So how does AMD pull this one through?

[CHADBOGA]

the x86 license and current tech AMD has under it's IP alone is worth keeping it going.

How are IBM and VIA going with their x86 licensed products?

 

[Phynaz]

Someone explain the whole Global Foundries thing pls.

Like, why did AMD think that spinning it off was a good idea? Seems to me that immediately, you have higher costs after spin-off, and since all that AMD does is make chips, it would make sense to keep that in-house.

I mean, they could always take contracts with other firms like apparently GF is doing with Qualcomm. Might even end up making more money that way, by taking contract orders from the various ARM manufacturing chips, become like TMSC.

Was it b/c they needed capital?

The capital required to continually develop new process nodes is more that a company the size of AMD can afford. The idea was spin off the foundry and then get additional customers on board to help spread out the cost.

 

[frozentundra123456]

The capital required to continually develop new process nodes is more that a company the size of AMD can afford. The idea was spin off the foundry and then get additional customers on board to help spread out the cost.

If I recall the time frame correctly, it was also related to the huge outlay for the purchase of ATI, which burdened them with a lot of debt. I believe they were almost forced to divest their foundaries.

Edit: I was really in favor of the ATI purchase and had great expectations for APUs. Unfortuantely, APUs so far have been in a kind of middle ground that does not make it a compelling attraction to me in the desktop. It doesnt help that Llano was late and had mediocre CPU performance, giving Intel a chance to not really catch up, but to at least get out an igp that was decent for the average user. Perhaps with the console wins and new attempts to overcome the bandwidth issues, the purchase of ATI will yet pay off. But at best, they paid way too much.

 

[mrmt]

Like, why did AMD think that spinning it off was a good idea? Seems to me that immediately, you have higher costs after spin-off, and since all that AMD does is make chips, it would make sense to keep that in-house.

They didn't have a choice here. After they wrecked their balance sheet buying ATI, they couldn't afford 32nm factories, so they had to sell it and offload some debt to GLF, and because of that we got the WSA.

I mean, they could always take contracts with other firms like apparently GF is doing with Qualcomm. Might even end up making more money that way, by taking contract orders from the various ARM manufacturing chips, become like TMSC.

Whatever chances of AMD becoming a second TSMC died when they decided to adopt IBM SOI process. SOI wafers are expensive, no ARM manufacture would develop a low-cost solution for a SOI node.

Also AMD was capacity-constrained since 2000, so even if they had a suitable process they wouldn't have the spare capacity to fill with orders from other companies.

 

[rgallant]

They didn't have a choice here. After they wrecked their balance sheet buying ATI, they couldn't afford 32nm factories, so they had to sell it and offload some debt to GLF

-didn't the end of the world in 2008 happen or was it just a bad dream .[big buy using credit and market crashes for unknown number of years]

 

[pablo87]

I don't know why so many posters on these forums see spinning off ATi as viable or desirable. If they wanted to make money from their graphics IP, the better route would be to start licensing it to others to be integrated into SoCs- I'm sure Samsung would kill to get their hands on something as advanced as GCN.

The WSA is a flood and in floods you save what you can...

And in general, visual computing long-term has a much brighter future than x86. All these years x86 depended on a bloated o/s and a willing ecosystem for its success... Its not just tablets and smartphones either - Chromebook (Samsung hw) is a fine PC.

 

[Homeles]

Intel gets quite a bit of its income from Linux servers now.

 

[Dribble]

I think all thats really needed is for APUs to gain acceptance.

Once AMD starts makeing APUs that are akin to 7850's.....
ALOT of people wont bother with discrete cards anymore.

It ll basically kill off the intire low-end segment (then it becomes Intel IGP vs AMDs ~7850 level IGP).

That's what fan boys have been saying since the first APU came out. At the time everyone else said it's not possible, and time has show that yes indeed it's not possible.

Simple logic shows that - a high end gpu needs a lot of memory bandwidth and a lot of power. A cpu chip doesn't have the memory bandwidth - no fast 256bit HDDR5 bus here, even adding the pins to the chip for that sort of bandwidth would make it too expensive. What bandwidth there is has to be shared with the cpu anyway. On top of that the cpu is designed to be cheap and simple - so quad memory buses (still way short of a discrete card) are not an option. Equally a cpu chip can't give 100W+ thermal budget to the gpu, or power usage when both got going would be silly huge and very hard to dissipate - once again remember this is designed to be cheap and simple, not something that requires liquid cooling to work.

On top of that even if it was possible most people wouldn't care - they don't use their pc's for high end gaming. They aren't even buying pc's much, most of the money goes on phones and tablets now.

AMD as a company is not moving forward in the right markets (mobile, server, super computer), has too much debt to manoeuvre (can't invest enough to break into those markets) and a millstone around it's neck with the GloFo contract (forces them to buy lots of chips whether they want them or not from an uncompetitive fab). Management is also top heavy and incompetent. It's not pretty at all.

The future:
1) they go bust and all the bright people and patents go to other companies who might be able to make something happen.
2) they do a ViA and continue to exist in a little way.
3) they find the next Steve Jobs, do an Apple and rise like a colossus from the ashes.

3 is the dream, 2 is what happens if their management finally does something right, 1 is probably realistically the best use of the talent they have.

 

[NTMBK]

That's what fan boys have been saying since the first APU came out. At the time everyone else said it's not possible, and time has show that yes indeed it's not possible.

Simple logic shows that - a high end gpu needs a lot of memory bandwidth and a lot of power. A cpu chip doesn't have the memory bandwidth - no fast 256bit HDDR5 bus here, even adding the pins to the chip for that sort of bandwidth would make it too expensive. What bandwidth there is has to be shared with the cpu anyway. On top of that the cpu is designed to be cheap and simple - so quad memory buses (still way short of a discrete card) are not an option. Equally a cpu chip can't give 100W+ thermal budget to the gpu, or power usage when both got going would be silly huge and very hard to dissipate - once again remember this is designed to be cheap and simple, not something that requires liquid cooling to work.

On top of that even if it was possible most people wouldn't care - they don't use their pc's for high end gaming. They aren't even buying pc's much, most of the money goes on phones and tablets now.

AMD as a company is not moving forward in the right markets (mobile, server, super computer), has too much debt to manoeuvre (can't invest enough to break into those markets) and a millstone around it's neck with the GloFo contract (forces them to buy lots of chips whether they want them or not from an uncompetitive fab). Management is also top heavy and incompetent. It's not pretty at all.

The future:
1) they go bust and all the bright people and patents go to other companies who might be able to make something happen.
2) they do a ViA and continue to exist in a little way.
3) they find the next Steve Jobs, do an Apple and rise like a colossus from the ashes.

3 is the dream, 2 is what happens if their management finally does something right, 1 is probably realistically the best use of the talent they have.

Kaveri is rumoured to be using GDDR5. The PS4 chip is already confirmed to be using GDDR5. And Intel's top-end mobile Haswell has a big fat dollop of eDRAM on package. So your theory of APUs never being able to break through the bandwidth wall seems to be a bit flawed...

 

[Arkadrel]

Simple logic shows that - a high end gpu needs a lot of memory bandwidth and a lot of power. A cpu chip doesn't have the memory bandwidth

Think most people know this.
However thats about to change.

http://www.xbitlabs.com/news/cpu/di..._Kaveri_APU_Supports_GDDR5_Memory_Report.html

With the next-generation high-performance APU code-named Kaveri, AMD will finally unleash its Radeon HD’s potential thanks to a secret weapon, 128-bit GDDR5 memory controller.

This means we ll see system memory reach something like:

128bit x 1.750 (7ghz GDDR5) / 8 x 4 = 112 GB/s
system memory bandwidth (should be enough to feed a APU's IGP).

...Equally a cpu chip can't give 100W+ thermal budget to the gpu, or power usage when both got going would be silly huge and very hard to dissipate...

A 7790 uses around 69watts while running Crysis 2 benchmark (TPU review).
Now a 7790 has alot of things active, that a APU wouldnt have....
ei. no discrete GDDR5 (that would be part of the system memory, not the GPU/APU's power usage).

Lets say a APU's IGP with 7790 level, uses 50watts while playing games.
Now add a 65watt TPD level CPU design.

~115watt TPD APU with 7790 level of performance.

It is possible.
It should have been done long ago.

 

[frozentundra123456]

Kaveri is rumoured to be using GDDR5. The PS4 chip is already confirmed to be using GDDR5. And Intel's top-end mobile Haswell has a big fat dollop of eDRAM on package. So your theory of APUs never being able to break through the bandwidth wall seems to be a bit flawed...

Kaveri is supposed to have 512sp, is it not? That is 7750 level, not 7850 level. Every new generation of apu has supposed to be the one that reaches mid range discrete levels of performance, but has ultimately not even come close.

 

[Idontcare]

The future:
1) they go bust and all the bright people and patents go to other companies who might be able to make something happen.
2) they do a ViA and continue to exist in a little way.
3) they find the next Steve Jobs, do an Apple and rise like a colossus from the ashes.

3 is the dream, 2 is what happens if their management finally does something right, 1 is probably realistically the best use of the talent they have.

If DEC had not done (1) then AMD would never had been able to develop the original K7 Athlon nor its successors. It very much saved AMD's bacon.

Who knows who's bacon will be saved should AMD do a DEC?

 

[Arachnotronic]

If DEC had not done (1) then AMD would never had been able to develop the original K7 Athlon nor its successors. It very much saved AMD's bacon.

Who knows who's bacon will be saved should AMD do a DEC?

Nobody's. Samsung already poached the low power core team at AMD

 

[Fox5]

Nobody's. Samsung already poached the low power core team at AMD

And qualcomm poached the high performance one.

 

[Concillian]

Someone explain the whole Global Foundries thing pls.

Like, why did AMD think that spinning it off was a good idea? Seems to me that immediately, you have higher costs after spin-off, and since all that AMD does is make chips, it would make sense to keep that in-house.

I mean, they could always take contracts with other firms like apparently GF is doing with Qualcomm. Might even end up making more money that way, by taking contract orders from the various ARM manufacturing chips, become like TMSC.

Was it b/c they needed capital?

Business deals like this are so intertwined with the ludicrous wall street systems that I'm not sure people like us can ever fully understand.

My thought is that it helped them hide a huge quantity of debt in confidential contractual agreements to use Glo Flo capacity. This kept some of the financials from needing to be immediately released to investors. I think you're right on the capital front, doing things how they did them, it was more like a loan than a deal. The confidentiality of the agreements meant they could slow leak the bad financial news so the stock wouldn't slide quickly, giving time for higherups to dump their stock through automatic sales rather than large one time sales that would set off the analyst alarm bells.

 

[mrmt]

And qualcomm poached the high performance one.

I didn't know this one. Got a source?

 

[Idontcare]

I didn't know this one. Got a source?

Not sure if this is ho he was referring to, but maybe?

http://www.tomshardware.com/news/eric-demers-amd-qualcomm-new-job,15689.html

 

[CHADBOGA]

Nobody's. Samsung already poached the low power core team at AMD

And qualcomm poached the high performance one.

Don't forget that AMD have Jim Keller, more experienced than many young Intel engineers. :awe:

 

[mrmt]

Not sure if this is ho he was referring to, but maybe?

http://www.tomshardware.com/news/eric-demers-amd-qualcomm-new-job,15689.html

Not sure it is. They lost just a CTO here, while the Samsung there is far more widespread. Even Otelini mentioned that poaching in a conference.

In any case, it seems that AMD lost most of the team that designed their Kabini core. Let's see if AMD can recover from this one, or if they will start bolting new things in the core without any kind of new development as they did with the A64 core.

 

[Idontcare]

Don't forget that AMD have Jim Keller, more experienced than many young Intel engineers. :awe:

You have to understand, Keller is the one.

How manu Keanu's does AMD need to take down the machine?

 

[Genx87]

I don't know if it would/will be the savior of AMD, but I am quite confident that Intel will eventually do a massive faceplant. No company in the history of companies has managed to avoid that fate.

If Intel cant break into the mobile markets and control it. I expect that faceplant to happen within the next decade. PC shipments are stagnating and or slipping. Hard to grow a business in a shrinking market.

 

[Genx87]

Bad weather is to be expected in the business ocean but the Gloflo WSA has been a perfect storm. Consider that 90% or so of the financial loss in 2012 was WSA related. Besides, the total cost of building at Gloflo is much higher than at TSMC so under an arms' length arrangement, AMD would have made money in 2012.

Its a major burden that hasn't entirely reared its head - beyond the next 4 quarters, nobody actually knows how much AMD has to buy from GF during the FOLLOWING 10 years - we do know its exclusive so regardless of GF's competence (or lack thereof), AMD is stuck - its not inconceivable that they would design the greatest processor and still go bankrupt because GF couldn't build it.

Complete disclosure would likely cause the share price to tank, and most of the remaining key staff to leave. Being conflicted and <insert adjective of choice here>, the current BoD will not address this until the company is past the point of no return. SOP.

There may be 1 loophole: an investor takes a 40% position ($800MM?), forces a vote on spinning off ATI to shareholders, and then restructures AMD.

Otherwise, despite the best efforts of the spinmeisters in renaming gaming consoles "embedded devices", the writing is on the tablet - AMD is going down.

//rant

I havent been paying enough attention to how the GF deal is structured. But somebody pointed out that it appears that AMD is being gutted to pay for GF. Think there is any truth to that?

 

[Idontcare]

I havent been paying enough attention to how the GF deal is structured. But somebody pointed out that it appears that AMD is being gutted to pay for GF. Think there is any truth to that?

That is a good way of putting it.

While the WSA is really costing AMD an arm and a leg in present times, what is really hurting AMD's future is the exclusivity requirement (AMD must use GloFo or pay huge penalties for an exclusivity waiver on a quarterly basis, as they are doing for all APU's fabbed at TSMC).

GloFo is what now, nearly 2yrs behind TSMC in getting 28nm into volume production?

If AMD were free to conduct their business as a fabless company, picking the foundry that had the best technology, prices, and time schedule that fit AMD's needs then AMD's future would be of their own making.

But as it stands now it is not, they have an exclusivity contract that stands until 2024

Given GloFo's track record, I don't see AMD surviving until then in the face of all their fabless competitors who are able to freely choose their foundry partner.

 

[mrmt]

You have to understand, Keller is the one.

I guess we now have a better understanding of the recent personnel cuts at AMD.

It's not that the company could not afford those engineers, it's just that they were made redundant by Keller. Why bother spending millions in thousands of engineers when *the* man can do the entire job faster and more efficiently.

 

[CHADBOGA]

That is a good way of putting it.

While the WSA is really costing AMD an arm and a leg in present times, what is really hurting AMD's future is the exclusivity requirement (AMD must use GloFo or pay huge penalties for an exclusivity waiver on a quarterly basis, as they are doing for all APU's fabbed at TSMC).

GloFo is what now, nearly 2yrs behind TSMC in getting 28nm into volume production?

If AMD were free to conduct their business as a fabless company, picking the foundry that had the best technology, prices, and time schedule that fit AMD's needs then AMD's future would be of their own making.

But as it stands now it is not, they have an exclusivity contract that stands until 2024

Given GloFo's track record, I don't see AMD surviving until then in the face of all their fabless competitors who are able to freely choose their foundry partner.

Perhaps one day AMD fans will finally be able to be honest in assessing the role that Hector Ruiz played in destroying AMD.

Hector gave himself a special multi-million dollar bonus for completing the sale of AMD's fabs and then arranged for himself to be the CEO of the newly purchased fabs, where he could continue to rake in the millions.

What a hero. :awe: