How Will AMD stay alive?

[jennyh]

Only kidding. I know it wasn't useless.

I'm okay with JennyH being skeptical on this one. After all, no one ships SRAMs as products. How are we to know how many of those SRAMs are functional, even if there are the 2.9 billion cells per die? The rules of evidence, though, should be company-agnostic and you can therefore cast doubt upon the IBM claim to have a working cell a year ago. Did they demo it? To whom?

I will say, though, that this is exactly where Intel wanted to be at this point in time, and from this Tick-Tock looks to me (from an admittedly non-Fab background) to be right on track.

When IBM has 28nm (or even 32nm) ready to ship as revenue, perhaps this will be a more interesting conversation. Until then... well, we'll see whether Ivy Bridge or Bulldozer Shrink hits market first. Or even in the same year.

Good of you to be cautious and reasonable about it archi.

I could easily say, when intel has a production cpu better than 45nm, maybe we'll start talking about 32nm then. At this very moment you don't. Is it imminent? It sure looks that way but it's not here is it?

What I'm trying to do here is show that the gap has almost certainly closed. Perhaps not at 32nm, but at 22nm IBM and AMD were capable of doing something they weren't capable of doing before.

1 year ago it was only a few SRAM cells. A few days ago, intel have a couple of billion of them on a wafer. It's a lot closer than 45nm was or 32nm is likely to be, that's all I'm saying.

 

[archibael]

I could easily say, when intel has a production cpu better than 45nm, maybe we'll start talking about 32nm then. At this very moment you don't. Is it imminent? It sure looks that way but it's not here is it?

Sure, you could say that. Although having engineering samples in the hands of customers and demoing the actual running of real software would tend to lend credence to those who assert it's ready to ship.

What I'm trying to do here is show that the gap has almost certainly closed. Perhaps not at 32nm, but at 22nm IBM and AMD were capable of doing something they weren't capable of doing before.

1 year ago it was only a few SRAM cells. A few days ago, intel have a couple of billion of them on a wafer. It's a lot closer than 45nm was or 32nm is likely to be, that's all I'm saying.

I suppose that all depends on how fast the "single SRAM cell" --> "large numbers of SRAM arrays on a wafer" transition is.

If it's a year, IBM and the consortium may have caught up to some significant extent. If it's significantly longer than that, then they haven't. I don't know when we had a 1270 SRAM cell working. Could have been three years ago for all I know. Could have been last quarter. [shrug]

 

[jennyh]

I'm not disagreeing with that and I'm 100% sure that it is almost ready for shipping. Until it does though, we can't treat it as anything except speculation.

IBM could have 32nm engineering samples ready as well, it would be pretty amazing if they don't tbh.

 

[jennyh]

I suppose that all depends on how fast the "single SRAM cell" --> "large numbers of SRAM arrays on a wafer" transition is.

If it's a year, IBM and the consortium may have caught up to some significant extent. If it's significantly longer than that, then they haven't. I don't know when we had a 1270 SRAM cell working. Could have been three years ago for all I know. Could have been last quarter. [shrug]

The ability to move to 22nm isn't all about intel's capability to get there though. It is depending on others ability too, ie Nikon and Canon etc.

I'm sure if you hunt around a bit, somebody down in lithography should be able to update you on what the score is with Nikon, I believe a deal was struck there a year ago or so?

Either way, the tools need to be available, and intel sure aint manufacturing lithography tools 3 years in advance of those who make a business out of it.

 

[archibael]

Last I heard IBM's nowhere near ready to ship 32nm, but they could surprise us all.

I can't treat Clarkdale as speculation, or else I would have missed watching Dancing With The Stars this week. YMMV.

 

[yomamafor1]

The purpose of SRAM vehicle is not about whether it works or not, but rather the transistor density a particular process node can cram in onto a space. Although Intel historically has the best process node in the industry, we don't know about the 22nm process node's performance until the next IEDM.

And it looks like Intel's SRAM chips do indeed work:
http://www.gizmag.com/intel-unveils-worlds-first-working-22nm-chips/12936/


And for the record, AMD's process was developed by a consortium led by IBM, so AMD actually did not actually actively take part in it. The process was first developed by IBM in East Fishkill, then move to AMD's fabrication plant in Dresden for optimization. Now that AMD has spin off the fabrication plant to Globalfoundries, there's no way to know when they will adopt the 22nm process (maybe in 2~3 years).

 

[randomizer]

Have we actually answered the question "Is AMD doomed?" yet?

 

[jennyh]

That's what I mean, IBM could be about to surprise us all.

See, the whole IBM alliance is a much tighter knit unit. Look at AMD's 5870 for example of how to keep something under wraps. Sure the odd benchmark came out, and some early indications of its shaders and clocks weren't all that far out.

But who knew about eyefinity before september 10th? Did you archibael? I sure as hell didn't and I haven't spoken to anyone who had even the faintest indication that it was gonna happen.

Did Nvidia know about it? We will know the answer to that when the g300 finally gets released.

Last I knew, AMD were nowhere near ready to move gaming from one screen onto 3 in the blink of an eye, but it seems they did just that without anyone knowing about it beforehand. Just because you aren't fed up reading about it doesn't mean it's impossible, it can also mean it's been kept secret...somehow...in this most difficult of all industries to keep anything secret.

 

[TechnologyCoordinator]

True, they sell them and call them Barcelona.

Oh snap.





....I mean.... WOW!

 

[Upendra09]

@ random not reallly

 

[jennyh]

Have we actually answered the question "Is AMD doomed?" yet?

No that's being reserved for a future thread. Tomorrow is my best guess, but until it has been released I couldn't possibly say for sure. 😀

 

[JAYDEEJOHN]

Thing is, unless it starts really paying off having more cores, if BD has a similar IPC to Intel, it doesnt and wont matter how much of a lead Intel may have for next node, since really, thats not where their true lead is anyways.
Oh sure, 4-8%, but thats all, everything else is pure IPC.
CPUs are too slow moving for real impacts per node swith, unlike gpus, where 10% is considered the norm, or double the perf, and thats just a dumb shrink, not new arch

 

[fazers_on_stun]

AhAint no 'gaurantees' in this game

Hmm, that's funny - you seem to make "gaurantees" pretty frequently.

I can see it now - my new Michelin tires are "gauranteed" by jennyh to last 100,000 miles, or 20 ft. outside the shop, take yer pick! 😀

BTW, I see from TC's link ("ownership" tab) on AMD financials that Rivet & a bunch of other AMD senior personnel are uniformly selling their AMD stock - every last one of them! What does that tell you?? Hint: grab while the grabbin's good!!

 

[TechnologyCoordinator]

Have we actually answered the question "Is AMD doomed?" yet?

Yes.

 

[Upendra09]

what do you mean by future thread?

 

[fazers_on_stun]

And in case you want some actual facts about the original topic of "Is AMD Doomed?", here are financial analysts estimates for AMD's 3Q 2009:

Most positive Q3 estimate: -.36 per share (hint, the minus sign isn't good)
Most negative Q3 estimate: -.58 per share
Average Q3 estimate: -.45 per share
Q2 2009 Actual: -.53

And I know you don't like it when people list sources and back up facts and opinions, but here's the SOURCE: http://investing.businessweek.com/research/stocks/earnings/earnings.asp?ric=AMD

Interesting - I hadn't seen that before. If the middle estimate holds true, another $300M loss. It would be much greater of course if AMD still had any ATI goodwill to write off, but they spent that dime already 😀.

 

[jennyh]

BTW, I see from TC's link ("ownership" tab) on AMD financials that Rivet & a bunch of other AMD senior personnel are uniformly selling their AMD stock - every last one of them! What does that tell you?? Hint: grab while the grabbin's good!!

Last time you said that AMD tripled in 6 months 😀 😀

 

[archibael]

Oh snap.





....I mean.... WOW!

It was a cheap shot, and I regret it somewhat.

I'll try not to do it again. Often.

 

[fazers_on_stun]

AHA, I think fazers has a skeleton in his closet too TC
http://technofazer.blogspot.com/2008/09/industrys-first-true-quad-core-x86.html

LOL - that's my evil twin brother . I'd have set my phaser to vaporize with him, but our Mom would probably object 😀. She never let us run with scissors, play with sharp objects, etc - real spoilsport! [:thegreatgrapeape]

 

[Major7up]

AMD is not doomed, but they may evolve into a different market segment that they have historically and compete less directly with Intel. I could see AMD becoming a 3rd party chipset maker like VIA though I would prefer to see them continue making actual CPUs to compete with Intel. But I certainly don't see them going away, even if they declare bankruptcy, they will survive in some fashion.

 

[jennyh]

Fazers if you had a twin brother I'm sure he'd be saying you were the evil one lol 😀

 

[TechnologyCoordinator]

Last time you said that AMD tripled in 6 months 😀 😀

I don't deny that AMD stock has done well for people that have bought at the low point, but don't forget about how the people who bought at $10, $20, $30, and even as high as $40 feel, not so good.

However, if AMD survives (and I hope they do), they would be a most excellent investment. A return to profitability would more than likely send stock prices into the teens (not a dramatic jump, but a slow return to profitability I feel would send the stock prices up to that point).

Of course, the above paragraph is pure speculation, but the top paragraph is nothing but historical stock data.

 

[TechnologyCoordinator]

even if they declare bankruptcy, they will survive in some fashion.

I agree.

Although there might be issues with the x86 cross license agreement. However, it is in Intel's best interest to be 'flexible' with it in the event of an AMD bankruptcy. Weak competition is better than no-competition and increased government regulation.

 

[jennyh]

When ATIC and AMD formed GlobalFoundries they had a plan. In order to dominate the cpu industry, you have to own x86. Intel weren't gonna just give it up were they?

Through AMD, ATIC can take control of x86, and therefore the world cpu market. And only through AMD can that happen. There is no other way of doing it except beating intel at their own monster, and only AMD can do that and have the history of doing it.

That's pure speculation of course, but it must feel nice to have oil barons billions behind you when you set out to achieve something. Who else except AMD can achieve ATIC's goals?

Go on, answer it.

 

[yomamafor1]

:lol:

AMD's x86 license was actually given by Intel. So Intel can just yank the plug, and AMD can produce x86 chips no more. (although Intel wouldn't do that because it will be slapped with a huge "monopoly" banner).