[Digitimes] Kaveri delayed

[Idontcare]

I don't really mind fanboys who 'cheer' something.

OTOH, posting non-stop negative/attack posts/threads against the 'enemy' of your favorite 'team' seems to be almost entirely acceptable here. Non-stop 'pro' threads don't bother me much. Non-stop 'bash' threads just seem like a negative environment to me personally, and this is a relatively new development in the years that I've been here (I'm a relative newcomer, visited since '03, member since '06).

I'm not questioning the policies, but I think that many would agree that this incessant negativity and hostility is not a fun or valuable part of the forums, and in fact is a bane against both common sense and the reasonable exchange of thoughts and ideas.

I'm a big fan of AMD GPUs, and some of their CPUs. I'm a big fan of certain Intel processors, while I think many of them are poor and/or overpriced. I currently own a mix of AMD/NV/Intel/Apple/etc products, and they all work fine for what I expect of them. I find myself defending the forum against trolls here constantly who love to bash some company over and over, but it seems to be a losing war. Just look at the state of this forum now. Virtually every thread derails into flaming, hostility, and mean-spirited sniping. Half the threads start out with that very intention. You can ding me for that if you wish, but I am starting to feel like the effort to wade through the fanboys, shills, and above all, the incessant negative/attack poster is just not worth it.

When you read your post in toto, complete with the sig that is not captured when quoting a member's post, I think you'll agree the solution is self-evident.

__________________
Death is the answer.

We (the vastly outnumbered moderators) can no more stop the negativity than we can encourage the positivity.

The people who have the power to stop the negativity are the very people who presently encourage it...stop posting in those threads, stop bumping them to the surface, stop giving the negativity an audience.

If a small group of individuals want to band together and trash-talk the competition of their favorite team then just leave them to their pit of despair and misery, don't waste your time going into the thread, spare yourself the pain of bearing witness to what goes on in their self-fashioned echo chamber of a thread.

The moderators really cannot stop that kind of forum activity without completely gutting the vibrancy of the forum itself. They are part-and-parcel. No one cheers "their own team" during a baseball game without tossing out a little trash talk about the other team.

The psychological parallels run deep and to every extreme, that is why they are called "fans", be it of technology or of sports. So you have to approach with an appropriate sense of expectation in those threads.

Like I said, the forums are intentionally open-ended in the sense that members are expected to take the good with the bad from their personal perspective of it. Hopefully there is enough good to keep you coming back despite the bad.

That said, it is still a tricky balance to strike because we do not want the hooligans and trolls here. And if a rowdy group of fans decide to riot and make trouble in the rest of the community then they need to be vacationed and put in the drunk tanks to cool off for a day or three.

That is where the community gets the community it deserves. If you see hooliganism and don't report it to the authorities then it will run rampant and unchecked. But not everything that looks like a troll is a troll, so be prepared for the mods to decide in their favor if we conclude they are acting (perhaps irrationally) as nothing more than a fan without the intent to troll.

 

[pablo87]

APU wise, AMD should just kill Kaveri - by the time they ship, the GPU tech. will be 2.5 years old, and built on process tech. that is hopelessly behind. And more generally, they should just stop with the dual APU development programs.

Instead, they should refresh Kabini with a GPU that is >> Kaveri's on TSMC's next generation mfg. process, followed later by Kabini 3 with an improved core in 2015 and so on - APU TICK TOCK so to speak. Straightforward, easy to understand, cost effective.

That said, I wouldn't give up on big core. While Intel has a big lead, they have been so focused on IGP and SOCs, they've lost focus on what got them here. After so many years, they don't even have a desktop quad core under $100 - core count wise, their value Consumer offerings are about to be overtaken by tablets and smartphones. In my mind, no matter what, a PC should always provide much better performance than a smartphone or tablet.

To me, the schwerpunkt for AMD big core should be quad core, low cost. No APUs. Its been stated that Kaveri is 220mm^2 with GPU. What would it be without, 120mm^2? Also, since the FCH is integrated, the mb cost should be much lower. The objective should be a profitable non OC itx MB* with GPU less Kaveri for $100, with a subsidized discrete GPU bundle. Now that would be disruptive not only from a cost perspective but from a go to market perspective - while Intel promotes integrated, AMD promotes discrete. Clear differentiation.

Edit - the beauty of this approach for AMD is that its not dependent on offering performance equivalent to Intel - it would win because it does away with integrated graphics, its quad core and its sub $100, a package that Intel can't or won't put together.

* AMD needs to have more influence and control on mb designs and pricing - they can't rely on the mb vendors who push come to shove, will always pander to Intel.

 

[Enigmoid]

APU wise, AMD should just kill Kaveri - by the time they ship, the GPU tech. will be 2.5 years old, and built on process tech. that is hopelessly behind. And more generally, they should just stop with the dual APU development programs.

Instead, they should refresh Kabini with a GPU that is >> Kaveri's on TSMC's next generation mfg. process, followed later by Kabini 3 with an improved core in 2015 and so on - APU TICK TOCK so to speak. Straightforward, easy to understand, cost effective.

That said, I wouldn't give up on big core. While Intel has a big lead, they have been so focused on IGP and SOCs, they've lost focus on what got them here. After so many years, they don't even have a desktop quad core under $100 - core count wise, their value Consumer offerings are about to be overtaken by tablets and smartphones. In my mind, no matter what, a PC should always provide much better performance than a smartphone or tablet.

To me, the schwerpunkt for AMD big core should be quad core, low cost. No APUs. Its been stated that Kaveri is 220mm^2 with GPU. What would it be without, 120mm^2? Also, since the FCH is integrated, the mb cost should be much lower. The objective should be a profitable non OC itx MB* with GPU less Kaveri for $100, with a subsidized discrete GPU bundle. Now that would be disruptive not only from a cost perspective but from a go to market perspective - while Intel promotes integrated, AMD promotes discrete. Clear differentiation.

Edit - the beauty of this approach for AMD is that its not dependent on offering performance equivalent to Intel - it would win because it does away with integrated graphics, its quad core and its sub $100, a package that Intel can't or won't put together.

* AMD needs to have more influence and control on mb designs and pricing - they can't rely on the mb vendors who push come to shove, will always pander to Intel.

Sure the tech will be a little old but there is nothing better than it currently. 28 nm is still here.

Kabini would completely need a rework to be anything other than low power (expanded PCIe support, dual channel RAM). Not to mention that Kabini's cpu is so weak it bottlenecks its own gpu.

I agree that with current pricing AMD is basically throwing the GPU on there for free.

 

[ShintaiDK]

Also, since the FCH is integrated, the mb cost should be much lower.

The FCH is not integrated.

 

[PPB]

APU wise, AMD should just kill Kaveri - by the time they ship, the GPU tech. will be 2.5 years old, and built on process tech. that is hopelessly behind. And more generally, they should just stop with the dual APU development programs.

Instead, they should refresh Kabini with a GPU that is >> Kaveri's on TSMC's next generation mfg. process, followed later by Kabini 3 with an improved core in 2015 and so on - APU TICK TOCK so to speak. Straightforward, easy to understand, cost effective.

That said, I wouldn't give up on big core. While Intel has a big lead, they have been so focused on IGP and SOCs, they've lost focus on what got them here. After so many years, they don't even have a desktop quad core under $100 - core count wise, their value Consumer offerings are about to be overtaken by tablets and smartphones. In my mind, no matter what, a PC should always provide much better performance than a smartphone or tablet.

To me, the schwerpunkt for AMD big core should be quad core, low cost. No APUs. Its been stated that Kaveri is 220mm^2 with GPU. What would it be without, 120mm^2? Also, since the FCH is integrated, the mb cost should be much lower. The objective should be a profitable non OC itx MB* with GPU less Kaveri for $100, with a subsidized discrete GPU bundle. Now that would be disruptive not only from a cost perspective but from a go to market perspective - while Intel promotes integrated, AMD promotes discrete. Clear differentiation.

Edit - the beauty of this approach for AMD is that its not dependent on offering performance equivalent to Intel - it would win because it does away with integrated graphics, its quad core and its sub $100, a package that Intel can't or won't put together.

* AMD needs to have more influence and control on mb designs and pricing - they can't rely on the mb vendors who push come to shove, will always pander to Intel.

Methinks Kaveri is shipping with GCN 1.1 as it's better suited for the HSA thing than 1.0

 

[pablo87]

Sure the tech will be a little old but there is nothing better than it currently. 28 nm is still here.

Kabini would completely need a rework to be anything other than low power (expanded PCIe support, dual channel RAM). Not to mention that Kabini's cpu is so weak it bottlenecks its own gpu.

I agree that with current pricing AMD is basically throwing the GPU on there for free.

Expanded PCIe is not essential though in that space - more of an expensive nice to have. Kabini should have been dual channel to begin with.

I wasn't able to confirm TSMC's 20nm schedule from general sources but I have to imagine its sometime in 2014 - should that not coincide with a new Kabini (at this point, I've given up on GF) just like GCN was at 28nm?

 

[pablo87]

The FCH is not integrated.

The more reason to CANX Kaveri tbh at least in its current iteration.

 

[Enigmoid]

Expanded PCIe is not essential though in that space - more of an expensive nice to have. Kabini should have been dual channel to begin with.

I wasn't able to confirm TSMC's 20nm schedule from general sources but I have to imagine its sometime in 2014 - should that not coincide with a new Kabini (at this point, I've given up on GF) just like GCN was at 28nm?

Isn't the GPU lane for kabini PCIe 2.0 x4?

Fine for ultra low power/ low power devices but for desktop more lanes would be nice.

 

[krumme]

APU wise, AMD should just kill Kaveri - by the time they ship, the GPU tech. will be 2.5 years old, and built on process tech. that is hopelessly behind. And more generally, they should just stop with the dual APU development programs.

Instead, they should refresh Kabini with a GPU that is >> Kaveri's on TSMC's next generation mfg. process, followed later by Kabini 3 with an improved core in 2015 and so on - APU TICK TOCK so to speak. Straightforward, easy to understand, cost effective.

That said, I wouldn't give up on big core. While Intel has a big lead, they have been so focused on IGP and SOCs, they've lost focus on what got them here. After so many years, they don't even have a desktop quad core under $100 - core count wise, their value Consumer offerings are about to be overtaken by tablets and smartphones. In my mind, no matter what, a PC should always provide much better performance than a smartphone or tablet.

To me, the schwerpunkt for AMD big core should be quad core, low cost. No APUs. Its been stated that Kaveri is 220mm^2 with GPU. What would it be without, 120mm^2? Also, since the FCH is integrated, the mb cost should be much lower. The objective should be a profitable non OC itx MB* with GPU less Kaveri for $100, with a subsidized discrete GPU bundle. Now that would be disruptive not only from a cost perspective but from a go to market perspective - while Intel promotes integrated, AMD promotes discrete. Clear differentiation.

Edit - the beauty of this approach for AMD is that its not dependent on offering performance equivalent to Intel - it would win because it does away with integrated graphics, its quad core and its sub $100, a package that Intel can't or won't put together.

* AMD needs to have more influence and control on mb designs and pricing - they can't rely on the mb vendors who push come to shove, will always pander to Intel.

Intel will see there is no profit with their cost structure competing with arm. Its was like IBM competing with Intel.
They will therefore have to use their fab capacity for other customers.
In a market where tsmc is accelerating GF is just plain doomed with owners that dont know anything about the business.
That means mubadala will dump amd.
The relationship is bad for all.
In that situation perhaps 5 years from now, big cores from amd like kaveri is probably profitable on tsmc process but i think its going the arm way anyway. Amd might just trim the organization further and just stop kaveri like products - if they are allowed by mubadala.
If they are not allowed i would just delay it

Its a mess...

 

[SiliconWars]

You need to understand the mission to understand why mubadala isn't getting out of this business.

http://en.wikipedia.org/wiki/Mubadala_Development_Company

Mubadala’s mandate is to facilitate the diversification of Abu Dhabi’s economy. Its focus is on managing long-term, capital-intensive investments that deliver strong financial returns and tangible social benefits for the Emirate.

Globalfoundries is the fastest growing semi in the world.
And still people claim that GF is doing a bad job...unbelievable.

TSMC - 26 years old
Globalfoundries - 4 years old

Thank god myopic tech forum "analysts" aren't running the company or their whole ethos would have failed at the first step when they couldn't immediately usurp the established industry giants. Dear oh dear.

 

[Arkaign]

When you read your post in toto, complete with the sig that is not captured when quoting a member's post, I think you'll agree the solution is self-evident.



We (the vastly outnumbered moderators) can no more stop the negativity than we can encourage the positivity.

The people who have the power to stop the negativity are the very people who presently encourage it...stop posting in those threads, stop bumping them to the surface, stop giving the negativity an audience.

If a small group of individuals want to band together and trash-talk the competition of their favorite team then just leave them to their pit of despair and misery, don't waste your time going into the thread, spare yourself the pain of bearing witness to what goes on in their self-fashioned echo chamber of a thread.

The moderators really cannot stop that kind of forum activity without completely gutting the vibrancy of the forum itself. They are part-and-parcel. No one cheers "their own team" during a baseball game without tossing out a little trash talk about the other team.

The psychological parallels run deep and to every extreme, that is why they are called "fans", be it of technology or of sports. So you have to approach with an appropriate sense of expectation in those threads.

Like I said, the forums are intentionally open-ended in the sense that members are expected to take the good with the bad from their personal perspective of it. Hopefully there is enough good to keep you coming back despite the bad.

That said, it is still a tricky balance to strike because we do not want the hooligans and trolls here. And if a rowdy group of fans decide to riot and make trouble in the rest of the community then they need to be vacationed and put in the drunk tanks to cool off for a day or three.

That is where the community gets the community it deserves. If you see hooliganism and don't report it to the authorities then it will run rampant and unchecked. But not everything that looks like a troll is a troll, so be prepared for the mods to decide in their favor if we conclude they are acting (perhaps irrationally) as nothing more than a fan without the intent to troll.

Good points, IDC. I'll refrain from honoring the trolls with responses. It's useless anyway. The intelligent members already know these trolls given their records. Basically anyone who posts primarily attack/bash threads has already lost all credibility in the minds of a reasonable person.

 

[Idontcare]

You need to understand the mission to understand why mubadala isn't getting out of this business.

http://en.wikipedia.org/wiki/Mubadala_Development_Company

Globalfoundries is the fastest growing semi in the world.
And still people claim that GF is doing a bad job...unbelievable.

TSMC - 26 years old
Globalfoundries - 4 years old

Thank god myopic tech forum "analysts" aren't running the company or their whole ethos would have failed at the first step when they couldn't immediately usurp the established industry giants. Dear oh dear.

The added hyperbole doesn't do your post justice, you'd have better reception within the community of you dialed that back.

Getting to the meat of your post, people are right to be reserved and pessimistic regarding GloFo's prospects. History in the sector is why.

TSMC left UMC, Chartered, and SMIC in the dust, and they were all established foundries with track records.

Its not exactly unfathomable to consider the possibility of TSMC doing it again.

What happened to Transmeta? Centaur Technology? Cyrix?

All had great momentum only to quickly wane in marketshare and revenue as they simply couldn't afford to keep up with the R&D budgets of their vastly larger competitors.

Yes GloFo is growing fast in terms of percentage, smaller companies can do that more easily than larger ones, but R&D budgets are dictated by absolute dollars and not percentages of growth, and that is why they are lagging TSMC in node-release timeline.

How much of GloFo's revenue is from 28nm? How much of TSMC's revenue is from 28nm? That alone determines how much budget can reasonably be expected to be pumped into 20nm and 16/14nm development.

 

[pablo87]

Intel will see there is no profit with their cost structure competing with arm. Its was like IBM competing with Intel.
They will therefore have to use their fab capacity for other customers.
In a market where tsmc is accelerating GF is just plain doomed with owners that dont know anything about the business.
That means mubadala will dump amd.
The relationship is bad for all.
In that situation perhaps 5 years from now, big cores from amd like kaveri is probably profitable on tsmc process but i think its going the arm way anyway. Amd might just trim the organization further and just stop kaveri like products - if they are allowed by mubadala.
If they are not allowed i would just delay it

Its a mess...

Great post. The sooner the better re. Mubadala liquidating GF.

 

[TerryMathews]

That said, I wouldn't give up on big core. While Intel has a big lead, they have been so focused on IGP and SOCs, they've lost focus on what got them here. After so many years, they don't even have a desktop quad core under $100

You're forgetting one thing: Intel has a number of chips in their product line that are feature reduced intentionally to segment their markets.

If they needed a Pentium-label quad core to remain competitive, all they would have to do is stop lobotomizing them quite so much. It's not like they need to R&D a new chip. Just cripple it a little less.

 

[pablo87]

You're forgetting one thing: Intel has a number of chips in their product line that are feature reduced intentionally to segment their markets.

If they needed a Pentium-label quad core to remain competitive, all they would have to do is stop lobotomizing them quite so much. It's not like they need to R&D a new chip. Just cripple it a little less.

Yes they have the capability so why haven't they done it? The problem runs much deeper.

 

[TerryMathews]

Yes they have the capability so why haven't they done it? The problem runs much deeper.

They haven't done it because they don't have to. Intel's development hasn't just outstripped the competition, it's outstripped demand.

In the past, that lack of urgency led to Intel getting overtaken. Time will tell if that will happen again.

 

[krumme]

Great post. The sooner the better re. Mubadala liquidating GF.

I guess Mubadala is some sort of political controlled looking by the crazy behavior and the fantacy ppt gf provides for the emperor.
If thats the case differentiating between revenue and profit can sometimes be a challenge. GF generate a lot og big numbers.
But at some time the political focus changes quicly.
I dont see the political interests in GF now:
The ambition to build a fab the worst place on earth outside of the poles have ended. What gf provides now is just big numbers until a new playing field is found.
I can see better use of the money from a political perspective. I think its a disaster in all dimensions.

 

[krumme]

Getting something to grow in 50c is hopeless but here is an alternative proposal for the gf waste of money.
Ditch gf. Keep 51% amd and get someone big on board that knows business fx samsung.
Build a amd team over the next 10 years for some specific part of design or preferably coding. Fabs demand to many different competences. It needs to be simple and focused on the competence level.
Build colege and adapt a univercity to that specific target. Use 15 years to get there. Invest heavily on large concentrated research groups on both colege and univercity level.
The spinoff will probably end somewhre else, but the start focus keeps it on same track and gives a unique position.

 

[SiliconWars]

The added hyperbole doesn't do your post justice, you'd have better reception within the community of you dialed that back.

I can only read so much short-termism before it begins to grate heavily on my nerves. Half the reason they whole planet is in such a mess is because of it.

Getting to the meat of your post, people are right to be reserved and pessimistic regarding GloFo's prospects. History in the sector is why.

TSMC left UMC, Chartered, and SMIC in the dust, and they were all established foundries with track records.

And in 4 short years GF has overtaken those as well, making themselves clear #2. What else were they supposed to do?

Yes GloFo is growing fast in terms of percentage, smaller companies can do that more easily than larger ones, but R&D budgets are dictated by absolute dollars and not percentages of growth, and that is why they are lagging TSMC in node-release timeline.

Here is the problem. What else are they supposed to do? What were people expecting? That TSMC be supplanted as the top foundry in under 3 years? How was GF going to do that seeing as it takes 3 years to build ONE fab?

TSMC are spending ~$1.4 billion on R&D this year. GF are spending $2 billion on one new R&D facility this year, but again it takes time to build. They are planning a 450mm $10 billion+ fab in NY. Yet people are still under the belief that they are going to drop out at the first sign of trouble? They didn't even start the 32nm node they are left with (and which 28nm is heavily based on), yet people think they are going to drop out because they are a year behind TSMC?

GF are at the stage of building the infrastructure to take on TSMC - they can't just magically turn bricks and morter into actual buildings overnight however.

How much of GloFo's revenue is from 28nm? How much of TSMC's revenue is from 28nm? That alone determines how much budget can reasonably be expected to be pumped into 20nm and 16/14nm development.

If GF was being funded by their own revenue then yes, it would be a problem. When the government of Abu Dhabi has a vested interest in the company it makes it somewhat different to the others like UMC and SMIC who are trying to do that.

These guys live in a life of luxury because of oil that is running out. They know what the replacement for oil is - microchips. Believe me they are all highly motivated to keep their lifestyles and will spend staggering amounts of dollars right now to ensure they continue it in the future. I think having a plan that lasts longer than 3 years is why most Westerners can't get their head around why it can possibly work.

 

[NTMBK]

They know what the replacement for oil is - microchips.

So instead of today's oil wars, in the future we will have... silicon wars?

 

[MisterMac]

Silicon:

That also means AMD short of creating a miracle ARM product or x86 product within the next few years will eventually bankrupt or become a VIA.
GF Owners are purposely bleeding it dry - just to take less of a financial hit.

If your analogy of longterm playbooks are correct.


One could argue they wouldn't need AMDs petty revenue - but clearly they're taking all they can get.

 

[SiliconWars]

Silicon:

That also means AMD short of creating a miracle ARM product or x86 product within the next few years will eventually bankrupt or become a VIA.
GF Owners are purposely bleeding it dry - just to take less of a financial hit.

If your analogy of longterm playbooks are correct.


One could argue they wouldn't need AMDs petty revenue - but clearly they're taking all they can get.

AMD's revenue isn't petty, it's a quarter of GF's entire revenue. They are not bleeding AMD dry either - they are offering AMD what is almost certainly very competitive rates in the industry. It's AMD who can't fulfil their side of the agreement.

When GF bought Chartered they got Chartered's customers and revenues (they were losing money btw). Buying AMD's fabs got them AMD's custom. The important thing was to expand the fab business but not have a massive income hole while doing so. That's why they bought fabs that already had customers, and that's why they still have the same customers while they attract new ones.

So GF has AMD's and ex-Chartered revenues, both of which were not enough to turn a profit for either company. GF are not magicians who can turn two unprofitable companies into one profitable company overnight. But all the time they are expanding while not losing huge amount of cash - this is important because they still need to show results to the Abu Dhabi government in order to continue to get investment. These guys aren't fools, they aren't going to throw money at bad investments.

There is not another company in the world who would have done it differently to the way GF is currently building their business. GF has a vested interest in keeping AMD afloat because they are dependent on AMD's revenue's (currently) in order to make their own numbers. It's guaranteed revenue, but they can't give wafers away for free, all they can do is hope that AMD gets their act together and keeps their side of the agreement as best as they can. There is no chance of them bleeding AMD dry because once again - that would be a short term benefit with bad long-term consequences. How do you think Abu Dhabi would react if GF's revenue suddenly dropped by a quarter in a year? If AMD goes under what benefit would that be to GF?

 

[mrmt]

How much of GloFo's revenue is from 28nm? How much of TSMC's revenue is from 28nm? That alone determines how much budget can reasonably be expected to be pumped into 20nm and 16/14nm development.

Another point is how much GLF 28nm still demands in R&D and how much TSMC 28nm still demands in R&D. I doubt that TSMC is spending too much on it, while GLF has yet to bring any meaningful 28nm product to the market.

 

[SiliconWars]

The seemingly slow-moving nature of the industry is probably the main reason why people have getting their heads around it, but it's just not possible to make the kind of financial and technological advances that are being demanded of GF in such a short space of time.

Some of you might enjoy this - http://www.youtube.com/watch?v=ed1WexCZgFs

As you can see the guys at GF are not sitting on their asses smoking cigars while counting AMD's revenue. They have 20nm silicon working and ready but they just can't start mass manufacturing of it until they actually have the fab ready. It's coming, there's no chance of GF not making 20nm, or 14nm after that. You just have to give them the time that it naturally takes to make it happen.

 

[SiliconWars]

TSMC's 28nm should be ~30% of their revenue this year - http://www.taipeitimes.com/News/biz/archives/2013/06/06/2003564057

GF's 32nm would have been ~25% of their revenue last year.

Another thing that most people (including myself) keep forgetting is that GF were the first in the foundry business to ship HKMG. God only knows what happened between 32 and 28 but it was probably either a serious problem with gate-first (which again was not a fault of GF's - gate-first was planned long before GF or ATIC existed), or it was because of a planned transition to bulk. Either way there are obvious reasons for why GF's 28nm was late, but that doesn't have to mean that 20nm will be late, or 14nm - the real comparison should be GF's 32nm to TSMC's 28nm, and GF was 6 months ahead.