom line up. It's been pretty obvious that AMD has almost 0 influence when it comes to the OEMs that design the end products. I actually put off replacing my notebook for Llano however the selection was poor. What sort of business decisions result in having a 2 or 3 Intel Core i notebooks from each major brand that fit my requirements but not a single Llano version? It's a bit frustrating from a consumer perspective and the trend so far continues with a dozen 14 inch Intel models in the under $550 category on Newegg but only a single AMD model.

Even if Kaveri shows up in retail before the end of the year will it end up in any decent notebooks?

 

[monstercameron]

you guys are arguing far-reaching hypotheticals.

 

[mrmt]

You don't know what kind of prices AMD is getting at GF. All we know about it is that GF has shown every willingness to amend the WSA when it was beneficial to AMD to do so, and will be amending it again, no doubt to reflect AMD's inability to pay $1.5 billion per year on wafers. I'm pretty sure GF doesn't have to do this, but they are because they realise that AMD is an important partner.

The fact thar AMD was trying to themselves from glf says a lot about hw happy they are with their foundry partner.

Again, I just don't see how this is a bad deal for AMD, because they can't tear up their contract and act like they never spun off the fabs? What would GF be getting out of that?

TSMC doesn't have any capacity, they never do. Hardly anyone does and it won't be long before Samsung has theirs sold again as well. Wafer starts are at a premium.

AMD made the choice to scale back on production - they could have kept production at maximum and dropped prices but instead they decided to scale back and cut costs. That was their choice - they aren't being held to ransom by GF they are simply paying the agreed cost of cancelling some wafers. Should GF pay for AMD's unwillingness to stick to the agreed wafer starts?

TSMC is bringing Apple on board but cannot bring AMD? Really?

 

[SiliconWars]

TSMC is bringing Apple on board but cannot bring AMD? Really?

How much do you think Apple is paying for this? Or do you think TSMC is doing it out of the goodness of their heart?

 

[mrmt]

How much do you think Apple is paying for this? Or do you think TSMC is doing it out of the goodness of their heart?

The capacity is there. Just look at where nvidia and qualcomm searched for more capacity when they needed. Same for amd. When glf screwed up their 28nm process, they again turned to tsmc to field jaguar on tsmc.

 

[podspi]

AMD is a technology company not a company that exists to make shareholders money. The lack of dividend payments should be enough to keep that kind of investor away, so if anyone was expecting anything more then they should have done their homework a bit better.

That is the point of EVERY computer. Technology companies create new technologies to make the owners of the company (the shareholders) money. That's the point.

It has been clear for quite some time that the purchase of ATI was a huge mistake for AMD. Either they should have merged with NVDA (as more or less equals), or they should have just licensed GPU IP, something that is done all the time when producing SoCs...

 

[SiliconWars]

Basically what most of you are saying is that AMD should get first dibs on GF's cutting-edge capacity, but only so long as they want it. If they didn't want it, GF would foot the bill.

Try to imagine what kind of poker face they'd need to pull that off at the discussion table. What is in this for GF? There were other companies out there, Chartered for one, UMC, SMIC...all of these would have been viable targets for the initial buyout. Do you really think they were so desperate for Dresden that they'd have agreed to basically sponsor AMD with free wafers at AMD's whim? It was AMD who went to ATIC btw - not the other way around.

 

[SiliconWars]

The capacity is there. Just look at where nvidia and qualcomm searched for more capacity when they needed. Same for amd. When glf screwed up their 28nm process, they again turned to tsmc to field jaguar on tsmc.

Where did Nvidia and Qualcomm get more capacity? If you mean the few thousand extra wafers that TSMC got for them then that is nowhere near what AMD would have needed. AMD would need 60K+ 32nm SOI wpm, all 3 of which TSMC could never provide in a viable timeframe.

 

[mrmt]

Where did Nvidia and Qualcomm get more capacity? If you mean the few thousand extra wafers that TSMC got for them then that is nowhere near what AMD would have needed. AMD would need 60K+ 32nm SOI wpm, all 3 of which TSMC could never provide in a viable timeframe.

What about ibm for SOI?

 

[Vesku]

Basically what most of you are saying is that AMD should get first dibs on GF's cutting-edge capacity, but only so long as they want it. If they didn't want it, GF would foot the bill.

Try to imagine what kind of poker face they'd need to pull that off at the discussion table. What is in this for GF? There were other companies out there, Chartered for one, UMC, SMIC...all of these would have been viable targets for the initial buyout. Do you really think they were so desperate for Dresden that they'd have agreed to basically sponsor AMD with free wafers at AMD's whim? It was AMD who went to ATIC btw - not the other way around.

There was no 28nm production to call dibs on until this year despite Globalfoundries claiming it's been available since 2011 via press releases: http://www.globalfoundries.com/newsletter/28nm_frontier.aspx

Yet the first GF 28nm LP chips are just now showing up: http://www.globalfoundries.com/newsroom/2013/20130617.aspx

Were those press releases made so they could hold AMD's feet to the fire that is the WSA? "Pay us to not push through 28nm wafers with no working dies on them"?

 

[rgallant]

just asking
-if gf gets up to speed on it's 28nm , could amd shift the console chips over to gf , if they do any guess on the wafer volume to fulfill the WSA.
-would it be enough so amd could move some higher tech. chips to 20nm TSMC in 2014 ?

 

[SiliconWars]

What about ibm for SOI?

Ok first of all, why would AMD help create GF, then ditch them for the same process except at IBM, which would almost certainly be more expensive?

Then you have the fact that GF are actually in the fab business while IBM has been scaling down their manufacturing for years. Who would you go with if your business depended on it?

All these ideas sound great until you realise that AMD basically had their ex fab, their ex fab guys and first dibs on their cutting-edge process.

*Why* would they say no to that? It was simply creating a new entity that provided them with everything they already had, except on a client basis. You heard what JHH said about "it's AMD's fab"?

Would you be saying this about GF if they hadn't screwed up 32nm or were so late with 28nm? No you wouldn't. That means you are looking at unforeseeable technical problems as the reason for being bad business decisions. The business decision was the only one that made sense, the technical problems (and the collapsing PC market) are what has made it look bad for AMD now - none of those were obvious at the time the deal was struck.

What if GF had provided a superb 32nm, 28nm and was racing towards 22nm in great shape? It would have been the exact same business decision, but now it would have been a great one. This is why the argument that it was a bad business decision doesn't make sense.

 

[SiliconWars]

just asking
-if gf gets up to speed on it's 28nm , could amd shift the console chips over to gf , if they do any guess on the wafer volume to fulfill the WSA.

Yes Jaguar was designed for this kind of portability. It'll be on GF's 28nm as soon as it's ready, if it isn't already.

-would it be enough so amd could move some higher tech. chips to 20nm TSMC in 2014 ?

TSMC will have so much demand for 20nm in 2014 it's doubtful if AMD will even have a chip to release on 20nm next year. Even if they did, Jaguar and discrete graphics would be far ahead of any high-end CPU.

 

[rgallant]

Yes Jaguar was designed for this kind of portability. It'll be on GF's 28nm as soon as it's ready, if it isn't already.



TSMC will have so much demand for 20nm in 2014 it's doubtful if AMD will even have a chip to release on 20nm next year. Even if they did, Jaguar and discrete graphics would be far ahead of any high-end CPU.

ok thanks

 

[Vesku]

What if GF had provided a superb 32nm, 28nm and was racing towards 22nm in great shape? It would have been the exact same business decision, but now it would have been a great one. This is why the argument that it was a bad business decision doesn't make sense.

No. Becoming a fabless company by signing an agreement to be bound tightly, no agreed wafer pricing framework and exclusive 28nm agreement that had to be bought out of at a total cost of ~$700 million, to the fab you are spinning off is a bad business decision regardless of circumstances. If GF had performed well for AMD all that would change is the bad decision wouldn't be affecting AMD's business.

http://www.zdnet.com/blog/btl/amd-ditches-stake-in-globalfoundries/70775

I'd lay all the blame on AMD management if it weren't for the fact GF took on Ruiz as Chairman. Must have been a lot of camaraderie during the WSA negotiations. In my opinion AMD management AND the Abu Dhabi investors really pulled a number on the other AMD shareholders. Shame AMD management never formed a similar relationship with any OEMs, at least then they'd have a dedicated product channel even if it was bleeding them dry.

 

[SiliconWars]

No, becoming a fabless company by signing an agreement to be bound tightly to the fab you are spinning off is a bad business decision regardless of circumstances. If GF had performed well for AMD all that would change is the bad decision wouldn't be affecting AMD's business.

If GF had performed well and got the interest of other clients, and AMD had lost their wafers due to an inability to compete on price, you'd be saying the opposite ie asking why AMD didn't have it written in stone that they had to have first dibs on wafers for a set period of years. To me that one is justifiable - spinning off the fab which the majority of their business depended on, and not ensuring it was available to them - and only them - for a grace period would be borderline criminal.

Hindsight, while wonderful, is not something you can hold a company to task for. The decision that both parties made was the only sensible business decision that could be made.

 

[SiliconWars]

Ok so basically at the time of the deal, Dresden would have been 100% of AMD's cpu production.

You think that AMD should have spun it off and just behaved like a normal client to GF, putting their entire business at risk. There is only really ONE fab, 60k+ wafers per month that AMD has run their entire cpu business on, and you think that they should compete with all others for wafer starts at it?

That would have killed them stone dead in weeks. Intel could have bought every one of those wafers for 10% higher price and not even noticed it on their balance sheet. AMD needed the guarantee of those wafers because that *was* their business at the time. Even now with things at their worst point Dresden still supplies them with 60% of their CPU's. They need that capacity as a guarantee.

 

[Vesku]

They should have at least negotiated a framework for how much wafers would cost, isn't that a basic aspect of going from in house to a spin off? Even if that framework was not as favorable as a truly independent deal it would certainly be better than having no way to estimate production costs. Who was GF going to sell 45 and 32nm wafers to? You think Intel would decide to spend a billion+ a year just to trash those wafers (that would probably force the FTC to act) when even smaller fabless chip makers were only enticed to GF at 28nm? Just AMD and Rockchip so far with GF 28nm and if I had to guess GF probably did some enticing to get Rockchip on board.

 

[frozentundra123456]

Ok first of all, why would AMD help create GF, then ditch them for the same process except at IBM, which would almost certainly be more expensive?

Then you have the fact that GF are actually in the fab business while IBM has been scaling down their manufacturing for years. Who would you go with if your business depended on it?

All these ideas sound great until you realise that AMD basically had their ex fab, their ex fab guys and first dibs on their cutting-edge process.

*Why* would they say no to that? It was simply creating a new entity that provided them with everything they already had, except on a client basis. You heard what JHH said about "it's AMD's fab"?

Would you be saying this about GF if they hadn't screwed up 32nm or were so late with 28nm? No you wouldn't. That means you are looking at unforeseeable technical problems as the reason for being bad business decisions. The business decision was the only one that made sense, the technical problems (and the collapsing PC market) are what has made it look bad for AMD now - none of those were obvious at the time the deal was struck.

What if GF had provided a superb 32nm, 28nm and was racing towards 22nm in great shape? It would have been the exact same business decision, but now it would have been a great one. This is why the argument that it was a bad business decision doesn't make sense.

Yea, but that is EXACTLY the point. GF has done a lousy job and AMD is tied to them. Kind of like giving a quarterback a big contract. If he wins the super bowl, nobody cares, but if the team are continual losers, the deal will be harshly criticized. Like it or not results are what count.

 

[SiliconWars]

Yes I think Intel would spend a billion+ a year to kill off AMD. They've spent a helluva lot more trying to in the past. The real risk would be from a company like Nvidia or Apple or any number of companies who have more buying power than AMD does.

Fab 1 was AMD's entire CPU business - I wouldn't be surprised if it was AMD who asked for the rights to it until 2024.

There must have been some kind of wafer pricing agreement made, that is just obvious. You have to also consider that GF was doing SOI solely for AMD, that is R&D and wafers that nobody else was likely to want. GF needed to know that AMD wasn't going to leave them sitting with empty fab space and no compensation?

In all of this people are acting like the guys at GF are mugs with no business sense. You can't really believe that GF would accept some of the deals you think AMD should have got? The deal would have suited both for similar reasons - AMD needed the guarantee of wafers, GF needed the guarantee of a high volume customer for a set number of years. The only problem is the bad process, which was going to screw both over anyway...

 

[SiliconWars]

Yea, but that is EXACTLY the point. GF has done a lousy job and AMD is tied to them. Kind of like giving a quarterback a big contract. If he wins the super bowl, nobody cares, but if the team are continual losers, the deal will be harshly criticized. Like it or not results are what count.

And if GF churns out an amazing 20nm that everybody wants, but AMD has first refusal on, I'm sure you'll all be here apologising for suggesting that AMD got a bad deal out if it.

 

[Vesku]

Intel couldn't buy all of GFs wafers for the same reason they can't buy AMD now (they have more than enough cash to do so) and close it down, the US FTC and DoJ would be forced to act. That's assuming the Intel CEO could convince his board to let him spend several billion in the attempt. Heck if Intel were completely unrestricted by antitrust rules it could buy ARM via intermediaries and slowly shut it down.

GF's whole business plan was to expand capacity to compete with UMC and TSMC, not sure how exactly GF could both have an amazing 20nm and at the same time so little production volume that AMD finds it needs dibs.

 

[SiliconWars]

They couldn't do that for the same reason they can't buy AMD now (they have more than enough cash to do so) and close it down, the US FTC and DoJ would be forced to act. Heck if Intel were completely unrestricted by antitrust rules it could buy ARM via intermediaries and slowly shut it down.

Let's say AMD operated as a normal client of GF's, same rights as everybody else, and GF got other big clients interested. They would have been forced into an almost instant downsizing due to their inability to produce anything like the required number of chips the business was operating at.

Intel doesn't need to get involved, any one of the ARM guys would do. Qualcomm was long rumoured to be interested, no way could AMD compete with them and there would be no antitrust issues either. AMD would simply have cut their own throats by not guaranteeing capacity for a set number of years until they were better positioned to determine what direction their chipmaking took.

It was almost certainly AMD who asked for 15 years to do so.

 

[frozentundra123456]

And if GF churns out an amazing 20nm that everybody wants, but AMD has first refusal on, I'm sure you'll all be here apologising for suggesting that AMD got a bad deal out if it.

I haven't seen that 20nm yet, have you? In fact I havent even seen 28nm yet.