AMD Pays GF to get out of wafer commitment

[mrmt]

I thought they moved some APUs or CPUs production to TSMC, but don't know for sure.

Yes, this paragraph doesn't really clarify. Sounds like APUs/CPUs, but...

Current designs yes. Future designs will be GLF only.

 

[ShintaiDK]

So if you sum this all up it probably would have been better for AMD to keep the fabs themselves. They have no choice were to produce and now also no say what happens in foundry. Also if orders go down (fabs are not fully used) they suffer in both cases, probably even more now.

Sorry but that just proves how stupid their management was and the new guy now is putting the nails in the coffin. I mean if they have to lie and deflect all the time it's pretty obvious the truth isn't looking very good...

People thought it was good with ATIC and the "rich uncle". The threads about how AMD would catch up was endless.

But all they got was a sore rear end. ATIC almost got the fabs for free, and on top make AMD pay the running costs until they can get other customers. ATIC got zero long term interest in AMD.

 

[Vesku]

What can he do? AMD has an inferior product and has to charge more to break even. Should he dance gangnam style in from of the OEMs? It is not a problem sales alone can solve, this is an engineering problem. AMD have to develop better products or at least products that offers more reasonable pricing to OEM.

As a matter of fact, I think he is doing something. I remember Rory enthusing about Brazos (IIRC he said that Brazos was "beautiful"), and since then Brazos volumes just went up, effectively cannibalizing single module Trinity/Athlon sales in notebooks and even on desktops. This is not something a mere salesman can do, this has to pass through executive levels and you can bet this decision bears Rory approval. Brazos should be the only product making money for AMD.

I understand this decisions brings desperation to AMD fans of the Athlon 64 days, but that's what has to be done. Focus on products that bring money for the company, and not in what make fanboys happy.

Trinity should be well represented around the $400-500 mark but I barely see any compared to the plethora of pentium and celeron product. Trinity is good enough in that segment but you can't buy it if they don't make it.

Edit: Maybe Read is getting somewhere, Newegg Trinity listings in that range have tripled since I last checked about a month ago.

 

[podspi]

Just wanted to quote the numbers directly from mrmt's link since they address the question head-on:


So desktop GPU marketshare is 35.7% versus notebook marketshare being 34.2%.

In other words - its bleak across the board.

Ouch... though what is the excuse here, marketing? Perhaps Radeon is being hit by being associated with AMD D:?

Radeon 7K series are great competitors. I honestly don't think you can make the argument that current Radeon's aren't competitive like the FX-series is compared to the i7. You could make the argument that they are not as power-efficient, but gamers tend to not care, as long as the performance is there.

 

[mrmt]

Trinity should be well represented around the $400-500 mark but I barely see any compared to the plethora of pentium and celeron product. Trinity is good enough in that segment but you can't buy it if they don't make it.

Good for who? One has to ask the value of the Trinity line up when even AMD is cannibalizing its market. Instead of pushing single module die-salvaged Trinity parts, they are pushing Brazos to whatever design they can, even on desktops. One has to look just at the die size of Trinity and Pentium to see which product has the smallest price, and consequently better returns for the OEM.

 

[mrmt]

Radeon 7K series are great competitors. I honestly don't think you can make the argument that current Radeon's aren't competitive like the FX-series is compared to the i7. You could make the argument that they are not as power-efficient, but gamers tend to not care, as long as the performance is there.

Look at the die size of AMD GPU chips. They are either bigger or the same size as Nvidia chips. The problem? Nvidia has Quadro/Tesla, segments that make a lot of money for them. It was those segments that kept Nvidia on the black when they pushed their big GPGPU chips on the consumer market.

AMD has none of those premium segments, so they haven't the same flexibility with designs. They must always have smaller dies in order to afford smaller prices, otherwise the GPU segment will start to drain money, further weakening AMD balance sheet, a scenario that AMD has been very careful to avoid.

With GNC they pulled exactly the opposite. Big chips that costs on par or more than Nvidia chips. With GNC AMD practicably gave the GPU market to Nvidia, much like Bulldozer gave away the PC market to Intel. Sure, GNC performance is competitive, but be competitive or even beat your competitor is not enough to make money with a product.

 

[tviceman]

Ouch... though what is the excuse here, marketing? Perhaps Radeon is being hit by being associated with AMD D:?

Radeon 7K series are great competitors. I honestly don't think you can make the argument that current Radeon's aren't competitive like the FX-series is compared to the i7. You could make the argument that they are not as power-efficient, but gamers tend to not care, as long as the performance is there.

The hd7000 series lineup is great and arguably better than Nvidia's current Kepler lineup in many aspects. The problem with abysmal sales goes beyond marketing and "crap by brand name association." The first impressions of the hd7000 series cards, across the board, wasn't universally fantastic. Mostly, people felt all the intro prices were too high. Others felt the performance bump over last gen competition wasn't anything to get excited about. Whatever the reasons, first impressions tend to stick.

Then nvidia marched out the gate with a faster, cheaper, lower tdp chip. Right then and there the damage was done. Besides the fact that nvidia now claimed best perf/watt (one of the biggest knocks against nvidia in the past few years and the only variable completely out of AMD's control), AMD had plenty of time to react ahead of Kepler's launch by either lowering prices or releasing the higher binned Tahiti (hd7970ge) but they instead delayed doing either by enough time to leave additionally negative impressions beyond the initial low rumble of having too high of prices or not large enough performance increases.

 

[Vesku]

Good for who? One has to ask the value of the Trinity line up when even AMD is cannibalizing its market. Instead of pushing single module die-salvaged Trinity parts, they are pushing Brazos to whatever design they can, even on desktops. One has to look just at the die size of Trinity and Pentium to see which product has the smallest price, and consequently better returns for the OEM.

Pretty sure Brazos still has lower gross margins then Trinity. Perhaps I'm not recalling correctly but I believe it's one of the explanations AMD gave for the overall decline in margins (Brazos being a bigger % of sales).

 

[mrmt]

Pretty sure Brazos still has lower gross margins then Trinity. Perhaps I'm not recalling correctly but I believe it's one of the explanations AMD gave for the overall decline in margins (Brazos being a bigger % of sales).

Brazos has structurally lower margins than what historically was the standard margin for AMD high performance CPUs, the problem is that nowadays AMD cannot correctly design and manufacture a high performance CPU.

How much do you think AMD earns when a 246mm^2 SOI chip reaches the consumer at 80USD? Peanuts, I assure you. Just for comparison, Brazos die size is around 75mm^2.

You just have to look at what chips are responsible for the latest WSA snafu. It wasn't Brazos, but Trinity/Bulldoze.r

 

[krumme]

From here it looks like ATIC just f....d AMD, and AMD shareholders big time.

But anyway, the agreement with the seperation of GF and AMD, have left both company perhaps leaning on the other for progress, but have left both at the curb.

How have GF performed? Nothing but ppt slides and bs imho. Do we have any solid evidence of actual performance beyond 32nm? - i mean there must be something to it? What partners and agreements are there?

ATIC is a intellectual hole in the ground. Who the are they? How on earth can they get that sort of money?

(edit: ATIC no excuse for Hector and Dirk mess with this GF deal)

 

[OILFIELDTRASH]

GloFo is basically a loan-shark in this scenario, giving out balloon-payment loans at ridiculously high effective interest rates when it is all said and done, to a customer that wouldn't otherwise accept the "help" if it weren't for their desperate situation.

Some states have made so-called payday loans illegal because it preys on a very specific type of psychological weakness that has been well studied by sociologists.

"I will gladly pay you on Tuesday for a hamburger today"

AMD has been backed into a financial corner and they are now making desperate decisions that any other business would refuse to consider. The psychology is the same, and the ramifications will be too.

The problem for GloFo though is that this is nothing but bad PR for them, no loan shark wants to develop the reputation of being a loan-shark unless all they want for clients are desperate customers who have fallen on very hard times...all bonafide clients with good credit ratings will avoid such loan sharks and opt to take their loans from a much more reputable bank (or foundry in this case, i.e. TSMC).

Fabless companies will view GloFo's opportunistic contracts and inflexibility for what it is, and they will vote with their wallets and elect to keep their business at TSMC or UMC or Samsung.

Sounds like a solid business move to me. Heck I wish AMD would pay me 320 mil for doing nothing.

 

[Vesku]

Brazos has structurally lower margins than what historically was the standard margin for AMD high performance CPUs, the problem is that nowadays AMD cannot correctly design and manufacture a high performance CPU.

How much do you think AMD earns when a 246mm^2 SOI chip reaches the consumer at 80USD? Peanuts, I assure you. Just for comparison, Brazos die size is around 75mm^2.

You just have to look at what chips are responsible for the latest WSA snafu. It wasn't Brazos, but Trinity/Bulldoze.r

http://www.wikinvest.com/stock/Advanced_Micro_Devices_(AMD)/Data/Gross_Margin

If this data is accurate I'm not seeing Brazos having any better margins. The biggest hit was unsold inventory, presumably Llano and first gen bulldozer.

As far as I can see the only reason to pay GF to not produce Trinity chips is because AMD doesn't think they can sell them at a fast enough pace to not accumulate large inventories and risk another $100 million + write off.

Pentium b960 vs Trinity A6-4400M, similar price why can't AMD sell enough of these to justify having GF produce enough to meet their contractual obligations? I don't think it has much to do with the hardware. I think it's other business factors, AMD is sitting shyly near the punchbowl watching Intel work the dance hall.

 

[mrmt]

If this data is accurate I'm not seeing Brazos having any better margins. The biggest hit was unsold inventory, presumably Llano and first gen bulldozer.

I don't get it. Which data? AMD gross margins? You are assuming that Trinity gross margins is constant and that as Brazos grows in share the gross margin goes down? Yes, it would be a fair assumption if Bulldozer/Trinity volumes were constant, they aren't. These volumes are falling and they have to compensate somehow, and you can bet that they brought Trinity prices to the floor before paying 320 million to GLF not manufacture more chips for them.

As far as I can see the only reason to pay GF to not produce Trinity chips is because AMD doesn't think they can sell them at a fast enough pace to not accumulate large inventories and risk another $100 million + write off.

If they had any price latitude they would crater the prices and grab market share from Intel. Why don't they do that? Because they are already at their floor. It isn't interesting for them to sell more chips for lower prices because if they do they will incur in losses. Same with GPU.

Pentium b960 vs Trinity A6-4400M, similar price why can't AMD sell enough of these to justify having GF produce enough to meet their contractual obligations? I don't think it has much to do with the hardware. I think it's other business factors, AMD is sitting shyly near the punchbowl watching Intel work the dance hall.

Because the die size of a Pentium is half the size of a Trinity, which means that at the same price Intel will have >>a lot<< more of gross margin than AMD. In fact, for OEM Intel should be a lot more competitive than AMD. That's why their market share is plunging outside the Brazos world.

 

[Vesku]

I don't get it. Which data? AMD gross margins? You are assuming that Trinity gross margins is constant and that as Brazos grows in share the gross margin goes down? Yes, it would be a fair assumption if Bulldozer/Trinity volumes were constant, they aren't. These volumes are falling and they have to compensate somehow, and you can bet that they brought Trinity prices to the floor before paying 320 million to GLF not manufacture more chips for them.

That there was not much of a change in Gross Margins until they started having to write off inventory. Which brings us to that they can't shift enough A6-4400Ms that are priced similarly to Pentium b960s. This is taking the margin difference into account, this is their current pricing at current margins.

It's my position that if you magically gave the A6-4400Ms to Nvidia to sell that they'd be moving a lot more of them than AMD. That it's more than technical differences between Intel and AMD that is hindering AMD.

 

[Ajay]

Sounds like a solid business move to me. Heck I wish AMD would pay me 320 mil for doing nothing.

We need to get a business proposal to them quick. They clearly need a middleman to take a 10% commission on those deals

 

[mrmt]

That there was not much of a change in Gross Margins until they started having to write off inventory. Which brings us to that they can't shift enough A6-4400Ms that are priced similarly to Pentium b960s. This is taking the margin difference into account, this is their current pricing at current margins.

There is a reduction of the gross margins beyond the inventory impairment. Once you exclude the 100 million write off from the COGS in Q3 you get 38% gross margins.

But, what happened with the product mix in Q3? First Llano is being phased out, but more important, the good old Athlon X2, X3 and X4 are being phased out to give their place to Bulldozer derivatives. The reception of the new Trinity/Bulldozer line up isn't good to say the least, they simply cannot replace sales of their counterparts much less grow in sales, and they are usually bigger chips than the old ones.

We have a direct correlation here, once Bulldozer/Trinity went mainstream sales cratered AND gross margins plunged. So while I think Brazos gross margins aren't anything stellar, Bulldozer/Trinity margins are so awful that AMD prefers to shell out money to GLF not manufacture then. They didn't do the same with TSMC orders as far as we know.

 

[Vesku]

Going to try to condense it down:

~40% margin A6-4400M selling price ~= ~60% margin Pentium b960 selling price

Now why is AMD having such trouble moving parts like this?

IMO, if it was magically a Nvidia product it would move more without any further hits to margins. Assertion - AMD is it's own worst enemy.

 

[mrmt]

Going to try to condense it down:

~40% margin A6-4400M selling price ~= ~60% margin Pentium b960 selling price

Now why is AMD having such trouble moving parts like this?

You cannot make a raw comparison between Intel and AMD this way.

Intel 60% gross margin is boosted by the server and datacenter group. This business alone is a 10 billion dollars business, almost twice AMD sales in 2011 and with operating margins of 50%. Gross margins should be well in excess of 75% here. The PC and client group should have margins above 50%, but more likely on the lows 50.

But... let me put in perspective Pentium and Trinity to show you how bad is AMD situation. Trinity is more or less twice the size of Pentium processors, which means that Intel can get more or less twice the number of candidates per wafer, meaning that even at equal yields Pentium costs half of what Trinity costs to AMD. Once you factor that Intel yields are usually better than GLF, the balance shifts even more to Intel side. And what happens then?

I understand that your margins numbers are flawed, but as they would put Intel in a worse situation let's use them for a scenario.

Assume that COGS for a Pentium is 15USD and twice this value for Trinity. What happens once we apply your gross margins estimates? AMD needs to sell Trinity for 50USD to get 40% gross margins, but Intel needs to sell Pentium for 25USD to achieve 60% gross margins. Intel can make a hefty profit and still stay below AMD break even cost.

This is a very raw comparison and does not reflect a lot of differences between the companies, but you can get the picture. A6-4400M and Pentium b960 at the same price means that either Intel is making loads of money and AMD is making small amounts, or that Intel is making a fair amount of money and AMD is doing a small one, if any.

 

[Roland00Address]

But... let me put in perspective Pentium and Trinity to show you how bad is AMD situation. Trinity is more or less twice the size of Pentium processors, which means that Intel can get more or less twice the number of candidates per wafer, meaning that even at equal yields Pentium costs half of what Trinity costs to AMD. Once you factor that Intel yields are usually better than GLF, the balance shifts even more to Intel side. And what happens then?

I understand that your margins numbers are flawed, but as they would put Intel in a worse situation let's use them for a scenario.

Assume that COGS for a Pentium is 15USD and twice this value for Trinity. What happens once we apply your gross margins estimates? AMD needs to sell Trinity for 50USD to get 40% gross margins, but Intel needs to sell Pentium for 25USD to achieve 60% gross margins. Intel can make a hefty profit and still stay below AMD break even cost.

This is a very raw comparison and does not reflect a lot of differences between the companies, but you can get the picture. A6-4400M and Pentium b960 at the same price means that either Intel is making loads of money and AMD is making small amounts, or that Intel is making a fair amount of money and AMD is doing a small one, if any.

The thing is AMD and Intel are not at war with each other. They are not trying to destroy each other.

Intel is trying to do profit maximization so they will price the product competitively so people will buy the Pentium, but they will not price it too low if they lose money since the increase volume does not make up for the decreased average selling price. Furthermore Intel wants AMD to at least survive so they won't be charged with being a monopoly.

AMD is also trying to do profit maximization, but in reality it is just trying to make any form of profit and stay above water.

There is no reason ever Intel will price a pentium at 25 dollars.

------

In theory since the A6-4400m is a single module part with half the shaders they could always release a separate die for it. A 123mm^2 part is a lot more cost effective than a 246mm^2 part. For comparison a sandybridge pentium is either a gt1 part with a 131mm^2 die or a cut down gt2 part with a 149mm^2 die. Then again the sandybridge pentium is 25% faster than the A6 in cpu task, the A6 has better graphics but it is such a low bar of graphics I wouldn't be happy with it.

 

[mrmt]

There is no reason ever Intel will price a pentium at 25 dollars.

To an OEM like Dell? Why not?

In theory since the A6-4400m is a single module part with half the shaders they could always release a separate die for it. A 123mm^2 part is a lot more cost effective than a 246mm^2 part. For comparison a sandybridge pentium is either a gt1 part with a 131mm^2 die or a cut down gt2 part with a 149mm^2 die. Then again the sandybridge pentium is 25% faster than the A6 in cpu task, the A6 has better graphics but it is such a low bar of graphics I wouldn't be happy with it.

Yes, a single module die should at least have better economics than fused off dual module parts, but once you look at Trinity shipment numbers you see that the volume does not justify the additional effort to build and validate this SKU. It is more or less what happens with Bulldozer, where the quad core is the eight core die but with two modules fused off.

 

[pablo87]

From here it looks like ATIC just f....d AMD, and AMD shareholders big time.

But anyway, the agreement with the seperation of GF and AMD, have left both company perhaps leaning on the other for progress, but have left both at the curb.

How have GF performed? Nothing but ppt slides and bs imho. Do we have any solid evidence of actual performance beyond 32nm? - i mean there must be something to it? What partners and agreements are there?

ATIC is a intellectual hole in the ground. Who the are they? How on earth can they get that sort of money?

(edit: ATIC no excuse for Hector and Dirk mess with this GF deal)

Krumme, I give up to rationalize the irrational. AMD BoD should file chapter 11 to get out of such unreasonable agreement...between this penalty and poorer yields and higher packaging, cost of trinity/llano must be 6x Brazos!!!

ps - RR might be the icing of cake worst CEO though, seems he has no idea what he's doing.

 

[lamedude]

IANAL but I don't think that's an option. Under Prohibited Action in the cross license agreement it says

any Chapter 11 plan of reorganization or similar reorganization or liquidation plan that contemplates or would permit a Change of Control of such Person upon or following the effectiveness of such plan, unless such plan and any order confirming or otherwise approving such plan would and does also expressly bind such Person and its Subsidiaries, and each of their successors and assigns, to each and every term and provision of this Agreement, including without limitation Sections 5.2 and 7.2 hereof, upon and following any Change of Control;

 

[Vesku]

The list price on Pentium b960 is roughly the same as the list price for the A6-4400M. It's AMD not Intel that is taking the margin hit due to die size, etc. Heck, the A8-4500M isn't that much more than the A6-4400M, which is not true of going from the Pentium mobile series to the mobile i3 series for Intel (hefty price jump).

AMD, even though they are already taking the margin hit for competing with Intel just can't sell many of these chips. That is what paying to get out of wafer production tells us. I'm suggesting a less stigmatized tech company could move more of them at the current asking price (current margins).

 

[Ajay]

IANAL but I don't think that's an option. Under Prohibited Action in the cross license agreement it says

So that only applies to change of ownership (or the majority controlling party).

 

[mrmt]

The list price on Pentium b960 is roughly the same as the list price for the A6-4400M.

List price =! OEM price.

In the price formation process both can have very different markups and costs and still reach the consumer at the same price.

AMD, even though they are already taking the margin hit for competing with Intel just can't sell many of these chips. That is what paying to get out of wafer production tells us.

If AMD could make money, and I mean any money, by selling Trinity and Bulldozer why would they pay GLF to not manufacture chips for them? It would be an asinine decision even by AMD standards to shell out millions when you could just manufacture the chips and make a few bucks with them. A few bucks is better than nothing anyway.

The fact they chose the WSA amendment route means that there isn't market for the chips at the prices they need to make any money, so they either take losses at the hands of the OEM or at the hands of GLF.