The AMD Execution Thread [2007 - 2017]

[AlexV]

Probably because:
a)they don't have an architectural advantage(Pellerin is making up shit, they don't have a process technology advantage nor do they have a CPU architecture advantage, so that their 45nm will equal competitor's 32nm is pure rubbish)

b)Their 65nm is poorer in some aspects compared to what they had at 90nm, and the trend continues with the 45nm process being worse on some accounts than the 65nm one...worse enough to raise questions WRT whether or not it'll actually enable them to clock better then they are.

c)Solidly in development means about jacksquat when it comes to AMD-they've been solidly developing 65nm for quite a while and it's fairly lackluster, they've been solidly developing 45nm for quite a while too and they still haven't got samples back from the fab, although they were claiming they'll have 45nm parts out in 2008 and, generally, lately they always are solidly developing something that turns out underperforming at best, horrid at worse.

Summing up the above, and correlating it with what seems to be a turn towards more sincerity and less wanking with a healthy dose of realism added(see their latest analyst day materials....lackluster?yes, but at least those actually seem to be in-line with what AMD can do ATM and in the forseeable future), harping on about 32nm certainly would have no perceivable purpose, IMHO.

 

[3dilettante]

I don't want to parse Rivas's words to finely, but he only said they had 32nm advance work in SRAMs, not that they had 32nm SRAMs.
It's sort of like saying a car manufacturer has advance work in a prototype car, in that it has a transmission put together but nothing else.

PR speak is weasily anyway, though the language is not any more imprecise than any one-off comment would be.

I'm only bringing this up because of the wording of a story concerning IBM and AMD's developing separate 32nm SOI processes based on the common 45nm process.
http://www.semiconductor.net/article/CA6510988.html

IBM has demonstrated a 32nm SRAM for its own process.
AMD, if the article is not a mistatement, has not as of 12/9/07.

 

[MTd2]

IBM has demonstrated a 32nm SRAM for its own process.
AMD, if the article is not a mistatement, has not as of 12/9/07.

From that article: IBM and Advanced Micro Devices (AMD, Sunnyvale, Calif.) separately are developing a silicon-on-insulator (SOI) 32 nm technology, based on the HKMG gate-first solution developed for the 45 nm generation.

Does it mean that IBM process =AMD process, but different from the others?
Or does it mean that IBM process = others, but different from AMD ?

 

[INKster]

THG's got a curious take on the subject of possible merger bids:

http://www.tgdaily.com/content/view/35289/118/

 

[Geeforcer]

All together now: Cross-licensing agreement violation. I am stunned that supposedly competent analysts and journalists still manage to miss that key point and sit there musing why they are the only ones brilliant enough to envision the buyout scenario.

 

[Arun]

THG's got a curious take on the subject of possible merger bids:

I read until they said NVIDIA had unique 'simulation software' that allows them to always ship on first silicon. Then I rolled my eyes, glanced through the other paragraphs for 5 seconds, and closed the page.

They also fail to consider that when you buy a company, you need to pay more than its current market capitalization. Otherwise nobody is going to accept your offer. NVIDIA could buy AMD through a mostly-stock deal, and the resulting company would actually be cashflow-positive given NV's excellent cash generating position right now.

But that doesn't make it a viable proposition. Something else to consider is that NV's management seems to be guiding for substantial revenue contributions from MCPs/IGPs for Intel CPUs. It is not clear to me that they could keep that market if they acquire AMD. And then also, I'm not sure selling the GPU division is as easy as it sounds.

Spinning it off might have a fair amount of problems (both legal and practical) and besides Intel, I'm not sure who'd actually be interested in buying it. Obviously, selling it to Intel isn't a very good idea for strategic purposes - heck, I'd argue that even if it was free for NV to acquire AMD, if the condition was to sell the GPU unit to Intel, then it'd still be a better idea not to do anything.

I wouldn't exclude IBM and Samsung though. Both would have good reasons to be interested in most of AMD's business units. I do question the legal issues with Samsung buying the company though, but besides that it sounds viable to me. Which reminds me: Samsung also manufactures logic chips based on IBM's process. Who knows, maybe AMD's asset-light (aka 'asset-smart', ugh) strategy is somehow reliant on other companies in the alliance.

I'm so incensed, I'm hoping for an AMD apologist to post here so I can debunk any remaining excuses point by point.

Haha. Well, I'm not really an AMD apologist as most of you must have figured out by now (more of a Centaur apologist if anything, perhaps!) but I'll make a shot anyway:

1. Rivas didn't feel to me as if he had a perfect understanding of the business; his statements on chipsets, platforms and graphics were incredibly generic in the Q&A and so forth - so I wouldn't be surprised if the graphics/chipset roadmaps he gave are pretty much worthless. Although I'd still give a 60%+ probability they aren't, just a tad approximative and overly tied to platform plans.
2. Hector Ruiz said, as part of his ending speech, that they promised to 'underpromise and overdeliver'. There are two possible interpretations here: either they made their roadmaps much less aggressive to make sure to deliver, or they now have backup projects in place to be sure to always have fresh products on the market even if others are delayed.
3. If it's the latter, then we very well might see Bulldozer in 2009 after all, just that they're no longer willing to guarantee so. However, it does seem to me that 'Swift' is likely to be the notebook platform anyway unless it becomes obvious Bulldozer is on schedule really soon, because those need to be planned well in advance.

 

[MrGaribaldi]

But would that agreement stop a company from buying a controlling stake in AMD? and then taking control of the board?

 

[ShaidarHaran]

But would that agreement stop a company from buying a controlling stake in AMD? and then taking control of the board?

That would qualify as "buying the company".

 

[MrGaribaldi]

Ah, ok. To me having a 51-67% stake wouldn't be "buying the company", but rather a heavy investment. So tanks for the clarification (And too bad for AMD)

 

[MTd2]

Ah, ok. To me having a 51-67% stake wouldn't be "buying the company", but rather a heavy investment. So tanks for the clarification (And too bad for AMD)

Beware! Buying a significant share of AMD can cancel the cross lincenses with Intel.

 

[INKster]

Beware! Buying a significant share of AMD can cancel the cross lincenses with Intel.

That Abu Dhabi investment fund did (8.1% of AMD is still a significant amount of money), and yet nothing happened to the licenses...
The key aspects of the old agreement with Intel are still confidential, i think, so making that assumption without full knowledge of the matter may not be wise.

 

[MTd2]

That Abu Dhabi investment fund did (8.1% of AMD is still a significant amount of money), and yet nothing happened to the licenses...

Sure, but he said a higher amount, like more than 50%. The quantities are kept secret but certainly they must be below 50%.

 

[ShaidarHaran]

Sure, but he said a higher amount, like more than 50%. The quantities are kept secret but certainly they must be below 50%.

I believe the agreement states that if ownership of the company changes hands the agreement is null and void.

 

[mushroom]

Let's see: JDSU, Nortel, AOL, McClatchy, Corning, ...
And recently: Citigroup, Morgan Stanley, Merril Lynch, UBS, ... Those financials are all writing down loans pretending they'll have a value of zero even if, after the credit crisis, there's still a possibility of getting something for it, but buy getting them off the books entirely, they can only provide upside. (Note that many of those companies saw their stock rising after the write-down announcement.)

If you've never seen this kind of action before, you're really not following financial news...

actually, there will be more and more write downs coming from those same financials... i would tend to believe they are 'conservative' in their writedowns so far...

 

[3dcgi]

I read until they said NVIDIA had unique 'simulation software' that allows them to always ship on first silicon. Then I rolled my eyes, glanced through the other paragraphs for 5 seconds, and closed the page.

With tech like that they could put a lot of software companies out of business.

 

[MTd2]

R700 is 2008

No delays


We got a confirmation that R7xx generation is on schedule for 2008 and it won't be shifted to 2009 as many journalist believes. Many journos made the assumption based on Mario Rivas presentation at Investor days and Mario was talking about platforms and showed the slide with R7xx based platform in 2009.

According to current plans R7xx parts in a platform are scheduled for 2009. This doesn’t mean that the discrete products will be late, but at least toady it looks that ATI can delved them on time, at least at some point of 2008.

http://forum.beyond3d.com/showthread.php?t=41627&page=12

 

[ShaidarHaran]

I'll believe that when I hear it from a source that has at least some shred of credibility, and that isn't just quoting Fraud.

 

[Geo]

Faud has switched to calling it "R7xx" now, eh? Probably a good idea.

 

[3dilettante]

Rivas didn't feel to me as if he had a perfect understanding of the business; his statements on chipsets, platforms and graphics were incredibly generic in the Q&A and so forth - so I wouldn't be surprised if the graphics/chipset roadmaps he gave are pretty much worthless. Although I'd still give a 60%+ probability they aren't, just a tad approximative and overly tied to platform plans.

Is it a good sign that the executive vice president of AMD's Computing Solutions Group has an imperfect understanding of his department's product plans?

Besides, what are the odds that Rivas made those slides?
If he has such an imperfect knowledge of the business, he would have fobbed the job off to someone in his department who would have talked to the technical heads.

If those slides are an approximation of someone's knowledge of AMD's product plans, they aren't necessarily those of the presenter's.

Hector Ruiz said, as part of his ending speech, that they promised to 'underpromise and overdeliver'. There are two possible interpretations here: either they made their roadmaps much less aggressive to make sure to deliver, or they now have backup projects in place to be sure to always have fresh products on the market even if others are delayed.

If it's the latter, then we very well might see Bulldozer in 2009 after all, just that they're no longer willing to guarantee so. However, it does seem to me that 'Swift' is likely to be the notebook platform anyway unless it becomes obvious Bulldozer is on schedule really soon, because those need to be planned well in advance.

The Montreal chip is a possible spanner in the works.
If the Bulldozer server variant is not delayed, Montreal would need to exist for all of one quarter.
The additional cache and porting of a K10 core to the new socket is enough of a design effort to point towards a delay for Bulldozer.

The fact that Montreal is listed on the new socket also removes some of the possibilities for a Bulldozer delay.
It's obviously not the external system architecture holding things back, otherwise Montreal would be late as well.
It's either the core design or something on the other side of the HT and memory controller, leaving either the cores themselves or the part of the northbridge/uncore that they share.

If Bulldozer shows in 2009, AMD's own product decisions push it back to something closer to how Barcelona was supposedly available in 2007.

If the delay is upwards of six months, it sounds like a replay of the K8 replacements that weren't.

 

[INKster]

they promised to 'underpromise and overdeliver'

Did someone else spot any paradoxical claims in this sentence... ?