Sign in with
Sign up | Sign in
Your question

AMD makes 45nm test chips

Tags:
Last response: in CPUs
Share
February 7, 2007 1:04:49 PM

WHAT THE F...?
February 7, 2007 1:26:22 PM

Quote:
“65nm is kind of old hat. The focus is on 45nm. […]We’ll be producing early [45nm] products probably in Q2 of 2008, with full production in the second half,” Mr. Grose is reported to have said in an interview with Reuters.

They haven't even master making chips at 65nm and they are already jumping to 45nm? Hell, they might as well have jumped to 32nm and called it a day. Crappy yields, here we come!
Related resources
February 7, 2007 1:28:06 PM

xactly wot I thought! just 1 year of working with 65nm and then 45nm in '08, I don't believe this crap!
February 7, 2007 1:28:34 PM

I don't think AMD will be able to keep pace with Intel.
February 7, 2007 1:29:57 PM

Quote:
I don't think AMD will be able to keep pace with Intel.


But they sure will die trying...
February 7, 2007 1:31:27 PM

1 year? Its been less than 6 months that these chips have been marketable, and there are no perceivable differences over 90nm. 45nm will not only waste money but put them behind Intel. What the hell are they thinking? This better just be an experiment.
February 7, 2007 1:34:01 PM

Quote:
I don't think AMD will be able to keep pace with Intel.


But they sure will die trying...
Yep
February 7, 2007 1:42:17 PM

Quote:
But they sure will die trying...


Always better to die trying than giving up like a sore loser. Oh well... should make for an interesting year or so.... better get my popcorn ready!
February 7, 2007 1:44:36 PM

I always wonder, why AMD bought ATI in the first place, rather they could have built a new fab, they could have put some more money in R&D, can anyone tell me what benefit has ATI provided to AMD up till now? and now AMD is talking about 45nm process in '08.
February 7, 2007 1:46:02 PM

Quote:
But they sure will die trying...


Always better to die trying than giving up like a sore loser. Oh well... should make for an interesting year or so.... better get my popcorn ready!

My point was that AMD should NOT return to playing Intel's game, because they will not win that way.
February 7, 2007 1:50:46 PM

Quote:
I always wonder, why AMD bought ATI in the first place, rather they could have built a new fab, they could have put some more money in R&D, can anyone tell me what benefit has ATI provided to AMD up till now? and now AMD is talking about 45nm process in '08.

Specialists to start off immediately; I guess AMD had no good graphics professionals and those kind of people are pretty rare and already employed; not the kind you can find by journal announcements so the best deal was to buy all ATi.
February 7, 2007 1:53:30 PM

Ahh.... but even then, they should at least try to keep up with Intel's pace, make sure Intel gets the necessary competition to not create something like Netburst again. That way, Intel can also give AMD the competition it needs so not to get complacent, as C2D showed us when it came out.
February 7, 2007 1:56:47 PM

so have they come up with a product? according to my observation wot AMD needed was that to keep the lead, ATi wasn't going anywhere, ATi wasn't about to bought out by some other company.

People hit axe on the there foot, AMD hit there foot onto the axe.
February 7, 2007 1:57:19 PM

They want to catch Intel in the node changes - this is what it will take. Fast changes.
February 7, 2007 1:58:19 PM

sheesh whats all the hub-bub about. It says in the article that production will not begin for a year.

You guys seriously have issues. All I've heard in the past 6 months is whining and crying about AMD not innovating fast enough. Now AMD will drop from 90nm to 45nm in less than a year while increasing to 300mm wafers AND making major renditions to K8, and you guys are still crying??? WTF do they have to do to please you guys, Re-Invent the wheel???

This is nothing but great news! The cost savings from moving to both 45nm and 300mm wafers is tremendous. This means that from December 06 to May 08 that AMD has the capacity to MORE then double its manufacturing output without building a single new FAB. Obviously after the process matures.

If you take a step back and look at the larger picture youll realize fairly quickly that this is major solution to one of AMD's biggest problems of the past year, capacity strain.
February 7, 2007 2:12:18 PM

Quote:
They want to catch Intel in the node changes - this is what it will take. Fast changes.


Indeed. As long as AMD makes good use of their connections with IBM, I am confident they can pull off these fast changes.
February 7, 2007 2:19:05 PM

Quote:
so have they come up with a product? according to my observation wot AMD needed was that to keep the lead, ATi wasn't going anywhere, ATi wasn't about to bought out by some other company.

People hit axe on the there foot, AMD hit there foot onto the axe.


Actually, AMD's acquisition of ATI was mainly for chipsets and also extra revenues generated by ATI's discrete graphics section. By having ATI, AMD can effectively create their own complete platform for their CPUs, without worrying about whether third party companies, like Nvidia and VIA, can supply the necessary platforms for their processors. Intel makes their own platforms, and so if AMD can do so too, then they can take on Intel on a much more level playing field. Its all long-term strategies when dealing with acquisitions, because almost all of the time, they don't yield short-term benefits.
February 7, 2007 2:20:06 PM

Yes, and the whole market REALLY NEEDS AMD to be competitive, because they saw the drastic price drops and performance boosts the last 3 years brought. Everyone is going to support them, from Dell to IBM; this is the way to tremendously accelerate the PC future.
February 7, 2007 2:22:31 PM

Quote:
so have they come up with a product? according to my observation wot AMD needed was that to keep the lead, ATi wasn't going anywhere, ATi wasn't about to bought out by some other company.

People hit axe on the there foot, AMD hit there foot onto the axe.

Well, it looks so from the outside, but there are certainly reasons beyond our knowledge.
February 7, 2007 2:43:28 PM




I love being right. AMD is in an excellent position to pull this off. The reason why they couldn't do something like this before was because they didn't have enough Fabs.

Now they can offload to Chartered for 90nm and 65nm while installing 300mm wafers for one part of Fab30, then installing 45nm equipment in another part.

They seem to be executing rather well right now. The ATi acquisition seems to have gone well as they are releasing new chipsets and platforms.

It's not so far-fetched as immersion has been around since 2003 and they have said they will not use metal gates for Shanghai which is to be the first 45nm chip.

I bet if they had said they would not go to 45nm next year, people would have a different complaint.

We'll see if they show a wafer by July. They should have Fab38(45nm 300mm @ Fab30) online by then. Or they would have to in order to get chips out in Q208.
February 7, 2007 3:29:31 PM

Quote:
Yes, and the whole market REALLY NEEDS AMD to be competitive, because they saw the drastic price drops and performance boosts the last 3 years brought. Everyone is going to support them, from Dell to IBM; this is the way to tremendously accelerate the PC future.


Because cheap, powerful processors are the best... :D  Let's just hope Barcelona and 45nm really help put AMD back into the business of being the major thorn in Intel's fat ass...
February 7, 2007 3:46:10 PM

Quote:
1 year? Its been less than 6 months that these chips have been marketable, and there are no perceivable differences over 90nm. 45nm will not only waste money but put them behind Intel. What the hell are they thinking? This better just be an experiment.


I may be wrong, but I think the main problem with the 45nm so far is that they are still using old main architecture, but on a smaller scale. Probably saved them some manufactoring costs, but no other gain. Now if they make the internal structure of the cpu different, then the 45nm may come on strong.

Hope that makes sense. I'm still barely awake. Got to get some coffee. Either that, or go back to sleep.
February 7, 2007 3:47:48 PM

short article....but nonetheless good news!! I mean their only a couple weeks behind intel showing off penryn...i imagine both companies are going to have problems with 45nm but like all things...it'll get hammered out and all will be right in narnia.

45nm native quad cores though...a year and half away :(  oh well...not like i'd shell out anymore than 300 for a cpu...now when the quads start to cost 300 cause theirs 6 or 8 core beasts setting some awesome benchmarks...that'll be NiFtY...
February 7, 2007 4:11:35 PM

Every one here misses the biggest point. It is Big BLUE. AMD is "working jointly" with IBM. Read IBM is doing the work and AMD helps pay the bills. and shares the technology. If any one here thinks Intel can compete with IBM's Engineering Dept, I will give you a plastic cup and we will have a drug test.
February 7, 2007 4:17:41 PM



I don't know quite what is so surprising. Intel's 45nm test chips (sram) were announced in Jan of 06, working processors Jan 07, with production late 2007/early 2008. So AMD's timeframe is not that far off from Intel's, they are just 18 months behind process to process. Test chips Jan 07, working processors Jan 08, production late 2008 (he seems a bit optimistic though). Just have to wait and see how fast intel can go below 22nm with dry litho and if immersion litho can do what IBM/AMD thinks it can.

Centurion
February 7, 2007 4:20:32 PM

Quote:
If any one here thinks Intel can compete with IBM's Engineering Dept, I will give you a plastic cup and we will have a drug test.


They don't have to. Because everyone knows a shaved spider monkey could sell hair growth formulas better than IBM can market its PC products...
February 7, 2007 4:23:45 PM

Quote:
If any one here thinks Intel can compete with IBM's Engineering Dept, I will give you a plastic cup and we will have a drug test.


They don't have to. Because everyone knows a shaved spider monkey could sell hair growth formulas better than IBM can market its PC products...

Marketing has nothing to do with it. Just being IBM / AMD is marketing enough for a lot of purposes.

IBM registers thousands of patents a year just on random shite. They have some of the smartest and highest-paid engineers in the world working cranking out ideas and designs in such a broad field of technological applications it's staggering. IBM is highly revered for a reason, and for good reason.
February 7, 2007 7:44:30 PM



I don't know quite what is so surprising. Intel's 45nm test chips (sram) were announced in Jan of 06, working processors Jan 07, with production late 2007/early 2008. So AMD's timeframe is not that far off from Intel's, they are just 18 months behind process to process. Test chips Jan 07, working processors Jan 08, production late 2008 (he seems a bit optimistic though). Just have to wait and see how fast intel can go below 22nm with dry litho and if immersion litho can do what IBM/AMD thinks it can.

Centurion
It's got nothing to do with when they first tested 45nm. If amd launches the 45nm chips in Q2 '08 then they're 2 quarters behind Intel ~6months. You seem to dislike AMD, I've read your other posts.
February 7, 2007 10:55:03 PM

Quote:
sheesh whats all the hub-bub about. It says in the article that production will not begin for a year.

You guys seriously have issues. All I've heard in the past 6 months is whining and crying about AMD not innovating fast enough. Now AMD will drop from 90nm to 45nm in less than a year while increasing to 300mm wafers AND making major renditions to K8, and you guys are still crying??? WTF do they have to do to please you guys, Re-Invent the wheel???

This is nothing but great news! The cost savings from moving to both 45nm and 300mm wafers is tremendous. This means that from December 06 to May 08 that AMD has the capacity to MORE then double its manufacturing output without building a single new FAB. Obviously after the process matures.

If you take a step back and look at the larger picture youll realize fairly quickly that this is major solution to one of AMD's biggest problems of the past year, capacity strain.


I have said it countless times.... on this forum...PRO-AMD news is attacked all the time

Pro-INTEL news...not so much

I think people percieve INTEL can do whatever they claim and AMD cannot due to the size of each company
February 7, 2007 11:03:27 PM



I don't know quite what is so surprising. Intel's 45nm test chips (sram) were announced in Jan of 06, working processors Jan 07, with production late 2007/early 2008. So AMD's timeframe is not that far off from Intel's, they are just 18 months behind process to process. Test chips Jan 07, working processors Jan 08, production late 2008 (he seems a bit optimistic though). Just have to wait and see how fast intel can go below 22nm with dry litho and if immersion litho can do what IBM/AMD thinks it can.

Centurion
It's got nothing to do with when they first tested 45nm. If amd launches the 45nm chips in Q2 '08 then they're 2 quarters behind Intel ~6months. You seem to dislike AMD, I've read your other posts.


Yeah anything positve about AMD upsets him
only INTELIOTS get upset over positive AMD news....
February 7, 2007 11:10:36 PM

Most of them don't realize that even Intel's latest innovations were pulled out under AMDs pressure. Hadn't it been for them, now we'd be buying 2.0GHz P4s for $500. That is why their job deserves respect and personally, I don't even try to hide my admiration for them.
February 7, 2007 11:23:01 PM

heh, exactly.

Personally i hope AMD manages to pull off everything they claim they can, from monolithic quad cores, to 45nm... And force Intel to make another C2D revolution just to stay alive :p 
February 7, 2007 11:54:49 PM

All I have to say about this is simple.

Intel has such a lead on AMD with 65nm that it would actually make sense for AMD to just skip 65nm altogether and jump right into 45nm. All that time, money and effort spent trying to get 65nm would actually help them catch Intel at 45nm if they were to skip this die shrink.

Let's remember AMD now has IBM to help transition to 45nm now, so this claim of 45nm in Q2 2008 is not actually surprising.
February 8, 2007 12:00:48 AM

OK, Im opening myself up to flamebait here, but heres what the rumor mill has out and SOME if not all is true. With Barcelona we see amd catching intel in IPC (if not surpassing ) We see power usage being done 'smarter' , caching as well and who knows how thatll effect performance. Now we read that amd is also catching up in processing (45nm) If any or all this rings true and all things mature, amd will keep pace with intel as well as be robust with production costs lowering. I like others here would like to see this happen for all our sakes.
February 8, 2007 12:29:05 AM

Quote:
This better just be an experiment.


i would imagine it is.
they have to start sometime.

amd will probably start messin with 32nm with in the next 6 months.
not saying they can muster it. but they will probably start experimenting.
February 8, 2007 2:02:22 AM

Quote:
All I have to say about this is simple.

Intel has such a lead on AMD with 65nm that it would actually make sense for AMD to just skip 65nm altogether and jump right into 45nm. All that time, money and effort spent trying to get 65nm would actually help them catch Intel at 45nm if they were to skip this die shrink.

Let's remember AMD now has IBM to help transition to 45nm now, so this claim of 45nm in Q2 2008 is not actually surprising.


Skipping 65nm to go straight to 45nm isn't a viable thing to do.

First, IBM isn't on the leading edge of 45nm processes, like some would like to believe. In order to do so, they would actually have to spend a ton of money developing the process with equipment manufactuers, and that does get expensive. Really expensive, with different tool configurations, materials used, and gases/liquids used. It's not an easy transition, and it's far from easy to actually develop a new process at a smaller scale.

Second, AMD cannot afford to retool a FAB to do the same developmental stuff for 45nm, mostly cause of costs. Again, getting the right toolsets to do what they want isn't cheap. If a litho tool isn't right, a replacement tool could cost up to a few million for ONE. And having ONE tool to do all the lithography isn't cost effective.

A lot of the 45nm toolsets could be the same as 65nm, but with some major changes in material used and other process nightmares. If AMD doesn't have a viable 65nm process in place, and decides to jump to 45nm, even with IBM's help, it could be a long, expensive ride for both. AMD's best bet is to allow IBM and Intel to get 45nm at a stable state, and do a reconversion of toolsets that can be converted to 45nm easily, and without having to retool a fab.

That's my opinion, and that's all that it is.
February 8, 2007 2:29:02 AM

Quote:
All I have to say about this is simple.

Intel has such a lead on AMD with 65nm that it would actually make sense for AMD to just skip 65nm altogether and jump right into 45nm. All that time, money and effort spent trying to get 65nm would actually help them catch Intel at 45nm if they were to skip this die shrink.

Let's remember AMD now has IBM to help transition to 45nm now, so this claim of 45nm in Q2 2008 is not actually surprising.

AMD has had IBM for a while.
February 8, 2007 3:52:42 AM

I honestly think they're rushing to this because of the fact that Intel *will* have 45nm out the door a lot sooner than they're leading on or people believe...I'd hate to see them fall on their asses but if they rush into it that's exactly what's going to happen...

...and from what I've been reading immersion is a more expensive process than dry, so even if they do get 45nm out the door, I will guarantee the cost of the processors is going to far outweigh their performance value...
February 8, 2007 4:01:45 AM



I don't know quite what is so surprising. Intel's 45nm test chips (sram) were announced in Jan of 06, working processors Jan 07, with production late 2007/early 2008. So AMD's timeframe is not that far off from Intel's, they are just 18 months behind process to process. Test chips Jan 07, working processors Jan 08, production late 2008 (he seems a bit optimistic though). Just have to wait and see how fast intel can go below 22nm with dry litho and if immersion litho can do what IBM/AMD thinks it can.

Centurion


Centurion is quite correct.

This announcement puts AMD exactly at where intel was at in q1 of 2006. This is actually pretty bad news for AMD.

AMD also says first production of 45nm in q2 2008 -- this is actually a little bit later than intel, where first production of penryn in q1 of 2007. That's 1 full quarter slower from their test chips, but perhaps not unexpected since intel has a full development fab at their disposal where AMD needs to balance 65nm/revenue with R&D.

Ramp and debug should take intel to q4 2007 -- amd wants to do that in ~3 quarters, too. (shipment late 2008.) And that appears to be without hiK/MG.

Despite AMD's spin, it looks as if they are NOT catching up to intel anytime soon.
February 8, 2007 4:13:41 AM

Kool.Looks like AMD is getting their s*** together finally.Sorry but it's a long time coming.

Dahak

AMD X2-4400+@2.6 S-939
EVGA NF4 SLI MB
2X EVGA 7800GT IN SLI
2X 512MB CRUCIAL BALLISTIX DDR500
WD300GIG HD
EXTREME 19IN.MONITOR 1280X1024
THERMALTAKE TOUGHPOWER 850WATT PSU
COOLERMASTER MINI R120
February 8, 2007 4:26:11 AM



I don't know quite what is so surprising. Intel's 45nm test chips (sram) were announced in Jan of 06, working processors Jan 07, with production late 2007/early 2008. So AMD's timeframe is not that far off from Intel's, they are just 18 months behind process to process. Test chips Jan 07, working processors Jan 08, production late 2008 (he seems a bit optimistic though). Just have to wait and see how fast intel can go below 22nm with dry litho and if immersion litho can do what IBM/AMD thinks it can.

Centurion


Centurion is quite correct.

This announcement puts AMD exactly at where intel was at in q1 of 2006. This is actually pretty bad news for AMD.

AMD also says first production of 45nm in q2 2008 -- this is actually a little bit later than intel, where first production of penryn in q1 of 2007. That's 1 full quarter slower from their test chips, but perhaps not unexpected since intel has a full development fab at their disposal where AMD needs to balance 65nm/revenue with R&D.

Ramp and debug should take intel to q4 2007 -- amd wants to do that in ~3 quarters, too. (shipment late 2008.) And that appears to be without hiK/MG.

Despite AMD's spin, it looks as if they are NOT catching up to intel anytime soon.
1)WE don't know about high-k/metal gates. -speculation
2)These are test chips in q1 '07, just like you mentioned penyrn which was actually q4 '06
3) Amd will ship 45 nm in Q2 '08, intel hasn't said for sure, but somewhere between q1 '08 and q4 '07
4) it isn't first production in Q2 '08 it's actual production, as in chips that are going to be sold. This is 18 months, like amd said, and intel is right around 2 years, lowering the gap to roughly months.
Much of what you just said was wrong/speculated.
February 8, 2007 4:58:22 AM

I wonder how much support (and why?) IBM are putting into this in the background...
February 8, 2007 5:04:47 AM

I know it seems rather stupid to move onto the next level when you haven't perfected the level your currently on, but quite often when working with designs in general, you'll wipe the slate clean and start with a new beginning. AMD might have learnt an aweful lot about 65nm and may feel that they would be wasting further money by trying to fix what they've started. So buy jumping to the next level, they can close the "catch up to Intel" gap while having the extra cash from not spending it on finishing off 65nm, they will be able to tackle 45nm quite hard.

I understand this because it's what I've done with several projects of my own (Websites, PCB designs and Programs). Was half way through building a prototype RC Relay setup, found a lot of places I could improve it, instead of finishing it, I shifted onto the next version and saved the time and money of finishing off the development on something that was lesser then what I produced with the second version 8)

This is where AMD will catch up to Intel in one shot, or cut their wrists like an emo listening to Greenday.... :p 
February 8, 2007 5:10:39 AM

Quote:
I know it seems rather stupid to move onto the next level when you haven't perfected the level your currently on, but quite often when working with designs in general, you'll wipe the slate clean and start with a new beginning. AMD might have learnt an aweful lot about 65nm and may feel that they would be wasting further money by trying to fix what they've started. So buy jumping to the next level, they can close the "catch up to Intel" gap while having the extra cash from not spending it on finishing off 65nm, they will be able to tackle 45nm quite hard.

I understand this because it's what I've done with several projects of my own (Websites, PCB designs and Programs). Was half way through building a prototype RC Relay setup, found a lot of places I could improve it, instead of finishing it, I shifted onto the next version and saved the time and money of finishing off the development on something that was lesser then what I produced with the second version 8)

This is where AMD will catch up to Intel in one shot, or cut their wrists like an emo listening to Greenday.... :p 

I had the feeling AMD had a break through with their 45nm process when they announced that it would be 18 months from 65nm to 45nm. Maybe SOI was asking for immersion, I would almost guess, and that took away a lot of the problem in developing manufacturing. New processes require time, and they just must have had a breakthrough that allowed them to say that hey didn't need as much time as normal. I don't think it really happened like AMD says, where they are pushing for 18 months, I think something just happened that will allow them to launch in 18 months. Bah, I'm confused, do you get what I mean?
February 8, 2007 5:18:02 AM

Quote:
Immersion has nothing to do with having or not having SOI.... it is a lithography technique for access smaller dimensions withoug having to push down the wavelength for the exposure.

They are going to immersion because they cannot meet the dimensional requirements with dry.

I know, I was suggesting that it had something to do with a breakthrough in 45nm to make it 18 months. Because it almost seems like amd had trouble with 65nm because it didn't use immersion (I suggest this just from a lot of reading on amd, 65nm and immersion.).
February 8, 2007 5:27:58 AM

You understand though, Jack, that I didn't confuse Immersion and SOI? And that I think AMD got sort of lucky with 45nm? What are your thoughts on that.
February 8, 2007 5:31:21 AM

Let me express my ignorance of laser freq's... But could they be employing a tighter / higher frequency to attain the dry masking?
February 8, 2007 5:32:28 AM

Quote:

1)WE don't know about high-k/metal gates. -speculation


From the article:

The new chief of manufacturing at Advanced Micro Devices said the company has already made first test chips using 45nm process technology, however, he said the new fabrication process still does not employ high-k dielectrics and similar innovations that IBM and AMD developed collaboratively.

Quote:

2)These are test chips in q1 '07, just like you mentioned penyrn which was actually q4 '06


Respectfully, Rand, there is a world of difference between a test chip and a product. Penryn is a lead technology vehicle. A test chip means a specific thing: a set of circuitly (like SRAM and PLLs) that comprise the building blocks of a processor and qualify the overall process, but do not, in themselves, comprise a working CPU. (Note: I am intimately involved in design + process development so I know of which I speak.) Intel did indeed demonstrate a test chip in q1 of 06. If AMD is only doing so now, this is ~12 months behind intel.

Quote:

3) Amd will ship 45 nm in Q2 '08, intel hasn't said for sure, but somewhere between q1 '08 and q4 '07


Again, the quote is:

We’ll be producing early [45nm] products probably in Q2 of 2008


This is a far cry from volume production, and is, in fact what intel is doing just now. Intel is producing a functional Penryn A0. They are not selling it. Note the distinction. Historically, it's ~9 months from first silicon to HVM.

In fact ...

full production in the second half

seems to be on exactly this time table.

Quote:

4) it isn't first production in Q2 '08 it's actual production, as in chips that are going to be sold. This is 18 months, like amd said, and intel is right around 2 years, lowering the gap to roughly months.


Please provide a citation where they say HVM. I cite above, a source from AMD indicating this is in the latter half of '08, which I extrapolate to mean late q4, based on historical (not just amd) trends.

Quote:
Much of what you just said was wrong/speculated.


There is no doubt that I am speculating.. I do not work for AMD. Unless you do, then you are speculating as well. :D  However, I am attempting to parse precisely what the interview said. And his phrasing was critical.

However, I should retract a statement -- I did say amd was 18 months behind. This is erroneous, it appears as if they are 12 months behind on 45nm. So, they *are* catching up.[/quote][/b]
February 8, 2007 5:38:52 AM

You're cutting me with words that are inconclusive. For example,
Quote:

1)WE don't know about high-k/metal gates. -speculation


From the article:

The new chief of manufacturing at Advanced Micro Devices said the company has already made first test chips using 45nm process technology, however, he said the new fabrication process still does not employ high-k dielectrics and similar innovations that IBM and AMD developed collaboratively.
Right but he doesn't say that they won't be using high-k/metal gates, just that they aren't right now.

There is no argument here. I suppose if I test my chip at 45nm 55 years before you, but you release your chip 6 months after me, it makes a whole hell of a lot of a difference when I tested mine?
!