AMD's K8L to frag Intel's Clovertown by 40%
Intel BK by 2Q08 is set in the stone. 40% architectural advantage, that's K8 over Netburst all over again. Expect Patty to be begging on the streets soon.
"Based on what we know about our product and everything we understand about our competitor’s roadmap from all sources, we’re very bullish that when we introduce our native quad-core in the middle of the year, we’ll capture definitively the performance and performance [lead] and, by the way, do so in a way that doesn’t require new platform investments on the part of our OEM customers."
When Dirk Meyer the Alpha architect who says the above, Intel dudes know their days are numbered.
Just wait and see.
AMD's 4Q06 result is quite impressive. In this tough environment, AMD grew units by 19% sequentially.
posted by Sharikou, Ph. D. @ 1/23/2007 07:47:00 PM
90 Comments:
I have to agree, based on the benchmarks AMD showed, it is clearly 40% better in a "variety of workloads". I mean task manager alone was what 80% faster?
As it is now almost Feb, perhaps AMD can provide a better release date than "mid-2007" - is their manufacturing that unpredictable?
Ph.d's don't use terms like "frag"
Ph.d's can type in clear, concise, sentences, not broken english.
Congratulations, you succeed in proving yet again that your education is limited to G.E.D.
"AMD's 4Q06 result is quite impressive"
I must say the gross margin drop was quite impressive and that was EXCLUDING graphics biz...not quite as impressive as the GAAP loss, but still impressive. (And just think your Q3'06 GAAP loss predition was just a few quarters off, and a bit off on the company name).
"Overall server processor unit shipments were essentially flat compared to the third quarter and ASPs were down significantly."
Always a good business model to give up the most lucrative market and focus on the lower margin ones...gotta love Dell - let's increase our shipments by 19% but somehow lower earnings while doing so.
Oh and look at the inventories...
AMD market share in Q3 ~23% (forget if this is revenue or unit based)
19% sequential increase would mean max share of ~28%...of course the overall industry probably grew a bit in Q4 so you cant's just increase the Q3 market share by 19%....
Tell us again how this is a 40% market share runrate?
While waiting for you to spin AMD's Q4 2006 earnings, I read one of your older post "AMD: time for the kill".
In said post you make a statement...
Even with massive cuts on desktop CPUs, AMD can still profit nicely from Opteron, Turion 64 X2 and 4x4.
I won't say any more about that.
In the article in which you linked there is a comment "Barcelona's clock frequency will be lower than that for the company's dual-core chips."
My question is, can you give some kind of idea of what you think clock speeds will be?
Funny how AMD is now telling customers to go to K-late when it is even out yet.
What is funny about this is when Intel did this, all the AMD fans yelled Osbourne ffect...essentially now AMD is saying Opteron dual core is not even in the equation as to compete with Cloverton you should wait for K-late?
Wow, AMD must really be getting smacked in server space right now - they report very, very good growth in mobile, a 19% unit growth yet revenues and earnings are terrible - they must really be discounting Opterons to compete with Woodcrest in 1P and 2P space or losing serious market share in server space, or perhaps both.
When you factor in all the growth in mobile and 19% volume growth - ASP's must have gotten completely crushed (especially as mobile really isn't part of the price war and should have given a much better boost in earnings, margin and revenue...)
Any theories Sharikou on the seemingly 2 incongruous data points? (mobile and unit vol gorwing yet revenue flat and gross margin down?)
Interesting that you didn't post that AMD reported a 4th Q loss, some journalist you are, way to not see things from both sides, although i'm not suprized. I'd be pleased to wait once again for your predictions to be wrong once again. Once again you've proved yourself as having [P]ecker [H]ead [D]isorder
Yes, it is quite impressive to lose over $500M in 1 quarter for a company of AMD's size. The complete and utter lack of any credibility of their management is pretty astounding. The first time I ever read anything by you Sharikou was on the AMD stock message board on yahoo.com, seems the magnitude of your technical ineptness is matched only by your ineptness in all matters financial. Truly impressive!
LOL
AMD will be BK before Barcelona even goes to production.
Note the very carefuly wording by the AMD with words like " expected Clovertown"
They talk of 3.6 times a dual core Opetron at the same frequency then quietly ignore the fact the quadcore will have to run much slower clock frequency. No talk about they are comparing a dual core to a quadcore so 2x performance is a default, and how come on 65nm and with a big L3 cache its only 3.6 times with a new architecture. Sounds pretty lame to me a response for somethign that has been in the works since 2003 or something..
Beside shouldn't it beat INTEL Core2. Core2 has is old news. By the time Barcelona is here it will have a short quarter of glory before Penrym 45nm quadcores come out...
Bottom line K8L should be faster then Clovertown, the problem its 7 months late and in 9 months expect Penrym to be the talk of the industry on 45nm no less.
Too little to late..
Hey how about them quarterly earnings heah Sharikou? AMD sucking red and very weak forward looking guidance. How about them Sun Day adds and Suns surprise heah.
Poor AMD
Oh come on. This is just low, even for you. This is some bloke from AMD saying how great Barcelona will be. There's no benchmarks or anything. Yet you accept this and use it as a proof that Intel will go BK.
Yet when Intel provided early benchmarks for Conroe that were shown to tech websites like Hexus you claimed that they were clearly cheating, and rigging the AMD system for poor performance.
Sharikou has hit a new low!
OMG.. intel bk in 2008 ? i think you need a real job and ask your uni to revoke your phd...
K8L does look good but it will have to compete against 45Nm processors from Intel then, not Clovertown, so the comparison is not that valid. Will be interesting to see though.
In regards to Q4 figures the market share was the one bright spot. More worrying for AMD they increased revenue in 2006 over 2005 by 40+% but the profits were only up 9% due to massive expenditure on one thing or another.
They are going to find it tough until Q3 2007 and then they need to ramp K8L up quickly. Might be a tough 2007 in total.
Intel has now not got enough time left to go bankrupt by 2H2008, it's revenue forecast for Q107 looks pretty good and so they would have to lose all their money in 4 quarters. Not likely.
how come the total sales is lower than Q4 2005 ??
what about the sales revenue of ATI ??
Phew! I was afraid you would push the Burger King feast day back Sharikou! I'm glad it's still on for Q2 2008! :D
Indeed, VERY impressive!
AMD went to war with a billion dollar powerhouse (a.k.a monopolist), and yet came out victorious in ALL areas after four quarters of this.
The ONLY two places that AMD showed 'some' weakness are: margins and earnings. And yet, Intel, the Goliath which HAD to declare a full blown war on AMD in order to prevent profitability as to achieve the ultimate goal of limiting their expansion in production capabilities... well, FAILED!
Margins are considerably low, but better than I ever expected considering Intel's shear size. But most surprisingly, AMD did still post a profit (although small, it's STILL a profit).
And as proof, that DELL did NOT get the amazing discounts that you all believed: AMD's desktop shipments and mobile shipments were WAY up (a new record), but yet the margins in these areas were barely untouched (slightly lower). The only segments where the margins were hurt big-time were servers (and understandably).
Now AMD must keep it's strategy for only two more quarters of this, maybe results will be a little worst for the first half, so I wouldn't recommend buying AMD shares just yet, but don't wait too long. For the second half, there 'should' be a lot working in favor of AMD's strategy, such as 65nm, Barcelona, (a true) 4x4 platform, ATI division will be restructured, enhanced and integrated properly, etc. and most importantly better margins (for servers) by the end of 2007 (due to Barcelona).
For those accumulating shares during the first half, you should have no problem exiting 2007 with a DOUBLED investment.
is their manufacturing that unpredictable?
Intel 45nm is going to be released this year. Is their manufacturing that unpredictable?
Maybe you could post the day and the month it’s going to be released.
Oh and look at the inventories...
In case you didn’t read the news, AMD is going to build inventory, because of FAB30 transaction to FAB38. The FAB is going off-line for some time, if you don’t have inventory you have nothing to sell.
At least its good processors, not bad ones that Intel is still building (Celeron, P4 and PD).
Double Kill!
My question is, can you give some kind of idea of what you think clock speeds will be?
Sure. It will be in line with the clock of Intel quad (double dual core), less 12% lower clock, but not the doubled power consuming.
Multi Kill!
Funny how AMD is now telling customers to go to K-late when it is even out yet.
Funny how Intel still sell/manufacture Celeron, P4 and PD when Core 2 is already OUT.
Mega Kill
they are comparing a dual core to a quadcore so 2x performance is a default
It is? Who told you that, your 1st grade teacher?
Ultra Kill
expect Penrym to be the talk of the industry on 45nm no less
Haaaa let me see, AMD 90nm to 65nm nothing special. Intel 65nm to 45nm is super duper.
Monster Kill
You can tell it's not good when AMD starts crying "MONOPOLY" again.
It must be convenient to have such an excuse everytime you screw things up.
AMD's new sales pitch from the conference call: "Best value in the industry" - Rivet. ROFL! Sounds pre-Athlon era to me.
My humble op.: AMD will have to provide their own (i.e. ATI) chipsets for the most demanding applications (i.e. servers), which will have to be technically outstanding, or else they will not compete as favourably as needed. So their B-plan would remain in mobile integrated solutions, which seems a rather cheap bussiness.
I dare to say AMD may not get a second chance for a clear technology leadership unless there will be some major flaw in Intel's product line uncovered.
The first K8L batch will be codenamed "Sharikou edition" featuring 1 Hz clock and blasting C2D by 180% requiring 1 W power supply. Will be mine!!
Not taking the 600M loss into consideration, yes AMD is doing fine *rofl*
AMD is 1 step behind Intel in EVERYTHING, how the heck could they compete?
Can anyone comment on:
1. Expected performance improvement of AMD's 65nm process over the next six months?
2. What such potential process improvement means to the product mix and pricing.
3. What % of the installed Opteron server base is expected to upgrade to Barcelona (assuming AMD is able to convince those considering a new system from switching to Intel).
4. How many chips will that be in Q3 and Q4 and at what price would you expect those chips to start?
Thanks to anyone offering serious consideration.
Sharikou, why do you still insist that Intel will be bankrupt by Q2'08? Can't you see the facts before your very eyes? AMD is in a LOT worse shape than Intel and even I don't think they will be bankrupt by then. Since there is absolutely zero penalty when you are proven wrong, there is no reason for you not to keep repeating this bogus claim but doesn't it hurt your integrity? Do you have any integrity? From what I have seen on this blog, I think the answer is NO.
And now something for all Intel suckers :
http://www.theinquirer.net/default.aspx?article=37154
CRY CRY CRY CRY CRY CRY CRY Intel |
"fanbuoys" !
Sharikou: let's celebrate Intel's last days !!!
Sharikou, Your prediction seem to fail miserably.. Even the weather man has been more accurate than you are.
For those accumulating shares during the first half, you should have no problem exiting 2007 with a DOUBLED investment.
Uhh, right. AMD stock is performing so well (down 7+% right now on the day).
I actually think AMD is a good company and it will rebound, and I'm sorry for legitimate investors who've lost money on them lately, but...
MAN am I laughing my ass off at all you pathetic AMDroid losers who sock all your money into AMD stock because you're semi-retarded. It's _GREAT_ to know how much money you've lost on AMD stock. I actually get a smile each time I think about it.
I may invest in AMD when they hit like 15. That way it feels kind of like I'm actually taking your money away from you because I'm not an idiot who invested in them when they were 20-25+.
Ph.d's can type in clear, concise, sentences, not broken english.
But it’s the only way to make sure that morons understand.
Especially Fanatics Intel guys like you when don’t have arguments write bullshit like you.
Yes, the smell that is coming from the big green K-chip company is... dead meat.
K8L won't ship in volume until mid-2008. By that time, AMD will have lost 50% of their server share.
Opteron/DDR2 is a dog that should never have been made. Basically a complete waste of all the momentum AMD had built in the market.
Same goes with AM2 DDR2 Dog Meat.
Nothing can change the fact that AMD is run by a bunch of retards that get their orders from Bill Gates.
There is no innovation at AMD, just paper launches and long academic discourses from Sleepy Hector.
Yawn....
The "LOSS" was almost 100% in charges taken for the integration of the 2 companies. A company WANTS to take as big of a chunk of their expenses in ONE qaurter ONLY. Throw in some office movement expenses, and all kinds of other things, and charge it all in that ONE quarter... That's how everybody does it! Next quarter, AMD will NOT report a loss... $550 million of the $574 was "integration" related, and of course, AMD threw in $27 million for employee stock-based compensation... Just lump it all in the same quarter you are going to have big expenses, so they only show 1 quarter with large expenses.
AMD increased their revenue 33% (excluding ATI revenues) year over year - Way to go! ($3.935 Billion to $5.251 Billion)
Intel BK in 2008? All AMD has to do is make them unable to pay their debts... then banks and other lenders take over. I believe most people here think that Intel will have to sell ZERO chips to go bankrupt! Not so... Make them not able to pay their bills, and banks come'a'knocking! If Intel's Average Sales Price per chip drops drastically, they will have to cut staff drastically to stay ahead of the Bank hounds barking at their tail. Intel has $9 BILLION in debt... (http://sharikou.blogspot.com/2006/10/another-look-at-intels-balance-sheet.html)
If Intel wants to turn to 45nm process on most of their fabs, they will need BILLIONS in upgrade costs... remember, that's why AMD STAYED at 90nm... postponing the Billions in upgrade costs. But 45nm might be their only resource to combat their powerhungry processors...
This will be an interesting year.
AMD AM2 to frag AMD by 40%.
It was a shame and a big mistake to throw up socket 939 owners. AM2 processors performance-wise are a flop when compared with the good "old" skt 939 CPUs clock-to-clock, minimal or not any speed gain at all. Well, I need to upgrade my "939" system, I cannot drop in a fresh, faster 5000+ or better CPU, my only chance is switching to a C2D system and as me many "939" people are going to switch. Bye bye AMD, you have fired your feet, and yes, Intel has Clovertown CPUs, you have Clown CPUs!
This blog is dead.........period!!!
enumae dijo...
While waiting for you to spin AMD's Q4 2006 earnings, I read one of your older post "AMD: time for the kill".
In said post you make a statement...
Even with massive cuts on desktop CPUs, AMD can still profit nicely from Opteron, Turion 64 X2 and 4x4.
I won't say any more about that.
In the article in which you linked there is a comment "Barcelona's clock frequency will be lower than that for the company's dual-core chips."
My question is, can you give some kind of idea of what you think clock speeds will be?
8:28 PM, January 23, 2007
interestingly that a person who claims to be smart and constantly prove the "FACTS" hasnt researched that the loss is due of the ATI adquisition.
if you get the ATI adquisition out, AMD still earned money.
gotta love oportunists!
Quite the string of downgrades. OUCH!!!
Advanced Micro (AMD - commentary - Cramer's Take - Rating) downgraded at Friedman, Billings: AMD was downgraded from Market Perform to Underperform at Friedman, Billings Ramsey. $10 price target. Company said that the situation was far worse than previously expected. AMD has also lost its technical advantage, and is facing a CapEx crunch in 2007.
Advanced Micro estimates cut at Prudential: Prudential is cutting its 2007 estimates on AMD from 42 cents a share to a loss of $1.01 a share due to expectations of very aggressive pricing environment.
Advanced Micro price target cut at Goldman: Goldman is slashing its target on AMD to $12 from $16 after company reported weak Q4 results. Disappointments included gross margins of 39.7% and higher R&D and interest expenses. Believe business model is structurally flawed. Maintain Sell rating.
Advanced Micro downgraded at Credit Suisse: Credit Suisse is downgrading AMD to Underperform from Neutral. Cites continued pressures in CPU sector and competition. Lowers target price to $13 from $21.
The AMD results were so amazing that the stock only dropped 7% today while Intel was up 1.5%. Also, Barcelona will not be launching at speeds past 2.2GHz so the 40% advantage will be negated. Expect pricing starting at $499 for the 1.8GHz part and full production ramp will not occur until late Q3. The current Brisbane 65nm chips will not make it past 3.2GHz if they even launch it at that speed, 3.0GHz should paper launch shortly.
AMD has made some bad decisions in the last few months when it comes to desktops. First they stopped doing 939 and like mentioned before a lot of single core 939 users would have gone to double core at these prices, but they cannot. Secondly the 4x4 system that did nothing for their image and is hardly selling like hot cakes, thirdly they release a 65nm product that is slightly slower and does not overclock as well. The 65nm product makes them more money but it does not encourage people to sing their praises. In fact all the praises are now being sung for Intel and it looks like staying that way for another 6 months.
Rather than talking about AMD against a monopoly I think it is better to start looking at home and your own product line. Intel did that with P4 and it did them Ok when they decided to change.
I heard without charges that the loss was .02 per share. Much better than the 1.5 bn, billion with a b kids, that Intel made.
I have no doubt that Barcelona will be a nice chip. Then Intel can jack up the GHZ on Conroe to 3.5 or so
Hector Ruiz should be arrested and sent to Gitmo for lying to investors a month ago.
Actually AMD reported a profit drop due to price competition.
AMD also reported a market share increase.
Even the great conroe has been unable to stop intels market share slide.
Intel is laying off another 11000 employees and closing more facilities.
Intel 45nm is going to be released this year. Is their manufacturing that unpredictable?
Maybe you could post the day and the month it’s going to be released.
Intel has stated Q3/Q4'07 that is still 6 months away. AMD on the other hand continues to say mid-2007 which could mean as early as Apr (if you consider mid-2007 to mean Q2-Q3'07). As wafers need to start at least 3-4 months prior to completion of wafer processing and packaging, if Apr is a real date those wafers should have started already. Thus they should be able to refine a 6 month window which starts in 2 months?!?! One wold think they would know if these wafers have started and with their AMAZING APM3.0 technology they should be able to calculate WIP turns and figure out when the wafers will be out of the fab.
"And as proof, that DELL did NOT get the amazing discounts that you all believed: AMD's desktop shipments and mobile shipments were WAY up (a new record), but yet the margins in these areas were barely untouched (slightly lower). The only segments where the margins were hurt big-time were servers (and understandably"
Funny here I was thinking the first part of the Dell-AMD deal was limited to servers (you know prior to the Sept-ish release of Mobile/desktop).....I'm pretty sure I read that somewhere.
You also do realize as AMD has been transitioning to 300mm (we'll ignore 65nm as those didn;' really ship in Q4). 300mm should yield ~30% cost reduction over 200mm (yeah I know everyone will say the area scales 2.25X, but equipment price goes up, chemical consumption. consumables go up, more expensive (OHV)automation system for 300mm, etc..)
So if 300mm is coming on line to a greater % of overall products and margins are flat, that likely means pricing is falling as fast as the 300mm cost savings or 300mm yield is crap (unlikely).
I would speculate that desktop and mobile prices are dropping (especially with Dell) as AMD should be seeing a margin increase merely due to production with 300mm.
I know this is a lot to comprehend for most of you wishful thinkers - but flat margins coupled with lower production cost (due to 300mm transition) means lower prices or someone is embezzling lots of money...
" expect Penrym to be the talk of the industry on 45nm no less...."
Even as an Intel fan I think it is irrational for people to think this chip to be significantly better. You will see significant power reductions (which could in theory allow somewhat higher clock rates), but I think what most people fail to realize is generally speaking you get ~30-40% TRANSISTOR performance increase but this rarely translates into a direct 1 for 1 clockspeed (fmax) increase as there are things other than sheer transistor switching speed which can limit clock frequency.
I would expect 10%, maybe 20% clock gain coupled with some power reductions - helpful but by no means will this be a breakthrough in performance compared to 65nm.
Keep in mond Penryn is not much different architecturally over it's 65nm brethren.
The Intel fans stating this are in my mind no different than the irrational Sharikou who expected a 30-40% performance gain on K8 simply by moving to 65nm.
"Intel has $9 BILLION in debt..."
While this sounds like a huge # let's try to put this in perspective.
9Bil is ~1 quarter of revenue.
Intel spends ~750Mil-$1Bil each quarter on dividend distribution to stockholders and stock repurchases.
You need to look at some normalize numbers like debt: equity ratio, etc to get an understanding of what the absolute debt really means.
If a millionaire owes $20,000 on a car loan is that susbtantial? What about that same debt if the person takes in $40,000 per year?
"Margins are considerably low, but better than I ever expected considering Intel's shear size"
I love this - you were expecting a relative 25-33% drop in gross margin?
If you have such clarity what is your expectation on Q1 gross margin?
Can I assume you also assumed that intel's quarter on quarter gross margin would be flat even during a price war?
"Funny how Intel still sell/manufacture Celeron, P4 and PD when Core 2 is already OUT.
Mega Kill"
What's even funnier is that with this still being 75% of Intel's production Intel still manages to make more profit and have higher overall gross margin with this heavy a mix of n-1 technology parts!
All the wall street action doesn't change the fact that AMD64 K series is still technically superior to any INTEL in terms of scalability and data thru-put K8l will only re-enforce this advantage.
There is no K8L on the market.
“K8L won't ship in volume until mid-2008. By that time, AMD will have lost 50% of their server share.”
Pardon me buddy; what happened to the speed crown and all that goes with it. Have you so quickly forgotten? Intel fanboy arguments will BK in a 2Q 07.
Anonymous said...
"...if you get the ATI adquisition out, AMD still earned money.
gotta love oportunists!"
I understand that, but what he stated/predicted was wrong.
His article was claiming that AMD could do better in a price war by being able to ride the profits of Server, 4x4 and Mobile processors.
Since the Server ASP is down substantially, 4x4 was a flop (until K8L), Mobile was the onlything that kept AMD alive considering Desktop was flat, it would be fair to say he was wrong.
The whole point of my post, Intel took money from AMD in the area Sharikou claimed they couldn't... Servers.
If Intel wants to turn to 45nm process on most of their fabs, they will need BILLIONS in upgrade costs... remember, that's why AMD STAYED at 90nm... postponing the Billions in upgrade costs. But 45nm might be their only resource to combat their powerhungry processors...
This will be an interesting year.
But they don't need to convert MOST of the fabs ... only the ones which will produce the maximum performance oriented parts - the ones that hurt AMD most in the most demanding segments ...
I think Sharikou should put his money where his mouth is. If Intel really does BK 2Q08, I (and I'm sure many readers) will admit that there's truth in what he writes and give him all the credit for predicting Intel's demise. However, if Intel does *not* BK 2Q08, I think Sharikou should make no excuses, admit he was wrong, and post on this forum an entry to that effect.
Even the motherboard manufacturers are now reporting my previous Q3 launch statement along with low demand (channel/reseller, not Dell) until then with the AM2 series. The next couple of quarters will not be good for AMD, maybe the R600 will at least pull the graphics division back up to break even or better. ..
You can tell it's not good when AMD starts crying "MONOPOLY" again.
It must be convenient to have such an excuse everytime you screw things up.
AMD's new sales pitch from the conference call: "Best value in the industry" - Rivet. ROFL! Sounds pre-Athlon era to me.
They ARE up against a monopoly. A "do whatever" monopoly at that.
AMD just has balls of steel. They decided to take the ATi hit and the lower pirces hit in the same quarter.
That's actually GOOD planning. Now in Q1 they can actually stop the bleeding since by the end of Q1 they will hve all of the K8 line switched over to 65nm. OF course they will stil make 90nm but I can see Chartered going 65nm real fast.
That way they can use Fab30 for Opteron and Turion and Fab 36/Fab 7 for Brisbane and Barcelona.
Yes, the smell that is coming from the big green K-chip company is... dead meat.
K8L won't ship in volume until mid-2008. By that time, AMD will have lost 50% of their server share.
Opteron/DDR2 is a dog that should never have been made. Basically a complete waste of all the momentum AMD had built in the market.
Same goes with AM2 DDR2 Dog Meat.
Nothing can change the fact that AMD is run by a bunch of retards that get their orders from Bill Gates.
There is no innovation at AMD, just paper launches and long academic discourses from Sleepy Hector.
Yawn....
You guys make me sick. Isn't there enough war without you guys brining one to my PC?
AMD will still own TPC-H so they will still get people to buy Opteron.
Server companies don't just switch out infrastructure, so Intel needs growth to get server business.
8 Opteron cores beat 8 Core 2 cores in TPC-H.
I guess if Barcelona is as fast as they say you'll be eating your words along with CoolnQuiet, IMC, HTX, AMD64, etc.
How is that not innovation?
Sharikou you really need to cut off these anonymous stalkers (I mean posters).
Speaking of 40%, how's that 40% market share runrate by end Q4'06 prediction working out for you?
23% * 119% is ~27-28% overall and that's best case as it is not accounting for overall industry growth in Q4...
Please enlighten us with your latest runrate calculations for Q4
It's year 2007 already, Dr.
Will u shut down this blog
next year if Intel doesn't go BK??
AMD have come through the worst. Now Intel have plenty to worry about-Clowntown won't cut it and all Intel have left are an assortment of processor names-not REAL products.
Lest I forget, AMD have SEVERAL BILLION $ to collect from Intel!
Since the Server ASP is down substantially, 4x4 was a flop (until K8L)
I have a hunch that 4x4 was just a "public beta test" with already existing but not polished chipset and quickly repackaged k8 cores. AMD surely knew the performance wasn't marketable enough and they just did it instead of a "paper lounch", maybe they wanted the feedback from the public.
There is always a possibility that they are working in-house on a better performing chipset to work with yet-to be released k8l units. If my prediction is correct and 4x4 rev.2 will actually rock (which would be quite possible if Barcelona performs as suggested), media and reviewers will do quite a bit of free advertising when comparing the "new 4x4" to the "power hungry old behemoth".
Maybe AMD just sacrificed a dead-end project (a project without any real future in it's present state) ...
Then again it could just be a flop ...
I guess if Barcelona is as fast as they say you'll be eating your words along with CoolnQuiet, IMC, HTX, AMD64, etc.
Wow! CoolnQuiet! Awesome! Too bad Intel did it with SpeedStep first. IMC? Intel did that first too with Timna - even if it didn't ship. The Alpha processor was the first shipping CPU with an IMC. 64bit? The MIPS processor was 64bit back in 1991!
Next please!
What will Barcelona's strongest feature be? It's performance? It's energy efficiency? Maybe, but IMHO, it will be it's drop-in compatibility. Not only will they get new customers, but just about everyone who currently owns existing platforms will more than likely upgrade... HUGE DEMAND!!!
AMD's V.P said:
because AMD has made it easy for customers to drop the quad-core solution into their existing equipment, customer acceptance will be rapid and broad-based.
If Intel wants to turn to 45nm process on most of their fabs, they will need BILLIONS in upgrade costs... remember, that's why AMD STAYED at 90nm... postponing the Billions in upgrade costs.
Finally, people are starting to understand the idea!
I cannot drop in a fresh, faster 5000+ or better CPU, my only chance is switching to a C2D system and as me many "939" people are going to switch.
Do a bit of research and you'll quickly learn that Intel's upgrade path is horrendous in comparison to AMD's. AMD loyalists have saved MILLIONS in the last decade, while Intel loyalists have spent MILLIONS on such upgrade paths.
Quite the string of downgrades. OUCH!!!
Once you understand that stocks have cycles, you can profit from it. In this case downgrades are a way for them to pick up massive amounts of shares and make the common guy lose A LOT of money. By mid 2006 (mark my word), there will be only UPGRADES for AMD which all these companies will slowly start liquidating (and the common guy will buy)... and the cycle will start all over again!
I love this - you were expecting a relative 25-33% drop in gross margin?
No, I was expecting worst! I was expecting a loss!
If you have such clarity what is your expectation on Q1 gross margin?
Couldn't really care! I know AMD, just like I know Intel. I invest in AMD for the company it will be in 5 to 10 years... NOT next quarter!
jeach said...
"...Not only will they get new customers, but just about everyone who currently owns existing platforms will more than likely upgrade... HUGE DEMAND!!!"
You make a great point, but there are some variable's that you may be overlooking.
No matter how good AMD's K8L ends up being, everyone in the position to buy a new system or upgrade an existing system will be looking to Intel to see if they have an answer.
2P and down Intel is extremely competitive, price and performance and have repositioned them selves in this space.
Another thing to consider is that, like 2006, Intel might end up showing Tigerton results, or 45nm processor specs and a future roadmap, and like the success they had in 2006 this could slow the adoption of AMD's K8L.
I am not saying that K8L will not be good, but I had said in the past, Intel has serious manufacturing muscle, and if they want to hurt AMD, they can do what they did in 2006 and lower prices with a competitive product.
Hypothetical... If you could not drop in quad cores and had to build a 2P system, would you go AMD if it gave you a 20% advantage but cost 30% more while the power was about even?
Anonymous idiot said...
It's year 2007 already, Dr.
Will u shut down this blog
next year if Intel doesn't go BK??
Why would the "Journal of Pervasive 64 bit Computing" shut down if Intel doesn't BK? This site is an unbiased look at the 64-bit computing industry.
"Wow! CoolnQuiet! Awesome! Too bad Intel did it with SpeedStep first. IMC? Intel did that first too with Timna - even if it didn't ship. The Alpha processor was the first shipping CPU with an IMC. 64bit? The MIPS processor was 64bit back in 1991!
Next please!"
A 64-bit, fully x86-compatible, power-efficient chip with efficient memory architecture (IMC) affordable and accessible by every PC owner.
Think about how Intel, the once dominating Kingship of x86 processors, with its 15x larger size could have missed this important milestone and let AMD took it. I know if you still don't respect AMD for what it did, you're definitely a *stupid* Intel fanboy.
Anonymous said...
Bottom line K8L should be faster then Clovertown, the problem its 7 months late and in 9 months expect Penrym to be the talk of the industry on 45nm no less.
You have failed to take into account the bottleneck of the ancient Intel platform. AMD's 4x4 has per core bandwidth, something that Intel's platform cannot provide. K8L will leave Penrym in tears.
"I have a hunch that 4x4 was just a "public beta test" with already existing but not polished chipset and quickly repackaged k8 cores."
The 4x4 was actually made for a very specific group of users. We know this when AMD claims it's only available on Socket F - no ordinary/budget concern consumer will consider it.
Hey Sharikou, if you allow me to go off topic for a moment with older subjects of yours...
Not only Sun is on a trademark power-trip these days (taken from a linux distrib site):
As most of you will know, Firefox & Thunderbird have been renamed in Debian due to trademark problems with Mozilla.org. [...] Icedove [...] is the new name for Thunderbird. Firefox will follow soon, being named Iceweasel.
See what I mean, the Linux community auto-destroys itself.
Think about how Intel, the once dominating Kingship of x86 processors,
LOL, past, present and future is all intel. INTEL has always hovered around 70-80% MS. ASPs are down but the profits continue to roll in at the billion+ / quarter. Feeding continued design, manufacturing and infrastructure leadership. Kudo's for AMD for surviving but they are irrelevant... they could die and to a larger order things don't change.
Trust me if INTEL was to go Hector and his band of castoff couldn't supply the world. Can you imagine the world dependent on IBM manufacturing prowess, AMD execution..
You'd all be wishing for a INTEL.. but no need to wish its all INTEL for the past 15 years and it'll still be all INTEL for the next 10 too.
No matter how good AMD's K8L ends up being, everyone in the position to buy a new system or upgrade an existing system will be looking to Intel to see if they have an answer.
In regards to 'buy a new system' I guess that could be debated. But in regards to upgrade, I really doubt they would go with Intel if it meant they have to buy a brand new platform.
Considering that you could 'theoretically' double the processing power of:
1) a 2P for $2000 to $3000, and
2) a 4P for $4000 to $6000.
All the while keeping the overall energy consumption at relatively the same level.
Remember that for every $1 invested, you will spend $0.50 in energy costs.
Now I'm not sure about the following, could someone answer:
1) Will Intel's 45nm quad core be native? I would assume!
2) Will it support per-core memory controllers? If not, will a 4P system have 16 separate memory controllers on each node?
3) AMD should be using small, shared level 3 'compressed' cache, which means the die footprint will be small. Also they will have an HT with a theoretical throughput of 5200MB/s. Since Intel will NOT have HT equivalent nor compressed cache (huge amounts resulting in a HUGE die footprint), will Intel's processors actually be capable of being energy efficient? My assumption here is that cache dissipates a lot of heat. The greater the area, the more heat, no?
Penryn was shown today running several applications under 64-bit Vista significantly faster than some 8-way Opteron servers. For those talking about Intel needing to spend billions of dollars on new 45nm fabs, they already have with two coming on line shortly and the third early next year. Barcelona is not coming until Q3 and it might be late Q3 before any hit retail. It will not be a significant amount of AMD's production until the second half of 2008, around the same time when AMD will try to launch 45nm production, close to a full year behind Intel, again.
You have failed to take into account the bottleneck of the ancient Intel platform. AMD's 4x4 has per core bandwidth, something that Intel's platform cannot provide. K8L will leave Penrym in tears.
Ooh, yeah, I forgot about that. I mean, look how that "per core bandwidth" really helps K8 and 4x4, I mean they mop up Core2Duo.
Oh, wait. No, they don't. Per core bandwidth means little in the real world and only helps on a few specialized tasks.
Damn, you've been foiled again!
jeach!
"1) Will Intel's 45nm quad core be native?"
Yes
jeach!
"2) Will it support per-core memory controllers?"
No
jeach!
"Since Intel will NOT have HT equivalent nor compressed cache (huge amounts resulting in a HUGE die footprint), will Intel's processors actually be capable of being energy efficient?"
Yes
jeach!
"My assumption here is that cache dissipates a lot of heat."
Wrong. Cache is the coolest part of a chip.
Quote: "Barcelona is not coming until Q3 and it might be late Q3 before any hit retail. It will not be a significant amount of AMD's production until the second half of 2008, around the same time when AMD will try to launch 45nm production, close to a full year behind Intel, again."
You seem to know MORE about AMD's production schedule than AMD themselves! Take a hike you stupid Intel Fanboy!
Some idiot wrote: "You have failed to take into account the bottleneck of the ancient Intel platform"
And how many times does has this extra bandwidth help AMD's superduperbandiwidthlon perform better than Core2Duo? ZERO!
How many times does AMD have to pay extra cost for this "special" but under utilized feature? 100% of the time!
jeach said...
"But in regards to upgrade, I really doubt they would go with Intel if it meant they have to buy a brand new platform."
Good point.
Just to throw another factor in, with Google and Sun beginning to use Intel again that would seem to say something about Intel's future platforms, and after reading the news today it would seem that Hyper Threading will be re-introduced in the Core Architecture.
Could be good or it could be bad, well have to wait and see.
PENIX said...
"Why would the "Journal of Pervasive 64 bit Computing" shut down if Intel doesn't BK? This site is an unbiased look at the 64-bit computing industry."
Are you that f..king stupid? Please poing me in the direction of ANYTHING on this blog that ISN'T biased?
I laughed out loud when I read this. Congratulations on proving yourself as retarded as the douche who started this blog.
roborat said...
And how many times does has this extra bandwidth help AMD's superduperbandiwidthlon perform better than Core2Duo? ZERO!
How many times does AMD have to pay extra cost for this "special" but under utilized feature? 100% of the time!
The same claims were made about AGP back when it first came out. They said the extra bandwidth would not be utilized and AGP was not needed. Now look where we are. AGP wasn't fast enough. Intel has failed to eliminate this bottleneck from their platform and because of it, AMD will crush them.
sharikou, earlier in this thread I noted that AMD is forecasting a Q1 GAAP loss, primarily due to another $120M ATI related charge, $30M of which is likely to be inventory writed down.
This is irrelevant. As long as AMD can ramp FAB36 on 65nm and gets K8L out in time, it can make all the money it can. As AMD takes more market share, Intel can't drop prices any more, or Intel will BK even faster.
"3) AMD should be using small, shared level 3 'compressed' cache, which means the die footprint will be small. Also they will have an HT with a theoretical throughput of 5200MB/s. Since Intel will NOT have HT equivalent nor compressed cache (huge amounts resulting in a HUGE die footprint), will Intel's processors actually be capable of being energy efficient? My assumption here is that cache dissipates a lot of heat. The greater the area, the more heat, no?"
Actually if you add up TOTAL L2/L3 cache between Intel and AMD the gap is no longer that large (8MB vs 6MB). This is probably why you no longer here Sharikou talking about how adding cache is a kludge or bandaid.
What is also interesting is the ZRAM talk has died down quite a bit (remember folks speculating whether L3 on K8l would be ZRAM?). Also if you follow the industry closely you will find some Intel results on alt SRAM approaches which also yield 4-5X area scaling (1C-1T cells and other non-conventional approaches when it comes to x86). Given the node scaling still seems to be o track I would expect cache sizes o continue to significantly increase (especially if some of the non-6T approaches are used!)
In reference to gross margin:
"No, I was expecting worst! I was expecting a loss"
Explain what gross margin would be with a loss, please? (and don't say 0%!) I'll give you a hint go back on AMD's previous quarters where they had losses and go look at what the gross margins were.
"If Intel wants to turn to 45nm process on most of their fabs, they will need BILLIONS in upgrade costs... remember, that's why AMD STAYED at 90nm... postponing the Billions in upgrade costs.
Finally, people are starting to understand the idea!"
As you have the financial sense of a mouse, I'll help you out a bit. The normalized cost per cm2 of Si goes up about 10% per generation (assuming wafer size stays constant). Die area scaling is at least ~33% (AMD right now) worst case and up to ~45% (Intel) best case for a tech node transition (assuming same architecture and constant cache size).
So i guess what you are saying is a 10% cost increase doesn't justify the 33%-45% die size reduction? And that's ignoring the technical performance benefits (power, transistor speed).
AMD didn't "CHOOSE" to wait on 65nm (and if you don't believe me why are they making such a big deal out of trying to accelerate 45nm). They went to "65"nm when it was ready or do you think they wanted to make fewer die/wafer at higher power when they could have switched to 65nm?
In your "financial" analysis you may also want to consider on a per fab basis you are talking about ~750Mil-1Bil equipment/year. ~1 Bil is already a sunk cost (the building itself) and the other ~2Bil is spread out over ~2 years depending on fab ramp rate. If AMD could finance a 5Bil ATI acquisition they easily could have afforded to pull in 65nm 6 months earlier if it was ready (which it wasn't) for ~$500Mil - the only financial loss for doing this would have been the NPV of pulling that $500Mil in 6 months early. Using a 15% return rate/year this would mean a net negative impact of <$50Mil.
On second thought you are clearly right - AMD didn't want to accelerate the 65nm ramp because they want to save ~50Mil by delaying the 65nm equipment purchase ~6 months.
Either that or perhaps they could sustain that ramp rate in the fab or 65nm wasn't ready any earlier than it was implemented (I believe it is both).
"And how many times does has this extra bandwidth help AMD's superduperbandiwidthlon perform better than Core2Duo? ZERO!"
maybe it doesn't help your toy application whose memory accesses are easily optimized, or whose data could happily sit altogether within a 2MB L2 cache.
maybe it doesn't help your toy application whose memory accesses are easily optimized, or whose data could happily sit altogether within a 2MB L2 cache.
Ooh, a sharikouism. You basically used the "Intel only good for super-low end" idiocy.
You call them toy applications, everybody else calls them "95% of the applications anybody runs".
"The same claims were made about AGP back when it first came out. They said the extra bandwidth would not be utilized and AGP was not needed. Now look where we are. AGP wasn't fast enough. Intel has failed to eliminate this bottleneck from their platform and because of it, AMD will crush them."
It's logic like this that makes you wonder...it's a question of TIMING! Was AMD crushed when they switched to DDR2 later than Intel? Of course not because AT THE TIME it wasn't a limiter, but they knew it would eventually be a problem and moved to it when the time was right from a market perspective.
The same thing goes for FSB vs IMC. Is it the most significant factor limiting CPU performance right now (other than 4P+ systems)? No it is not despite the yahoos who keep saying well what about applications that require >2MB... bottom line is everyday performance for THE MAINSTREAM computer user is not SIGNIFICANTLY hurt by use of FSB.
Will it be a limiter in the future? Probably. Will Intel swicth to an IMC? Yes, and they will likely try to time it to when it is really needed (I assume it will be after DDR2 transitions to DDR3 and you will see it on server products first and then potentially waterfalling down to DT/Mobile).
I still have not seen actual data across a BROAD set of applications that represent majority usage of computers that shows FSB takes a huge penalty hit over IMC (again setting aside 4P- where it clearly is a limiter). I'm sure Edward can pick out some obscure ones or server based ones but is that really representative of the overall market?
edward
"maybe it doesn't help your toy application whose memory accesses are easily optimized, or whose data could happily sit altogether within a 2MB L2 cache."
You do know that C2D has far superior memory access prediction than anything else on the market, do you?
So there will be another 120M hit on earnings in Q1, greater than computational products earnings in Q4, and based on the outlook statements, much greater than earnings in Q1. So, AMD themselves are predicting a GAAP loss again in Q1 '07. Not a good outlook. Those balls of steel are starting to rust- all the bleeding and whatnot.
But they also have a better cost structure because of 65nm so that Fab 36 can churn out up to 3.4X the amount of chips as the same amount of wafer starts as Fab30.
By concentrating on high-volume parts (3600+ and 4000+ Brisbanes are selling at Newegg RIGHT NOW), AMD can further save money since those chips will be very low-defect because of clockspeed (1.9,2.1GHz).
These will be perfect for the new HTPC designs they released during CES.
They will also have the influx of revenue from the R600 near the end of the quarter( not to mention all the other ATi property).
Looking at Intel though their ASP is supposedly up that only means that server is selling with growth rather than desktop (Core 2 prices are too low).
Their operating income, EPS, and profits are all down more than 30% YoY.
AMDs TOTAL hit was the $500M that came due with the acquisition.
If growth continues (Q4 saw 19%) they will offset a lot of charges with volume sales.
IF they can get Fab36 AND Chartered ramped to 100% 65nm by Q2, that will allow Kuma to be released in volume in Q3 at a better price than Brisbane.
If, in the same time, they can ramp just 1/3 of Fab 30 to 300mm wafers then that will mean twice the volume and will allow 1/3 to be taken totally off line and fitted for the first 45nm chips by the end of Q3.
"AMD can further save money since those chips will be very low-defect because of clockspeed (1.9,2.1GHz)."
I don't think the "defect" density has any to do with clockrate. Some defects prevent the circuit from even running; others make them run at extremely low speed (much less than 2GHz).
The yield of low-speed chips will be good, however, and it's clear that AMD's main priority is to get things out of the door (due to high demand). I guess they just don't want to spend too much time testing and binning. AMDZone was lucky enough to get a Brisbane 3600+ that OC's to 3.1GHz.
"I guess they just don't want to spend too much time testing and binning."
WTF? They still need to test and bin the chips...do you have any idea how this process works? (apparently not)
"They will also have the influx of revenue from the R600 near the end of the quarter( not to mention all the other ATi property)."
Influx? Perhaps 1-2 weeks of sales? How much do you really think this will impact overall quarterly earnings? "all of the other ATI property" was either at a lsight operating loss or slight operating gain (can't remember) so I wouldn't expect this to have much impact.
"If growth continues (Q4 saw 19%) they will offset a lot of charges with volume sales."
AMD themselves forecasted flat to slightly down for Q1, unless you think Q4-Q1 will yield another 19% growth? (or a 19% ASP decline to offset the flat to slightly down revenues AMD provided quidance on?)
"and fitted for the first 45nm chips by the end of Q3."
45nm? WTF? You talking Q3'07 or Q3'08? If you meant development chips my bad...but if you are talking production chips, Q3 is a bit early for HVM equipment installs.
"Looking at Intel though their ASP is supposedly up that only means that server is selling with growth rather than desktop (Core 2 prices are too low)"
The vast majority of Intel Core 2 production right now IS SERVER! I would much rather grow server share than desktop share...why is this an issue?
"IF they can get Fab36 AND Chartered ramped to 100% 65nm by Q2"
This will not happen according to AMD's statements - they expect most of the fab to be converted by end of Q2 but I don't think it will be 100%. Also Chartered volume is, what, 1-2K WSPM - that is negligible and also worse margin then making chips in house.
Finally what most people are ignoring is AMD's capital depreciation guidance is also up for 2007 (obviously due to the new equipment being installed). This will likely be offset by savings migrating greater mix to 65nm but it will eat into those savings.
Another shining example of people drinking the wrong Kool-aid...
thekhalif said:
But they also have a better cost structure because of 65nm so that Fab 36 can churn out up to 3.4X the amount of chips as the same amount of wafer starts as Fab30.
You're going to need to explain who in the hell going to 65nm is going to allow Fab36 to increase output by 3.4X. Did they double the cleanroom space over night or something? I think you're confused with the the transition from 200nm to 300nm which garners about 2.3X.
By concentrating on high-volume parts (3600+ and 4000+ Brisbanes are selling at Newegg RIGHT NOW), AMD can further save money since those chips will be very low-defect because of clockspeed (1.9,2.1GHz).
Clockspeed has NOTHING to due with defect density. It wouldn't make a difference if AMD had a 999GHz part. Defects are based on the cleaniness of the process and/or tool related issues during the manufacturing of semiconductors.
IF they can get Fab36 AND Chartered ramped to 100% 65nm by Q2, that will allow Kuma to be released in volume in Q3 at a better price than Brisbane.
If, in the same time, they can ramp just 1/3 of Fab 30 to 300mm wafers then that will mean twice the volume and will allow 1/3 to be taken totally off line and fitted for the first 45nm chips by the end of Q3.
I think you have AMD confused with the Willy Wonkea Chocolate factory or something else. AMD isn't going to see the kind of volume you're predicting in thirds from Q2 to Q3. Retooling a factory takes a few quarters, let alone requalifying part of the line for production. Also... AMD's own roadmap says they won't have 45nm until mid-2007.
""I guess they just don't want to spend too much time testing and binning."
WTF? They still need to test and bin the chips...do you have any idea how this process works? (apparently not)"
WTF? An X2 3600+ OC'd to 3.1GHz, surely they could've binned those chips better had they spent more time... do you have any logic - let alone idea how this process works? (apparently none)
""maybe it doesn't help your toy application whose memory accesses are easily optimized, or whose data could happily sit altogether within a 2MB L2 cache."
You do know that C2D has far superior memory access prediction than anything else on the market, do you?"
What's "far" superior? It makes 75% programs run 15% faster. IMO, it's better, but by no means deserve the "far" descriptive.
"WTF? An X2 3600+ OC'd to 3.1GHz, surely they could've binned those chips better had they spent more time... do you have any logic - let alone idea how this process works? (apparently none)"
(this was in response to: WTF? They still need to test and bin the chips...do you have any idea how this process works? (apparently not)")
So just so I understand this overclock you talk about to 3.1 was at normal voltages and was within the AMD TDP they stated? How does an OC mean they could bin parts higher?
So just so I understand AMD is either lower binning these parts intentionally or doesn't know they can be binned higher? This makes perfect sense as why sell a 4200 or 4600 when you can sell the same chip as a 3800 at lower margin?
I heard there was a P4 operating at >8Ghz, maybe that should've been binned higher? I'm neither an Intel or AMD fan - the people on this board need to get a clue on how manufacturing works before they comment on it.
Oh and by the way, you do realize when a part is binned at say 2.4 Ghz, it actually is capable of slightly more than that and the manufacturer factors in transistor performance degradation over the quoted lifetime of the part.
The only time I can see a manufacturer binning at lower speeds than capable is when there is a shortage, or if there is no competitive processor they will keep the top bin lower than theoretically possible (Conroe XE) or cherry picking for a lower power envelop (example AMD EE parts)
Here's one idiot's view of Sort/Test (I won't mention Ed by name...uh, I mean, Ricky, yeah that's, it Ricky)
Jim: Hey, Joe - plug that chip into our 2.2Ghz tester...does it work?
Joe: Yeah, I guess we should try it on the 2.4 GHz tester now, no?
Jim: Why bother, it works at 2.2GHz, let's just stamp it and get the next one.
Joe: You know what would be cool?
Jim: What?
Joe: If we had a tester that could determine actual speed and then after testing, we could mark it at the highest bin.
Jim: Come on man, next thing you know, you'll be telling me the technology exists to put a man on the moon.
Joe: You do know that was staged in a TV studio, right?
Well - some brain strain (I seem to recognize a pattern):
somebody figured out that
- low asking price combined
- with overclocking potential
- on stock voltage and air
generated quite a lot of FREE advertising (word-of-mouth, post-in-a-forum & review-sites) and SALES.
Why not utilize a --extremely very similar-- marketing tactic which worked well for the larger vendor - on a TEST batch that sells for peanuts and comes with "no warranty" for the highest clock achievable. IBM says it's Power6 will run at 5GHz on 65nm ... so why wouldn't a petite Brisbane go to 3.something?
Think different, remember ...
(no innovation required in this case)
(yaawhaa is the visual verification for this post, it must mean something)
"So just so I understand AMD is either lower binning these parts intentionally or doesn't know they can be binned higher? This makes perfect sense as why sell a 4200 or 4600 when you can sell the same chip as a 3800 at lower margin?"
Compared to that much you've said and that little you've actually spoken, I'll only say two sentences:
There are multiple degrees of binning, which affects the time from design to final product, a factor that was most important to AMD's first batch of Brisbane.
And to the other idiot anonymous who made up the conversation: you displayed your ignorance thoroughly.
Post a Comment
<< Home