The AMD Execution Thread [2007 - 2017]

[3dilettante]

Regardless, I do not expect a magic turnaround if Meyer is put in charge.

Ruiz may have been the captain of the ship, but the officers of the processor division were the ones who charted the course.

He does not appear to have the in-depth knowledge of AMD's products (he's on the record overestimating Barcelona's transistor count by about 50%) to show he took it upon himself to weigh the risks of AMD's process and processor transitions.

Somebody sold him on the path AMD took.
Who would that be but the one everyone now wants to run the company?

 

[Bouncing Zabaglione Bros.]

Problem is that AMDs current state is down to decisions they took a couple of years ago. They didn't come up with sucessors to Athlon and Opteron when their development window said they had to. Now they are up against Intel who themselves took a wrong turn and came back, and it took Intel a couple of years even with all the resources and money that AMD doesn't have.

Sure, you can sack the guys who steered the company to where it is now, but that doesn't dig you out of the hole AMD are currently standing in, and who knows how long it will take them to chase down Intel's Penryn and Nehelim over the coming year?

 

[ShaidarHaran]

I didn't mean to imply that sacking Ruiz was the magical cure-all solution, but *something* needs to happen and I think they need to start at the top. I don't expect AMD to return to competitiveness until 2009 at the earliest. Which is sad, because they weren't competitive at all this year (again, except at the low-end) and the same is true for most of 2006 as well.

 

[3dilettante]

AMD's only apparent chance at catching up in the processor race is Bulldozer, or an unexpectedly faster-clocking Shanghai.

The process decisions made by IBM and AMD are an unknown when it comes to the effects on performance and manufacturing.

That being said, Intel has taken the chance at a significantly revamped 45nm that has cut down on some very significant barriers to scaling and leakage control.

IBM and AMD have claimed their process can manage the same thing, but both have supplied little in the way of data to substantiate it.
AMD's silence in recent times has always been a better indicator than any PR slop it puts out.

For that reason, and the fact that Shanghai is a derivative of Barcelona, which is a derivative of Sledgehammer which is a derivative of the original K7, it doesn't look good for any huge gains with AMD's first 45nm products.
This is reinforced by the expectation that AMD's initial 45nm process will not have all the bells and whistles that it will upon maturation.

That leaves Bulldozer, a hopefully refactored design that isn't hobbled by signaling and variation problems that have killed K10's clock scaling.
Without more data, it's hard to see how well AMD's effort will pan out.

AMD's slides on Bulldozer emphasize HPC and indicate targeted performance is high for those sorts of applications.
Single-threaded is expected to be improved, though there's no way to judge how incremental such an improvement is from unitless arrows on a PR fluff slide.

 

[Arty]

Was this posted yet?

Would Nvidia Buy AMD?

 

[Geeforcer]

Regulatory hurdles aside, buying AMD right now is akin to moving in with an Ebola sufferer.

 

[AlexV]

Regulatory hurdles aside, buying AMD right now is akin to moving in with an Ebola sufferer.

They'll steal the sufferer's clothes and wife once hes past the point of no return though(as in I think that nV will care snagging a lot of things off of AMD if it goes further down the drain, as it will certainly be cheaper at that poing, and they may need to go into the CPU business at some point, and there's no alternate better key for that business-unless you think that they can buy Intel )

 

[hoho]

NV could surely use those FAB's. Too bad it can't use x86 license as it gets revoked if AMD gets sold.

Though I wonder what would happen if NV buys, say 33% share in AMD. Will license still apply?

 

[AlexV]

NV could surely use those FAB's. Too bad it can't use x86 license as it gets revoked if AMD gets sold.

Though I wonder what would happen if NV buys, say 33% share in AMD. Will license still apply?

I'm not sure about the FAB's...maybe they would, maybe they wouldn't want those, as they bring their whole slew of problems and would be quite expensive. The intellectual property and the engineers?Hell yeah. As for the x86 license, Intel can be bitchy about it, but in the end they can't truly placate nV IF nV actually decides to get one, so that's not a huge issue in itself. The question is when(not wheter, as I think that it's unavoidable in the end) will nV decide to get into CPUs.

 

[MTd2]

If you can buy fractions of a company, what is to buy a company anyway when it comes to licenses?

 

[dizietsma]

http://www.digitimes.com/systems/a20071121PD215.html

Looks like confidence in AMD's 45nm move is not very high and who can blame them, the 65nm node has proved tricky enough.

Is SOI running into a dead end ?

 

[MTd2]

Is SOI running into a dead end ?

AMD probably won´t use SOI for 45nm. IBM, which is doing a partnership with AMD for lower nodes, anounced at the same time earlier this year the use of High-K , based on a similar, but not the same, Hafnium composite that Intel is using now.

http://www.pennwellblogs.com/sst/eds_threads/2007/06/070615-ibm-hkmg-gate-first-processing.php

http://www.nanowerk.com/news/newsid=1404.php

http://www.semiconductor.net/article/CA6482837.html?nid=3660

http://www.dailytech.com/article.aspx?newsid=5885

 

[Bouncing Zabaglione Bros.]

http://www.digitimes.com/systems/a20071121PD215.html

Looks like confidence in AMD's 45nm move is not very high and who can blame them, the 65nm node has proved tricky enough.

I dunno, there can always be surprises. Look at the disaster that was ATI's R600 at 65nm compared to it's great RV670 at 55nm. Hopefully these guys are clever enough to learn from their mistakes.

 

[chavvdarrr]

I dunno, there can always be surprises. Look at the disaster that was ATI's R600 at 65nm compared to it's great RV670 at 55nm. Hopefully these guys are clever enough to learn from their mistakes.

is it so great? G92 is bigger, with more trannies, on bigger process, yet at full load dissipates less heat and its faster! (ok, less heat per card, more per chip due to different memory chips, but never-the-less...)

 

[Bouncing Zabaglione Bros.]

is it so great? G92 is bigger, with more trannies, on bigger process, yet at full load dissipates less heat and its faster! (ok, less heat per card, more per chip due to different memory chips, but never-the-less...)

If you'd made prejudgements of ATIs success at 55nm based on it's problems at 80nm, you'd have been very wrong. That's pretty much what Digitimes is doing by implying strongly that because AMD has had problems with it's 65nm process for CPUs, so its likely to have problems with it's 45nm process for CPUs.

If anything, I'd suggest that there are so many issues with AMD's 65nm, they've got no choice but to be better at 45 nm - they can hardly be worse!

 

[dizietsma]

AMD probably won´t use SOI for 45nm. IBM, which is doing a partnership with AMD for lower nodes, anounced at the same time earlier this year the use of High-K , based on a similar, but not the same, Hafnium composite that Intel is using now.

http://www.pennwellblogs.com/sst/eds_threads/2007/06/070615-ibm-hkmg-gate-first-processing.php

http://www.nanowerk.com/news/newsid=1404.php

http://www.semiconductor.net/article/CA6482837.html?nid=3660

http://www.dailytech.com/article.aspx?newsid=5885

That's interesting because the link below says that they will not use this till 32nm and will rely on SOI for 45nm

http://www.fabtech.org/content/view/3868/

I wonder who is right?

 

[MTd2]

If you'd made prejudgements of ATIs success at 55nm based on it's problems at 65nm, you'd have been very wrong. That's pretty much what Digitimes is doing by implying strongly that because AMD has had problems with it's 65nm process for CPUs, so it's likely to have problems with it's 45nm process for CPUs.

If anything, I'd suggest that there are so many issues with AMD's 65nm, they've got no choice but to be better at 45 nm - they can hardly be worse!

Exactly, somewhere above I stated that the technologies implemented on production are not always incremental, but can be result of an independent material research that will be integrated on a higher level of technology. Take nanowire or nanotubes. There no industrial application on chip prodution right now. They are independently researched. And anyway, whenever they are applied, it is not possible right now to predict right now how difficult it will be to get good yields.

 

[ShaidarHaran]

Regulatory hurdles aside, buying AMD right now is akin to moving in with an Ebola sufferer.

They'll steal the sufferer's clothes and wife once hes past the point of no return though(as in I think that nV will care snagging a lot of things off of AMD if it goes further down the drain, as it will certainly be cheaper at that poing, and they may need to go into the CPU business at some point, and there's no alternate better key for that business-unless you think that they can buy Intel )

If NV buys AMD I will declare a jihad against them. First 3dfx, then AMD & ATi = NV must die.

 

[_xxx_]

It's not like nV is responsible for the failures of other company's management though. If they did better (3dfx), they would still be around.

 

[ShaidarHaran]

It's not like nV is responsible for the failures of other company's management though. If they did better (3dfx), they would still be around.

No one ever said fandom had to be logical.