is AMD's native quad core feasible?

[wickedmonster]

Judging from how poorly AMD's 65nm transition has been, I wonder if it was economically feasible for them to do a quad core. Obviously, it's going to be a huge die, yields are going to be low, and the costs are going to be astronomical. I wonder if there is a market for it. Once Intel releases their 45nm quad core Penryn, they can undercut AMD in terms of cost. AMD won't be able to match. And if Intel's Penryn is faster, then AMD is going to be in alot of hurt.

 

[makaka]

yes the brisbane was bad experience for amd. but i hope that situation will change with time.in any case if the new K8L will be lower in performance than current conroe than amd can say bye bye to the big amount of market share

 

[ajfink]

Judging from how poorly AMD's 65nm transition has been, I wonder if it was economically feasible for them to do a quad core. Obviously, it's going to be a huge die, yields are going to be low, and the costs are going to be astronomical. I wonder if there is a market for it. Once Intel releases their 45nm quad core Penryn, they can undercut AMD in terms of cost. AMD won't be able to match. And if Intel's Penryn is faster, then AMD is going to be in alot of hurt.

What? They've had platters of the things on display for months and have a lot more production capacity now than they have had before. Yeah, the dies are huge, no doubt, but if performance is as AMD claims it won't be a problem, people will buy them for whatever semi-reasonable price AMD wants to sell them at. Yields shouldn't be too much of a problem, AMD is good at keeping them rather high once they get their process down. Penryn, while it will allow for clock speed boosts and adds SSE4 (K8L also has SSE4A) isn't really an architectural change, just a die shrink. I keep telling people to not expect C2D all over again...but they don't listen.

AMD isn't in the best spot right now, but they're not stupid.

 

[BaronMatrix]

Judging from how poorly AMD's 65nm transition has been, I wonder if it was economically feasible for them to do a quad core. Obviously, it's going to be a huge die, yields are going to be low, and the costs are going to be astronomical. I wonder if there is a market for it. Once Intel releases their 45nm quad core Penryn, they can undercut AMD in terms of cost. AMD won't be able to match. And if Intel's Penryn is faster, then AMD is going to be in alot of hurt.

The dual core 1MB chips were 231mm^2, Barcelona is said to be 283mm^2. That's not too bad. They managed to do a monolithic dual core. It's what they do.

Brisbane did excellently as a shrink from 130nm (ultimately). Had they not raised latency, no one would have said anything, since power is down.

 

[ajfink]

Judging from how poorly AMD's 65nm transition has been, I wonder if it was economically feasible for them to do a quad core. Obviously, it's going to be a huge die, yields are going to be low, and the costs are going to be astronomical. I wonder if there is a market for it. Once Intel releases their 45nm quad core Penryn, they can undercut AMD in terms of cost. AMD won't be able to match. And if Intel's Penryn is faster, then AMD is going to be in alot of hurt.

The dual core 1MB chips were 231mm^2, Barcelona is said to be 283mm^2. That's not too bad. They managed to do a monolithic dual core. It's what they do.

Brisbane did excellently as a shrink from 130nm (ultimately). Had they not raised latency, no one would have said anything, since power is down.

Note that it's really only impressive because of the fact that there is L2 and L3 cache.

 

[makaka]

well , from my view Intel can any time end this war .since Penryn was already released and it passe right now tests and conroe as all know proved to be better than A64.so if Intel will release more accessible price range conroe ,lets say between 80-120$ then amd can't keep up

 

[eregular]

Maybe AMD's Stars won't burn brightly enough until AM3...On board DDR3 controller...I mean Holy Crapstick Batman! 45nm native quad (and maybe by then kentsfield style octocores) Just wait till we have 256+ bit based cpu's....

 

[ajfink]

well , from my view Intel can any time end this war .since Penryn was already released and it passe right now tests and conroe as all know proved to be better than A64.so if Intel will release more accessible price range conroe ,lets say between 80-120$ then amd can't keep up

That's where the E4300 is going to be. Kind of a moot point to say AMD can't keep up...they've got very little going for them until K8L.

 

[Pippero]

Eh?
The E4300 costs exactly the same as the E6300... it should cost a few bucks left, but since it has a higher multiplier, its street price won't be as it should.

 

[ajfink]

Eh?
The E4300 costs exactly the same as the E6300... it should cost a few bucks left, but since it has a higher multiplier, its street price won't be as it should.

That's because it's new. Eventually prices will go to where they're supposed to be.

 

[Pippero]

It won't go to 80-120$ quickly, believe me.
Just look at the Netburst junk and how it is still priced...

 

[ajfink]

It won't go to 80-120$ quickly, believe me.
Just look at the Netburst junk and how it is still priced...

Different situation. Intel is keeping those prices high intentionally.

 

[BaronMatrix]

Judging from how poorly AMD's 65nm transition has been, I wonder if it was economically feasible for them to do a quad core. Obviously, it's going to be a huge die, yields are going to be low, and the costs are going to be astronomical. I wonder if there is a market for it. Once Intel releases their 45nm quad core Penryn, they can undercut AMD in terms of cost. AMD won't be able to match. And if Intel's Penryn is faster, then AMD is going to be in alot of hurt.

The dual core 1MB chips were 231mm^2, Barcelona is said to be 283mm^2. That's not too bad. They managed to do a monolithic dual core. It's what they do.

Brisbane did excellently as a shrink from 130nm (ultimately). Had they not raised latency, no one would have said anything, since power is down.

Note that it's really only impressive because of the fact that there is L2 and L3 cache.

And don't forget 4 cores. I would say this is a good sign that they are using the 65nm ZRam for this. They haven't really said but why license something if you're not going to use it?

 

[makaka]

It won't go to 80-120$ quickly, believe me.
Just look at the Netburst junk and how it is still priced...

Different situation. Intel is keeping those prices high intentionally.
True . this niche will be filled by new celeron 4xx and Pentium exxxx intel 2007 plan
then the remaining reserve of Netburst will cost in the 60-40$ range

 

[hannibal]

One thing has to be remembered:

If AMD fails -> the prosessor prices rokets sky high because there is not competition...

I hope that K8L is really good nad allso that Intels 45 is even better. Just hope that AMD 45 is not too late. I am not willing to play $1000 for any normal AMD or Intel prosessor and neither $2500 for top models!

I have had both Intel and AMD prosessors in my machines and I have to say that development has been for better after AMD managed to develop decent competition. Just hope that it will continue!

Huray for both teams! We are waiting for next round!

 

[makaka]

One thing has to be remembered:

If AMD fails -> the prosessor prices rokets sky high because there is not competition...

I hope that K8L is really good nad allso that Intels 45 is even better. Just hope that AMD 45 is not too late. I am not willing to play $1000 for any normal AMD or Intel prosessor and neither $2500 for top models!

I have had both Intel and AMD prosessors in my machines and I have to say that development has been for better after AMD managed to develop decent competition. Just hope that it will continue!

Huray for both teams! We are waiting for next round!

Totally agree if amd fail we will get them back to P4 days. i hope the same ...

 

[BaronMatrix]

Baron, you might have missed it.... the L1, L2, and L3 cache are normal SRAM, not ZRAM in AMD's 65 nm process -- the die shots have already been analyzed. Parrot had made this same conclusion erroneously, when I showed him the data it was clear.... all SRAM. Read through this thread to see the argument demonstrating no ZRAM, starting here:
http://forumz.tomshardware.com/modules.php?name=Forums&file=viewto...&p=

and here:

http://forumz.tomshardware.com/modules.php?name=Forums&file=viewto...&p=

ZRAM is still not ready for primetime, it is still way way too slow, even for L3 cache.

Ironically, ZRAM uses static electricity to function

Then you must have missed that ZRAM is ready for primetime at 65nm. It's possible that they will use it for Shanghai if it's not in yet.

My ultimate point was that the die is not that much bigger than 90nm X2. Here is last year's update to ZRAM.

PDF

I can only wonder WHEN it will be implemented. AMD did pay for a license.

 

[Pippero]

Yes, and that's the reason for introducing single core derivatives from C2D and Celeron or Pentium brand (or whatever it will be called).
The "Core 2" brand will not drop below 150$.

 

[mwswami]

Most likely, to keep the price hurt on AMD, Intel will introduce the Penryn chips at the higher end of the GHz scale first while keeping the prices close to the current Core 2 prices and lowering Core2 prices at the same time.

For example (numbers not to be taken literally) for DC desktop:

current:
Core 2 Duo: 2.93GHz EE ($1000), 2.66 ($500), 2.4 ($300), ...

with Penryn:
Penryn: 4.0 EE ($1000), 3.66 ($700), 3.33 ($500), 3.0 ($400)
Core 2 Duo : 2.66 ($300), $2.4 ($200), ...

 

[BaronMatrix]

Baron, you might have missed it.... the L1, L2, and L3 cache are normal SRAM, not ZRAM in AMD's 65 nm process -- the die shots have already been analyzed. Parrot had made this same conclusion erroneously, when I showed him the data it was clear.... all SRAM. Read through this thread to see the argument demonstrating no ZRAM, starting here:
http://forumz.tomshardware.com/modules.php?name=Forums&file=viewto...&p=

and here:

http://forumz.tomshardware.com/modules.php?name=Forums&file=viewto...&p=

ZRAM is still not ready for primetime, it is still way way too slow, even for L3 cache.

Ironically, ZRAM uses static electricity to function

Then you must have missed that ZRAM is ready for primetime at 65nm. It's possible that they will use it for Shanghai if it's not in yet.

My ultimate point was that the die is not that much bigger than 90nm X2. Here is last year's update to ZRAM.

PDF

I can only wonder WHEN it will be implemented. AMD did pay for a license.

Licensing a technology and making it work are two different things... they licensed Gen 1 ZRAM, so where is it? Perhaps this is why their L2 cache latency increased ... nahhh not really, because if they were using ZRAM they should have gotten a 60% die size reduction as opposed to 31%... hummmmmm, I will believe it when I see it.... just like high-K, Intel demonstrated functioning chips using high-K, therefore I believe Intel will make a high-K chip.... until AMD can show proof of concept, this is just headline grabbing, fool the rabid fanboy propoganda....

I doubt AMD is going to release a die shot of a quad core SRAM CPU only to remask and do it all over with ZRAM --- what waste of millions of bucks just to get a picture.

Why are you so easily fooled to the point you stretch a 'commerical chest thumping' into a hypthetical fact that only exists in your alternative universe.


Resources, resources, resources. If AMD had 12 Fabs they would have plenty of samples floating around. I would bet that there are some 45nm test vehicles at Fishkill.

Everyone knows this but have to have something negative to say.

 

[BaronMatrix]

Resources, resources, resources. If AMD had 12 Fabs they would have plenty of samples floating around. I would bet that there are some 45nm test vehicles at Fishkill.

Everyone knows this but have to have something negative to say.

To a point yes, however, how many times does someone need to say resources until it looks more like making excuses as opposed to being the reason.

No doubt there are 45 nm wafers running at Fishkill, if there weren't there is no hope for a mid 2008 45 nm release. And, resources is a poor excuse to begin with --- IBM has vastly more resources than Intel.... this should be no contest, IBM in front, yet Intel continues to surpass the competition...

So you mean Power5 isn't a better server proc? IBM doesn't do X86.

 

[Blacken]

Resources, resources, resources. If AMD had 12 Fabs they would have plenty of samples floating around. I would bet that there are some 45nm test vehicles at Fishkill.

Everyone knows this but have to have something negative to say.

To a point yes, however, how many times does someone need to say resources until it looks more like making excuses as opposed to being the reason.

No doubt there are 45 nm wafers running at Fishkill, if there weren't there is no hope for a mid 2008 45 nm release. And, resources is a poor excuse to begin with --- IBM has vastly more resources than Intel.... this should be no contest, IBM in front, yet Intel continues to surpass the competition...

So you mean Power5 isn't a better server proc? IBM doesn't do X86.

In the hell are you getting off? He never said anything about IBM having an X86 processer. Baron buddy, shouldn't you go eat mommies dinner she made for you downstair?

 

[BaronMatrix]

Resources, resources, resources. If AMD had 12 Fabs they would have plenty of samples floating around. I would bet that there are some 45nm test vehicles at Fishkill.

Everyone knows this but have to have something negative to say.

To a point yes, however, how many times does someone need to say resources until it looks more like making excuses as opposed to being the reason.

No doubt there are 45 nm wafers running at Fishkill, if there weren't there is no hope for a mid 2008 45 nm release. And, resources is a poor excuse to begin with --- IBM has vastly more resources than Intel.... this should be no contest, IBM in front, yet Intel continues to surpass the competition...

So you mean Power5 isn't a better server proc? IBM doesn't do X86.

Baron, this is a common means of losing an argument, resorting to putting words into one's mouth with a rhetorical question that is off point and changes the subject.

Process technology has always been, and for the forseeable future will remain number 1 at Intel, despite all the 'we have it too' headlines as well as 'looky what we did a lab, 500 GHz transistor'.... I am not comparing architecture so don't change the subject. My point is that this 'they have fewer resources there for cannot keep up' is getting old, and quite frankly not true --- IBM/Toshiba/AMD pooling resources to develop process technology provides vastly more resources than a single company (even the size of Intel) going it alone. Hence, my statement to a point yes... but pounding that point makes it look more like excuse making. Don't fault the delta of resources as a reason that AMD fails to deliver 65 nm in volume, quality, and comptetively.

You stated 65 nm likely implemented ZRAM and I pointed out that that is a pipedream at best, and that the current Barcelona die does not have ZRAM... this somehow evolved into you saying something as mundane as
'power 5 isn't as good'.... this deserves no reponse other than I have not studied power 5 performance in server apps, so unless I see data I would not know.... what is clear --- Intel could deliver to apple what IBM failed to do...

Intel is recognized as the industry leader in manufacturing and process technology and changing the subject with rhetorical, off the point questions will not change that.

Jack


That's how you prefer to read things. I guess it does keep flames flying but that's it. I said that if AMD had 12 fabs they would have more samples you said that's an old excuse and brought up IBM.

I then said that Power 5 is a great server chip. Intel is pushing this process advantage "line" to attempt to disparage AMD. Same old thing.

The fact that one little fab kept Intel at bay with larger processes means AMD is actually the better company for innovation. They HAVE TO BE. Barcelona will probably do exactly what Opteron did.

Kick Intel in the n u t s.
:lol:

 

[dasickninja]

One little FAB with a lot of help from big names you mean. *Clarification*

 

[BaronMatrix]

One little FAB with a lot of help from big names you mean. *Clarification*

Which big names are those? Links please?