Saturday, July 29, 2006

65nm AMD parts in the wild?

Questions, questions.

AMD's 90nm X2 3800+ frags Intel's 65nm Presler Pentium XE965. I expect AMD's 65nm parts to see 40% performance improvement over 90nm.

55 Comments:

Anonymous Anonymous said...

Its about time. How can AMD pretend to be a leader and be so late?

Yawn.. whats the news, they are a year late and its ain't even their own technology but IBM's.

7:57 PM, July 29, 2006  
Anonymous Anonymous said...

What does "ES" mean?

That means retail and I can buy today correct? Wow, I had no idea AMD would pull the 65nm launch in by >4 months.

8:40 PM, July 29, 2006  
Blogger Mad Mod Mike said...

"Its about time. How can AMD pretend to be a leader and be so late?

Yawn.. whats the news, they are a year late and its ain't even their own technology but IBM's."

Do Intel fanboys know anything? Do you ever shut up? Jesus...I feel like Gary Goleman from Tourgasm when he said to a heckler "Will somebody please put a c*ck in his mouth?"

8:46 PM, July 29, 2006  
Anonymous Anonymous said...

Its not unexpected, the reason DELL is going for AMD in its lineup.
AMD work beaver-like behind the scene
producing 65nm product shrinks of its line will so it can release in volume.
Intel has paper launches, nothing has changed.

9:31 PM, July 29, 2006  
Anonymous Anonymous said...

Madmodmike is right - clearly that first Intel fan post is wrong and doesn't know anything.

If AMD starts shipping in Dec06, they are ONLY 11 MONTHS behind the Intel 65nm launch not "a year". More lies and exaggerations from Intel fans!

(BTW - before I get flamed, this is meant to be sarcastic)

9:35 PM, July 29, 2006  
Anonymous Anonymous said...

More significant is the speculation that this part is a dual core with 2M L3 cache and K8L cpu spec's .?

10:09 PM, July 29, 2006  
Anonymous enumae said...

Mad Mod Mike said...

"Do Intel fanboys know anything? Do you ever shut up?"

Instead of bashing why dont you explain your point?

10:15 PM, July 29, 2006  
Anonymous Anonymous said...

These guys just have a good time. It could well be a 65nm Engineering Sample. But by no means could it possibly have an L3 cache, because it definately isn't a K8L. Looks an early sample anyway because of the low frequency.

10:26 PM, July 29, 2006  
Anonymous Anonymous said...

ES = Engineering Samples.

These represent probably first stepping with many bugs sent to OEM manufactures to start validation, platform stability and BIOS work.

Figure another 6 months best case before you see production then another 6 months before you see volume.

ES samples of Conroe were out there 9 months ago.

AMD is very very late, they are in trouble.

ES.. how sad. This is very bad news of how far they are behind.

10:54 PM, July 29, 2006  
Anonymous Anonymous said...

Well, that might really be a 65nm AMD but I wouldn't call it available. If I would then I could call those 4-corce Kentsfield available too ...
http://www.xtremesystems.org/forums/showthread.php?t=109054

12:50 AM, July 30, 2006  
Anonymous Anonymous said...

"Its about time. How can AMD pretend to be a leader and be so late?

Yawn.. whats the news, they are a year late and its ain't even their own technology but IBM's."

Wow..., don't be too optimist with what you say, unless you have the evidence...

12:52 AM, July 30, 2006  
Anonymous Anonymous said...

Well, that might really be a 65nm AMD but I wouldn't call it available. If I would then I could call those 4-corce Kentsfield available too ...
http://www.xtremesystems.org/forums/showthread.php?t=109054

12:56 AM, July 30, 2006  
Anonymous Anonymous said...

I would hope that AMD are rushing these out to combat Conroe. I don't think that will outperform Conroe but obviously you would hope they could reduce the price further so that people can then get a better price on either AMD or Conroe.

I also would like to see Intel and AMD making money even in this price war so they can still both invest in the future.

3:24 AM, July 30, 2006  
Anonymous Anonymous said...

"if AMD starts shipping in Dec06, they are ONLY 11 MONTHS behind the Intel 65nm launch not "a year". More lies and exaggerations from Intel fans!
(BTW - before I get flamed, this is meant to be sarcastic)"

Yeah AMD is one year behind Intel but it continues to gains market share and raise profits and Intel continues to loose Market share and profits, how does that make the stupid Intel fanboys feel like?
That's how retarded intel fanboys are, gloating when Intel as been constantly fragged by a company that is one year behind.

3:51 AM, July 30, 2006  
Blogger Mad Mod Mike said...

"Instead of bashing why dont you explain your point?"

Because Intel fanmonkeys don't listen to points, they just say "Conroe pwns!" - Gets old after awhile, eh?

"Figure another 6 months best case before you see production then another 6 months before you see volume."

Obviously you're a moron. I think it was AMD themselves who said Dec06 was 65nm for sure, hmm..

"ES samples of Conroe were out there 9 months ago."

So? Intel NEEDED Conroe out that early to hype it up.

"AMD is very very late, they are in trouble."

I bet you said the same thing when AMD was @ 130nm and the P4 reached 90nm...we all know how "in trouble" AMD was because of that.

"ES.. how sad. This is very bad news of how far they are behind."

Again...you are a moronic Intel fanmonkey who knows nothing. Because AMD is behind on changing process, means they live longer with mature yields and reduce costs to end-users; something Intel has never done until Core 2 Duo because they know K8 whipes the floor with everything they offer.

Jesus Fanmonkeys...Put a c*ck in your mouth and shut up!

4:22 AM, July 30, 2006  
Anonymous Edward said...

"These represent probably first stepping with many bugs sent to OEM manufactures to start validation, platform stability and BIOS work."

Engineering samples usually are tested in-house. Any decent company would do that (unless you assume AMD is not a decent company). Also, the sample-to-market time is usually less than 6 months - if longer, the develop group is in trouble.

I have no idea how Intel's Core 2 Duo were develop. But if as you said it took 9 months for them to go from ES to production, I'd say Intel's designers had dell a lot of trouble to get their baby out.

Beside, I don't believe your 9-month number. Lets say the latest Intel started volume production in April (2 months after the IDF show). That means as early as last July they had engineering sample ready. Duh... last July... do you have any proof that this is possible?

5:39 AM, July 30, 2006  
Anonymous Edward said...

"Its about time. How can AMD pretend to be a leader and be so late?
Yawn.. whats the news, they are a year late and its ain't even their own technology but IBM's.
"

AMD is a leader of processor architecture, not processing technology. There is little dispute that Intel is the leader of semiconductor production technology. Only few others can rival that (IBM, or maybe TI).

With less than 1/4 world market and 50% margin for a year(?), AMD is able to build a new fab and acquire ATi. Intel OTOH has had 3 times of this for like 15 or 20 years. It's no big deal that Intel has the best production technology; it's only amazing that with its superior processing technology, for the past 3 year, Intel's CPU performance still lagged AMD's.

AMD might be a year late with 65nm, but Intel was 3 years late with Conroe. Ever think of that?

5:52 AM, July 30, 2006  
Anonymous muziqaz said...

to that last anonymous guy:
amd is not in trouble.
intel is.look intel on 65nm can manage 65watt tdp for conroe, right?
amd can manage the same with 90nm.
now core duo is also on 65nm? and manages 25-30watts.amd turion x2 is on 90nm and manages pretty much the same thing.
that nanometer race is the same as megahertz race.at the end of the day intel will see the wall sooner than amd, becausde amd does not need less nm to stay competetive.
I'm just waiting to see what will buldozer with 65nm will do to merom [;

besides when I saw news about merom launch, none of the news sites stated TDP, only fsb, clockspeed, L2 size, and price.Why? Isn't the TDP most important thing in mobile cpu?

OT.Today I got some time to read the transcript of intels quarter finantial meeting(or how they call it). Intel showed interesting manners when answering the questions [[[;

6:33 AM, July 30, 2006  
Anonymous fire hector now said...

It doesn't matter how early or late AMD executes its strategy. It doesn't matter 130nm, 90nm, 65nm, etc.

For AMD's new strategy is all about making themselves into Intel's evil clone. The new AMD is anti-customer.

AMD put out a dud chip (AM2) and then made their chip even worse by removing half the cache (or 75% of it for their latest AM2).

Removing the cache doesn't help the customer. But it does make space for 2P on the enthusiast desktop -- AMD's 4$x4$. No sense in having a single AMD processor being able to compete with 2, is there? Not if you want to sell twice as many chips. Or now, twice as many graphics cards.

Now we have AMD+ATI. This means the end of affordable graphics as there will be no more competition between Nvidia and ATI. Instead we will see everything from ATI locked into AMD's platform.

All this crazy stuff AMD is doing is to make themselves into Intel. Instead of spending a little money and making their chips better.

It is a lesson Intel just learned the hard way, but AMD has apparently yet to learn. If you don't spend money making your chips better, you end up with massive problems.

AM2 is a dud. Opteron is overpriced. 4x4 is useless for most people (and looks to be just as expensive as a real 2P Opteron system). And the rest (Torrenza, etc) is just too vapor to even comment on.

Meanwhile you can buy the world's fastest x86 chip from the company that invented x86.

For most people, that is a no-brainer.

Meanwhile, AMD has nothing to offer the market except vapor technology and fire sales on old chips. Yawn.

The sooner AMD fires Hector and the rest of his loser management team, the better off AMD will be.

Hector is for AMD what Craig Barrett was to Intel. A crappy CEO that made a lot of acquisitions instead of focusing on core technology and business fundamentals.

7:15 AM, July 30, 2006  
Blogger Mad Mod Mike said...

"For AMD's new strategy is all about making themselves into Intel's evil clone. The new AMD is anti-customer."

You're a moron.

9:45 AM, July 30, 2006  
Anonymous Anonymous said...

What is different this time stupid AMD fanboys.

At 90nm INTEL was way ahead by a year, they had the best silicon but were struggling with a bad design. You guys know that I know that. They used their one year lead on silicon to try and stay close to the far superior AMD design. That again was the case at 65nm. THey were shipping CedarMill way earlier, debugging the process and driving to very high yields.

Then they introduced Core2. The benchmarks stand for themselves. The new architecture gave them 40%+ improvement compared to CedarMill and a good 20-30% over AMD's best.

In general a simple shrink even if you add more cache gets you 20% performance benchmark. Thus the to compre AMD's situation at 90nm is stupid. The playing field has changed.

Try something new. Until AMD introduces their new architecture they are at the very best case close in performance but behind on the majority of the line. So competing thru most of 2007 will be very difficult.

You fanboys will be making lots of noise when the new architecture and the 4x4 ships.. but if you think there was a shortage of Conroe at launch the shortage of the AMD parts will make the xbox 360 availability look like a joke.


SO stop comparing 90nm to 65nm situation the landscape has completed changed you idiots.

AMD was ahead at 90nm but at 65nm with Core2 the game is over. The fat eleplant has learned to dance even with stupid Craig and Paul as the ring leader. The engineers from IDC have spanked AMD's asses.

Its sucks but that is the truth boys.

Tell me why it won't be so thru the rest of 2007.. I'm eager to hear!

The Doctor

9:59 AM, July 30, 2006  
Anonymous Anonymous said...

"Yeah AMD is one year behind Intel but it continues to gains market share and raise profits and Intel continues to loose Market share and profits"

AMD gained 1% market share last quarter with parts that any objective person are superior to Intel's P4. Inthe one space where Intel is considered as good as/better than AMD chips (mobile), they gained 0.4%. With a competitive pdroduct now being ramped in 1P/2P server and desptop...

BOTH AMD's and Intel's profits went down last quarter from Q1'06. AMD raised what ~100-150mil in profits and Intel ~800mil?

10:07 AM, July 30, 2006  
Anonymous Anonymous said...

"The sooner AMD fires Hector and the rest of his loser management team, the better off AMD will be.

Hector is for AMD what Craig Barrett was to Intel. A crappy CEO that made a lot of acquisitions instead of focusing on core technology and business fundamentals."

While admittly I'm an Intel fan, this comment is plain wrong. Ruiz has been good for AMD as he, I believe, has made 2 key decisions.

1) you need to invest in capcity. While AMD is a little late with F36, if they can execut on new abs they will be in good shape (they would have a bunch more market share had that fab decision been made a year or 2 earlier)

2) He seems to have realized that beyond the retail market it is not just about chips - it's about having a stable platform (thus the ATI merge). This will give them a far better chance at mobile and business desktop markets. I'm not sure if the ATI merger was the way to about this but it probably is the fastest way to do it.

The bottom line - it is more than just products. K8 was a very good (and superior) architecture yet the market share gian was hindered by:

A) limited overall capacity (which AMD is now, finally, trying to correct)
B) slow market penetration - this oculd be debated for many reasons:
- Intel like it or not has better marketing and brand name, although AMD has improved on the brand side. - Dependency on 3rd party vendors for things like chipsets, without being able to provide the market a reference platform and the allow 3rd party vendors to join in when product ramps.
- Pricing flexibility; Intel has had, and continues to have, the ability to cut prices due to manufacturing capacity and the fact that they can absorb (short term) lower ASP's.

10:29 AM, July 30, 2006  
Blogger Sharikou, Ph. D said...

The sooner AMD fires Hector and the rest of his loser management team, the better off AMD will be.

Whenever Hector Ruiz's name is spoken at Intel board meetings, it's with deep fear. Ruiz is a genius, and he is cool headed military strategist who dares to make unexpected moves that will totally daze his enemy. The anti-trust lawsuit, the dual core duel, the DELL alliance, the ATI acquisition, all hit the pain spot. Intel is being forced to lay off massive number of workers. Intel revenue is dropping at alarming speed. With the current trend, Intel will BK in 5 to 7 quarters.

10:41 AM, July 30, 2006  
Anonymous Anonymous said...

" Anonymous said...

ES = Engineering Samples.

These represent probably first stepping with many bugs sent to OEM manufactures to start validation, platform stability and BIOS work.

Figure another 6 months best case before you see production then another 6 months before you see volume.

ES samples of Conroe were out there 9 months ago.

AMD is very very late, they are in trouble.

ES.. how sad. This is very bad news of how far they are behind. "

Last time I seen, Intel was late by 3 years, even with reduction to 65nm they were inferior to AMD for almost a year.

11:07 AM, July 30, 2006  
Anonymous Anonymous said...

"Whenever Hector Ruiz's name is spoken at Intel board meetings, it's with deep fear."

Didn't know you were on the Intel board, I'll have to check the stakeholder's report again.

All kidding aside the Dell and ATi moves were good accomplishments but the dual core duel hurt AMD more than it helped. Intel pushed dual core, intially, largely for strategic reasons as they knew AMD could not keep up with capacity (as evidenced by the fact that the majority of AMD's sales are still single core). Had AMD stayed single core they could have gained another ~10%-15% effective chip capacity (assuming ~30-35% of their chips are dual core now). This would have enabled them to gain more market share (and more revenue) while eventually moving to dual core when 300mm and 65nm came up. In my view it was a tactical error for AMD to get drawn into the dual-core fight earlier then they needed to.

11:21 AM, July 30, 2006  
Anonymous Anonymous said...

It's all about yields.

When AMD starts 65nm productin, it will be close to mature yield.
Unlike Intel, they can claimed to be the first, but it suffered from poor yields for months.

I guess Intel can afford to be wasteful.
Intel can afford to hire another design team in Israel while their engineers in the US tried to redesign P4, a donkey according to an Intel engineer in Israel.
Made me laugh when Pat Gelsinger is so proud of himself and takes credit.

12:24 PM, July 30, 2006  
Anonymous Anonymous said...

re:"In my view it was a tactical error for AMD to get drawn into the dual-core fight earlier then they needed to."

Wrong. A lot of Opteron system upgrades were to swap out a 1 core CPUs with 2 core CPUs.

Anyway. Decisions are based on customers' inputs/demands.

12:56 PM, July 30, 2006  
Anonymous Anonymous said...

"AM2 is a dud. Opteron is overpriced."

amd, as well as intel, is commercial company. so they have to get money to continue development and pay dividends.

opteron is priced exactly as it should be priced, fastest and perfect scaling cpu in 4x and 8x world. 2x is under pressure now, but its still quite competive.

AM2 is preparation to next gen core, k8l. it makes no big difference in most modern apps, but makes cosiderable difference in scientific FP loaded apps, look on eg spec fpu. also its quite smart to flood market with new socket, ahead of K8L release - look on conroe story, ppl can get some cpus on +50% of intel price, but u need to dig really hard to find mortherboard for it!

1:14 PM, July 30, 2006  
Anonymous Anonymous said...

Its about Yields..

I can say for a fact when INTEL ramps and transfers they have yields that are better then world class.

I've seen the yields that IBM tolerate, Japanese memory factories and INTEL. INTEL is world class when... Hector and others make the excuse about yields.

The bottom line is you only get the high yields. INTEL ramps early, hard and thus get high yields before everyone else.

Anyone who tells you anthing else knows nothing about silicon processing or yield improvement.

Keep believing that BS the AMD guys feed ya... vs learn to think and maybe learn yourself. But its hard to learn for many of the fanboys because they aren't even in the business so they don't know jack.

You know who

1:48 PM, July 30, 2006  
Anonymous enumae said...

Edward said...

"AMD might be a year late with 65nm, but Intel was 3 years late with Conroe. Ever think of that?"

Thats a great point.

2:01 PM, July 30, 2006  
Anonymous enumae said...

muzigaz said...

"Isn't the TDP most important thing in mobile cpu?"

I would think its more about battery life and performance.

2:03 PM, July 30, 2006  
Anonymous Anonymous said...

Late or or not I don't care.

Like in my personal life, late bloomer, or late developer didn't matter. In the end the one with the most money wins!
..

That is the bottom line.

who has made more money over the past 3 years, past 10 years... that same company will make more then AMD per MS % over the next 10 years too.

Give it up.. you lose face it.. suckers you have backed the wrong horse. I know its difficult to accept.

He who has the most money wins...

3:38 PM, July 30, 2006  
Anonymous Anonymous said...

"Figure another 6 months best case before you see production then another 6 months before you see volume."

MMM - "Obviously you're a moron. I think it was AMD themselves who said Dec06 was 65nm for sure, hmm.."

Madmodmike - Dec is ~6 months (OK, 5 if you want to get technical) from now...how is the previous poster a moron? Also AMD themselves said they won't see a 90nm/65nm crossover in F36 until midyear, throw in all of the 90nm F30 production and 50% overall crossover won't happen until at least H2'07. How again is the estimate of additional 6 months for volume a moronic comment?

"Because AMD is behind on changing process, means they live longer with mature yields and reduce costs to end-users"

As you like to (incorrectly) point out all of the time, if AMD is always starting out technology nodes at mature yields, how is cost improving over time?

With 2X # of die/wafer from one node of scaling, from a cost perspective you can take a large yield hit and still break even with the previous node (I'm ignoring some of the added technology node scaling cost from process complexity)

Note die #'s are not actual:
200 die/wafer(90nm) @ 100% yield = 200 good die out
400 die/wafer (65nm) @50% yield = 200 good die out

So assuming current yields are not 100% and the fact that AMD's yields won't be <50% (which with APM everyone claims they won't be), it makes no sense for AMD not to convert rapidly. The only reason why they haven't done so by now is either:
1) technology wasn't finished with development
2) yields are not good (<50%)
3) they can't get the equipment until now (some of the advanced litho tool have >1 year leadtime)

I would have to say up until now it's been item #1.

3:58 PM, July 30, 2006  
Blogger Sharikou, Ph. D said...

Also AMD themselves said they won't see a 90nm/65nm crossover in F36 until midyear

No. 65nm crossover by end of 4Q06 in terms of wafer starts. By mid-07, it will be 90% 65nm at FAB36. FAb30 and Chartered will be 90nm. Go listen to AMD June 1 analysts meeting.

The best thing is AMD will start 65nm at mature yields. Once AMD got the thing nailed, its APM allows very rapid transition.

4:05 PM, July 30, 2006  
Anonymous Anonymous said...

"Late or or not I don't care.

Like in my personal life, late bloomer, or late developer didn't matter. In the end the one with the most money wins!
..

That is the bottom line.

who has made more money over the past 3 years, past 10 years... that same company will make more then AMD per MS % over the next 10 years too.

Give it up.. you lose face it.. suckers you have backed the wrong horse. I know its difficult to accept.

He who has the most money wins... "

Damn... thanks for your personal story. I guess I invested in the wrong company by your theory in Google when Microsoft was the way to go...

Try Again...

"AM2 is a dud"

Good, don't buy one, and go get a Conroe NOW!! And when Intel's next chip comes out, you'll be forced once again to buy a new motherboard...

Come X-mas time, I'll get a Dual Socket Am2 and get one Dual (or quad?) CPU. When I can afford another one, I'll drop in another CPU for seamless upgrades. I'm an enthusiast, but also a smart shopper who looks out on the horizon.

I've had Intel computers in the past, and really like my Athlon64... I like choice, and not being forcefed crap. :)

4:25 PM, July 30, 2006  
Anonymous Anonymous said...

APM helps yields.. you are truely stupid. APM doesn't help yields.

Go study what APM is really, and then figure out how you drive yields. Then come back and tell us all what APM really does and why it doesn't help yields.

Try to stop posting stupid things

You are truely stupid.. You sound like Sharikou now .. pss that wasn't a compliment but you are probably to stupid to figure that out.

APM help yields.. to funny.

8:02 PM, July 30, 2006  
Anonymous Anonymous said...

I don't mean to burst your bubble, Sharikou, but quit trumpeting APM as some amazing technology unique to AMD. Every relatively advanced fab in the world has an APM-like system, and you know it. Just because AMD is the only one that bothers to brag about it doesn't maek them the only one that uses it.

Perhaps AMD's software is a tiny bit better than competitor's software, but that is the only difference that would exist.

8:25 PM, July 30, 2006  
Anonymous Anonymous said...

"Whenever Hector Ruiz's name is spoken at Intel board meetings, it's with deep fear"

Yes, Sharikou.. You would know.. that proves the rumors that you were an ex-canned Intel employee!

We understand your bitterness now.

Mad Cow Disease: can you name two english words that do not start with fan**** or moron..

Yeah, I did not think so..

8:51 PM, July 30, 2006  
Blogger Sharikou, Ph. D said...

"Whenever Hector Ruiz's name is spoken at Intel board meetings, it's with deep fear"

Yes, Sharikou.. You would know.. that proves the rumors that you were an ex-canned Intel employee


So, you must think I am Intel's chairman of the board?

BTW, I also know AMD people love Paul O and Craig B...

8:58 PM, July 30, 2006  
Anonymous BigBadWolf said...

With ATI acquisition AMD has ,once again caught Intel pants down.

Everyone in the world knew what was coming but the incompenet , ignorant and arrogant Intel management team.And Paul O is a total idiot.
I pity for you fannies.

Here is the example of sheer incompetency , just incase you haven't noticed.

http://www.theinquirer.net/default.aspx?article=33287

9:06 PM, July 30, 2006  
Anonymous BigBadWolf said...

Anonymous coward wrote ,

"Mad Cow Disease: can you name two english words that do not start with fan**** or moron..

Yeah, I did not think so.. "


HaHa i reckon shari has hit the tender spot.I hear you crying like a baby.OUCH

Now go back Fannboy and start singing INTEL IS GOD.

9:12 PM, July 30, 2006  
Anonymous Anonymous said...

"No. 65nm crossover by end of 4Q06 in terms of wafer starts. By mid-07, it will be 90% 65nm at FAB36. FAb30 and Chartered will be 90nm. Go listen to AMD June 1 analysts meeting."

Actually we were both - wrong AMD said crossover in Q1'07. At least I'll admit when I make a mistake.

Still haven't answered my question with F30 + F36 + chartered able to supply more than 100% of the world's CPU's in 2007 why bother with a NY fab, especially given the fact that F30 (200mm,90nm) will convert to 300nm@65nm in 2008 which would greatly increase that fabs production (according to you it is 4X from wafer size and node scaling). Seems like AMD will be way over capacity with all of these perfect yields and perfect ramps.

Oh and I forgot AMD also said they are on track for "volume" production on 45nm by mid-2008, which will double their capacity yet again.

Using your initial assumptions and factoring in 45nm conversion and F30 conversion AMD will be at 200-300% of the world's CPU capcity by end of 2008 - why bother with another 3Bil fab in NY?

Perhaps your intial calculation is a bt off and AMD is not building another 3Bil fab for no reason? AMD said they were targetting 30% - why would they be building more fabs and be so far off from your analysis (clearly AMD is not as smart as you are, they must have made an error on capacity planning?)

9:47 PM, July 30, 2006  
Blogger Sharikou, Ph. D said...

Still haven't answered my question with F30 + F36 + chartered able to supply more than 100% of the world's CPU's in 2007 why bother with a NY fab

Dude, OLPC alone needs a FAB, 1billion children out there. Add ATI graphics cards, cell phone chips, HDTV chips...AMD64 everywhere is the vision...

9:54 PM, July 30, 2006  
Anonymous Anonymous said...

"cell phone chips, HDTV chips...AMD64 everywhere is the vision..."

Didn't realize cell phone chips and HDTV chips were based on X86 64bit architectures. I guess I'll have to run some Sciencemark on my TV.

10:37 PM, July 30, 2006  
Anonymous Edward said...

"We understand your bitterness now."

I find such response rediculous. Can you guys focus on the points and facts, but not hypotheses in your head?

Even if Sharikou WAS an ex-Intel employee, you still logically understand nothing. Sometimes he has his points, sometimes he doesn't; it's completely irrelevant to where he worked. Not to mention all you said were no fact but your own assumptions.

Such responses like yours are disgusting. I wonder why you bother to post them? Go away and play elsewhere! Don't come and expose your childishness.

I also don't think Sharikou should've bothered to permit them, BTW.

1:06 AM, July 31, 2006  
Anonymous Edward said...

"Good, don't buy one, and go get a Conroe NOW!! And when Intel's next chip comes out, you'll be forced once again to buy a new motherboard..."

Exactly. For example, I won't buy Conroe, because I can drop an X2 4600+ into my 939 motherboard and beat out E6300 or E6400. That's less than $250. Had I bought ANY Intel-based machine 2 years ago, I'd have to spend twice that and reinstall the whole system in order to upgrade to dual-core.

So again if I buy a new dual-core system today, I'd buy AM2, because that MB can live past K8L and even future AM3 chips.

1:27 AM, July 31, 2006  
Anonymous Edward said...

"Didn't realize cell phone chips and HDTV chips were based on X86 64bit architectures. I guess I'll have to run some Sciencemark on my TV."

Didn't he say 'vision'? So in your dictionary 'vision' is used with the past tense?

1:29 AM, July 31, 2006  
Blogger Ho Ho said...

AMD64 everywhere is the vision

I would take a special-purpose CPU over general-x86 CPU any day. x86 is nice in general computing but it sucks in special purpose things. Especially in big number crunching like decoding HDTV and power usage as in cellphones. It's not that much about the CPU manufacturer but the ancient instruction set. x86 is almost 30 years old, after all.

4:16 AM, July 31, 2006  
Anonymous muziqaz said...

enumae, I was talking about cpu alone [;
then company introduces new line of mobile cpus, I expect them to display the vital things of that cpu.
I'm not interested in cache size on mobile cpu if they do not state heat wastage. CPU is one of the main things which decides the battery life and perf/watt.
[;

6:12 AM, July 31, 2006  
Anonymous The Sheepshagger said...

Man... There is a lot of low hanging fruit around here these days.

A couple finer points... APC doesn't increase yield, but it does enable tweaking the line for better bin splits at sort. This is something that allows a semi-conductor company to produce more 'boutique' or high end parts, but ultimately comes at the expense of budget chips. As usually, this means you yield faster parts from the center of the wafer, but this should NOT be confused with yielding more die per wafer. The faster chips were from regions that previously yielded good parts. That's it. APC has no bearing on suddenly creating MORE unusable die per wafer. Get it, got it, good?

Secondly... transitioning from 90nm to 65nm does NOT automatically equate to DOUBLING the die per wafer. Someone used the example of 90nm w/ a fictional 100% yield equally 200 chips and translated going to 65nm w/ 50% yield equaling 200 chips because there would be 400 possible die. It would be a WONDERFUL gift to all chip companies if that was the case but it just is not so. I think what the original poster was alluding to was retooling a factory running 200mm wafers to 300mm. Transition to smaller technologies allow can increase clock speed and reduce power consumption, however it is NOT a given with all chip designs and there is some work that always needs to be done to get there.

As far as what boardroom is scared of who's CEO... it's all bunk and amounts to nothing but idle schoolyard taunting equivalent to: "My dad can beat up your dad."

What I would like to say is that AMD is a much better company under the direction of Hector Ruiz than they ever were under that noodle head dilettante Jerry Sanders. (GOOGLE SEARCH: dilettante Jerry Sanders) Additionally, Intel is much better lead by Paul Tortellini than they ever were by Craig Barrett. Really and truly, he is the reason that Intel is turning around.

As for as the 'massive layoffs' at Intel... If you really don't know what you're talking about, then perhaps you shouldn't speak at all. The reason Intel lost their edge of the last 3-4 years is because of the bureaucracy to get things done in house. Know this to be true: Intel purchase a slew of companies between 1998 and 2004 and very few of the technologies acquired paned out. I can't address the exact number of companies that were purchased but it is somewhere between 30 and 50... most were never publicized. Intel purchased the technology and the people from these companies and headcount swelled by 12 - 15k.

In 2005 alone, Intel hired ~15k people; mostly people directly involved in staffing their new Fabs. Executive management was caught off-guard earlier this year when they were informed there were over 100k employees. You say there are 'massive layoffs' inside Intel and frankly, I laugh. In May, Intel leaned out the Human Resources depart of contract employees (read
~500 non-Intel employees). In June, Intel sold the XScale business to Marvel and 1600 employees when along. In July 1000 first line through executive level managers were let go and nearly ALL of these were people with redundant jobs. One example is there were 2 managers doing the same job between the 2 they had a single direct report. Will there be more layoffs? Perhaps.

In a typical year, Intel has 10-15% turn-over or natural attrition. So far between the last 3 months, roughly 3% of the payroll has been spun off to other companies or laid off. That being said, I HIGHLY doubt Intel will go for the whole yearly attrition figure in lay offs but if they did, you can beat your ass a leaner Intel is only going a force to be reckoned with.

11:37 AM, July 31, 2006  
Anonymous Edward said...

"Secondly... transitioning from 90nm to 65nm does NOT automatically equate to DOUBLING the die per wafer."

I don't understand. Sure the capacity increase won't be 100%, but assuming eventually they reach the same yield, wouldn't it be close to that, perhaps 80%? From what I can see, the second-order estimation would be (n^2 - 2n) where n is 90/65.

Also, why's there any connection between Intel "lost their edge" and "get things done in house"? Does Intel's in-house engineering really suck that much?

Any computer architect knew 3 years ago that NetBurst is inefficient. In fact, IIRC, it was mentioned and compared to p6 in H&P's textbook. Does this has anything to do with Intel's company purchase or other in-house activities? How do you justify such claim?

3:48 PM, July 31, 2006  
Anonymous Anonymous said...

"I don't understand. Sure the capacity increase won't be 100%, but assuming eventually they reach the same yield, wouldn't it be close to that, perhaps 80%? From what I can see, the second-order estimation would be (n^2 - 2n) where n is 90/65."

Edward: You are correct - physically there are ~2X the die/wafer (30% linear scaling means 0.7*0.7 or an area scaling of ~50%). So it is just a matter of how much the yield drops (if any). My point above was even if your yield happened to drop by 50% (which it would't) you would break even in terms of good die/wafer; everything better than a 50% yield drop means more good die /wafer.

So if you can get yield within 50% of the previous node, it starts to make sense economically to do the transition rather than "aging the process like a fine wine until mature yields are met" like some have claimed it is better to do. I made the 2X claim because of the ridiculous claims of "near perfect yield from APM" and "65nm will start at mature yield"

The reality is there is some time to get yield mature - if you look at AMD's own graphs from analyst day they have a plot of PRODUCTION VOLUME vs YIELD, with the yield plot going up over a certain amount of production volume before it hits the mature line. Thus the claim that AMD starts production at mature yield is BS or the graph would be just a bunch of flat horizontal lines - AMD's claim is that they are doing this on newer generations more quickly (not that the time period = 0)

The one point that does start out at mature yield is 90nm/SOI; what people are neglecting is the plot of 90nm which took time to get to yield and the 130nm/SOI line which showed a similar trend. All this point says is AMD was able to port the SOI process from 130nm to 90nm, not that 90nm started at mature yield - this took time as evidenced by the 90nm/bulk Si plot.

AMD clearly has made improvements in getting new process technology nodes yielding faster (I don't remember if it was the 0.25um or 0.18um that was a complete mess), but these claims of starting out at mature yield are absurd (as born out by AMD's only foils which show yield increasing to maturity with increasing production volume).

The line will get blurred even further as AMD is moving to a more incremental improvement approach for performance improvement vs Intel's approach of puuting in all of the improvements up front. I won't comment on which approach is better both approaches have their own set of pros & cons. I would speculate that AMD's yield plots will look more like a fine tooth saw tooth pattern with dips correseponding to when they introduce the incremental improvements whereas Intel's plot will look more like a rough saw tooth where there will be bigger dips at the beginning of each node but once it hits stability it will remain that way until the next node transition.

12:41 AM, August 03, 2006  
Anonymous Fragstorm said...

lol it will be a long time before amd gets 65 nm dies on their chips. With quad core and 90 nm that chip (4x4)is going to be huge (if amd even pulls it off this year), and extremely hard to keep cool. Which means crappy overclocking, and crazy power consumption. Not to mention the price of the chip itself, (which is rumored to be over 1500 usd just for production)

Intel is definetly loosing money now due to amd's price reductions, however their 65 nm fabrication process was an investment that will pay high dividends. Also the pentium
D's, 4's and extreme editions all still have alot to offer. (especially to the mainstream market which isn't just concerned with gaming!)

Quad core Kentsfields have already been released as engineering samples (ES for whoever asked), while amd has shown nothing of there 4x4!!!!

AMD is supposedly going to be releasing 4x4 (which is there ONLY HOPE given the performance of conroe and the yet to be released quad core Kentsfield).

Amd is a wounded animal. Pricecuts, which they are loosing money on are the only thing keeping them afloat. Answer me this question amd fan boys: IF THE PRICES OF AMD CHIPS WERE THE SAME AS THEY WERE 4 MONTHS AGO, WOULD YOU GO WITH CONROE? Only a chauvenistic, patronizing idiot would say no to that.
why? because the conroe is both more energy efficient, faster and overclockable than anything that amd can muster.

Amd and there line up of once dominant cpu's look frail in the face of conroe. Amd is a dying animal, no matter what the propaganda of this so called " journal of pervasive 64 bit computing BS tells you.
A better analogy is a colony of organisms that just got decimated by rubbing alcohol. Who knows, maybe they will mutate and spring back to life. (i'm hopin that they will, so that intel will have at least one competiter. It would be great if amd could top conroe (with price and performance) because i just want the best deal like everybody else;
that is found in Conroe

1:17 AM, August 13, 2006  
Anonymous FRAGSTORM said...

all you amd lovers have nothing to go on but your own stupidity! show some results!!!
(If you feel bad because you paid 700 dollars for an fx-62 when you could have paid 190 for an e6300, there are better ways to express your anger. The first way is to sell that piece of shit chip to some chump who doesnt know the difference. Or you can stick vehemenantly to your overpriced, weak , green ways and repudiate all evidence that you come by.
READ SOME (AS IN MORE THAN ONE, POST RELEASE DATE) REVIEWS BEFORE YOU TALK SHIT ABOUT CONROE!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!!!!!!!!!!!!!!!!!!!!!

1:24 AM, August 13, 2006  

Post a Comment

Links to this post:

Create a Link

<< Home