I thought I'd sum up what I thought were the key points Otellini made on day 1 of IDF. To watch the webcast, go to:
http://www.intel.com/idf/us/fall2006/webcast.htm
1. 5 million Core 2 shipped - this icludes desktop and notebook parts. This should put an end to speculation on yields. Specially the discussion over on Rahul Sood's blog where I commented on Merom notebook availability at launch being higher than usual while desktop being held back due to chipset availability issues:
http://voodoopc.blogspot.com/2006/09/be-on-lookout-for-pink-bunny-at-idf.html
It should also put an end to Sharikou's conjecture that Intel can't sell 400k Conroes:
http://sharikou.blogspot.com/2006/09/intel-needs-to-lay-off-80-of-its.html
2. Quad Core launch - Otellini announced the launch of their quad cores in two phases. November for servers and high end desktops. Q107 it becomes mainstream under the moniker Core 2 Quad. He claimed a 70% SPECINT increase for quad core over the Conroe Extreme version. This don't mean much until we see a series of robust benchmarks across. However, what this means is a 6 month lead over AMD...enough time to stabilise manufacturing and bring this into mainstream availability and pricing in Q1. As far as I'm concerned, in theory K8L should trounce 2 Core 2's integrated in the same package. However, this is not a done until we see K8L which is still months away. Even 4x4 is still to come. The demo from Remedy showing their new game Alan Wake was cool - http://www.alanwake.com/.
3. 45nm - 40 million 65nm processors shipped. But more importantly cross over from 90nm to 65nm completed. 45nm will give a 20% performance increase and a 5x reduction in power leakage. I'll deal with the market implication of this in a later post but let me tell you that it makes Penryn (the 45nm shrink derivative) due in 2H07 very interesting and potentially may give AMD a headache. Otellini showed 3 fabs in development to go online b/w 2H07 and 1H08 at an investment of 9 bln $s. First 45nm processor design to be complete in Q406 and 15 in total in development. 32nm comes in 2008 and he showed again the strategy to have a new m-architecture every 2 years. The biggie here is the 300% increase in performance per watt by 2010.
4. Woodcrest - 1 million Woodies shipped in 3 month. Woody is now 40% of the DP server market and over 50% of Intel's DP shipments. In addition, Rackable CEO gets up on stage and displays a mid sized rack that would put a 40 CPU Intel Quad core cluster in the top couple of hundred super computers and claims he has a bigger rack that would catapult into the top 80.
Sharikou has frequently complained that Woodcrest is nowhere to be seen. Now we know why - because customers are buying them before they hit the shelves -:) (do I need to put the cheeky monkey pic here too?)
5. Teraflop on a chip - Otellini displays a processor w/ 80 micro-cores transferring data to SRAM at over 1 teraflop. Claims technology will be available in 5 years. Speculation is this is the secret sauce for Gesher.
6. Wimax - Anand Chandrasekhar brought a UMPC w/ Wimax which was connnected to a new Volkswagen being developed. What was really cool was the UMPC wirelessly ran 2 video streams at the same time to 2 screens at the back of the car. Sweet...!
Overall - Otellini had some interesting things to show but the only thing close to a rabbit was the teraflop chip. Nett take away:
a) Intel's execution engine seems to be back on track somewhat. Just delivering what they committed is a good thing.
b) The Conroe issue for desktop is definitely not yields - as I said before, it's a chipset issue. 5 million Core 2's on desktop/notebook and 1 million Woodcrest is a good ramp. My prediction stands - Q3 will be better for AMD than Intel though not great for both. Momentum will swing back toward Intel in Q4 with them re-gaining market share. AMD will swing to a GAAP loss by Q207 once the ATI deal goes through.
c) Ultra mobile PC with Wimax has the potential to be a killer app if Intel can drive the system price down to 500$ as they are trying. Wimax is going to be big...and it will be big for Intel. AMD cannot catch them here.
d) Kentsfield will not be just a "we got there first" product as I've been thinking. By Q107 as yields stabilise this is going to kick some serious ass. With the 45nm Penryn derivative in 2H07 it may not be "Game. Set. Match. - AMD" as per current expectations. If K8L is priced in the high end and Kentsfield is lower priced the performance and thermals may not be that disastrous and may provide better value for money. Either ways, Intel has a 6+ month lead to market quad core to end users and position themselves as the leader.
Overall, solid progress though one was hoping to see more innovation. Teraflop on chip and Wimax are the two big things that have the ability to drive a big shift in the tech landscape. And the final thing that will change the tech landscape is Apple who were on stage too. These guys are on fire - HP & Dell look out. Read my post below for the full analysis.
Subscribe to:
Post Comments (Atom)
24 comments:
1. 5 million Core 2 shipped - this icludes desktop and notebook parts. This should put an end to speculation on yields.
You are aware that Intel makes 3.5 Million chips a week? If you add up how many weeks it took them to make these chips you'll realize that Intel will probably only make it to 25% by year's end.
However, what this means is a 6 month lead over AMD...enough time to stabilise manufacturing and bring this into mainstream availability and pricing in Q1.
Sorry, but no. Even Conroe won't be hitting that level by Q1 07. Kentsfield will be much less. To be honest, quad core doesn't hit mainstream until 45nm.
As far as I'm concerned, in theory K8L should trounce 2 Core 2's integrated in the same package.
K8L will trounce Kentsfield. However, the dual version of K8L also has to do pretty good against Conroe and K8L has to be good enough to not get overtaken in 2008.
it makes Penryn (the 45nm shrink derivative) due in 2H07 very interesting and potentially may give AMD a headache.
I doubt it. AMD will have 45nm about six months after Intel.
In addition, Rackable CEO gets up on stage and displays a mid sized rack that would put a 40 CPU Intel Quad core cluster in the top couple of hundred super computers and claims he has a bigger rack that would catapult into the top 80.
This isn't that impressive. This is a similar scale to the Octigabay system that Cray bought to offer a range below its Red Storm technology.
displays a processor w/ 80 micro-cores transferring data to SRAM at over 1 teraflop. Claims technology will be available in 5 years.
Five years? There will be an Opteron PetaFlop system in 2008. It would take 1,000 of these to equal the Opteron machine which might be a little denser. However, it still remains to be seen whether lightweight processors are the way to go. IBM used a similarly strategy with Blue Gene L.
Momentum will swing back toward Intel in Q4 with them re-gaining market share.
I don't see this happening. I'd say Intel's Q4 market share will be worse than what it was in 2005.
AMD will swing to a GAAP loss by Q207 once the ATI deal goes through.
I don't what you mean by a loss. Obviously this will add a couple Billion in debt but I don't think it will effect AMD's earning's in 2007.
You are aware that Intel makes 3.5 Million chips a week? If you add up how many weeks it took them to make these chips you'll realize that Intel will probably only make it to 25% by year's end.
Yes old boy - and 25% is exactly what Intel has said will be their Conroe mix in Q4 so what's the issue. They are building their business plan to this number. What this means is they are on track overall.
Sorry, but no. Even Conroe won't be hitting that level by Q1 07. Kentsfield will be much less. To be honest, quad core doesn't hit mainstream until 45nm.
Watch the webcast. Otellini announced Core 2 Quad at mainstream pricing in Q107. Tell me you know something he doesn't...actually, show me!
I doubt it. AMD will have 45nm about six months after Intel.
Prove it with a link please...
Five years? There will be an Opteron PetaFlop system in 2008. It would take 1,000 of these to equal the Opteron machine which might be a little denser. However, it still remains to be seen whether lightweight processors are the way to go. IBM used a similarly strategy with Blue Gene L.
I'll believe you when I see a demo. Heck, I'll even start to believe you if you can show me a roadmap published by AMD on this.
I don't see this happening. I'd say Intel's Q4 market share will be worse than what it was in 2005.
Your guess against mine. Let's wait and watch. However, Intel is already re-gaining market share against AMD in places like US retail. AMD get's Dell. But as I said, 2 years too late. Let's see how many AMD systems Dell manages to sell.
I don't what you mean by a loss. Obviously this will add a couple Billion in debt but I don't think it will effect AMD's earning's in 2007.
Read my earlier posts. AMD will announce a quarterly loss under Generally Accepted Accounting Principles by Q207. I do re-call you claiming with much bravado earlier that the debt of buying ATI will not affect AMD. Do I sense a slight change in your posture?
25% is exactly what Intel has said will be their Conroe mix in Q4 so what's the issue. They are building their business plan to this number. What this means is they are on track overall.
Well, on track is fine but this means that their product mix is still dominated by Prescott based processors.
Watch the webcast. Otellini announced Core 2 Quad at mainstream pricing in Q107.
I consider mainstream to be below $300. Is this what Kentsfield will be?
I doubt it. AMD will have 45nm about six months after Intel.
Prove it with a link please...
AMD plans to have 45nm in production mid year 2008. Unless Intel can move up a quarter this will be six months.
There will be an Opteron PetaFlop system in 2008.
I'll believe you when I see a demo. Heck, I'll even start to believe you if you can show me a roadmap published by AMD on this.
What roadmap are you talking about? This is a Cray supercomputer built with quad core Opterons. Cray already has the contract.
However, Intel is already re-gaining market share against AMD in places like US retail.
It did a little in Q2. However, it lost more in the Asian market.
AMD get's Dell. But as I said, 2 years too late.
This isn't true at all. Even in 2005 AMD didn't really have enough volume for Dell. Now is the right time.
Let's see how many AMD systems Dell manages to sell.
I would imagine this will be 10-20 Million systems by end of 2007.
Read my earlier posts. AMD will announce a quarterly loss under Generally Accepted Accounting Principles by Q207. I do re-call you claiming with much bravado earlier that the debt of buying ATI will not affect AMD. Do I sense a slight change in your posture?
No. I never go by profit. I go by gross revenue and margin. For example, Intel makes much more money than they report because they spend most of it on stock buyback. We need to count that. Also, we need to count capitol investments and R&D because these are things that make the company more money in the future.
I'm saying the same thing I've said before. AMD's 3rd and 4th quarter revenues will be up.
Well, on track is fine but this means that their product mix is still dominated by Prescott based processors.
Of course it is - on desktops and w/ Tulsa. So what - they're still going to sell a ton of Netburst parts. This is part of their plans. They will cross over on their desktop line in Q207. But note - their notebooks will cross over to Core 2 in Q406.
I consider mainstream to be below $300. Is this what Kentsfield will be?
Maintream refers to the ability to drive an end system price point for the average consumer. This system price point is in the 700-1000$ range depending on where you are on the planet. To do that, in addition to considering CPU price - you have to figure out the rest of the bill of materials.
Five years? There will be an Opteron PetaFlop system in 2008. It would take 1,000 of these to equal the Opteron machine which might be a little denser. However, it still remains to be seen whether lightweight processors are the way to go. IBM used a similarly strategy with Blue Gene L.
Dude - you're talking about one server sitting in a lab somewhere. Intel is displaying this technology to bring it commercially to the mass market. I'm reporting a CPU technology development, you're hanging onto a sliver of a thread around a whole server which is not just the CPU and comparing it to just the CPU. Apples to apples old fella....apples to apples please.
I don't what you mean by a loss. Obviously this will add a couple Billion in debt but I don't think it will effect AMD's earning's in 2007.
Read my earlier posts. AMD will have a relatively better Q3 than Intel when compared to what they told Wall St. But both will be bleeding. Come Q4 - AMD will have a worse quarter and momentum in terms of market share and margins will start to move up for Intel as the impact of OEMs bringing Core 2 desktops WORLDWIDE comes into play along with the natural time it takes to win deals in the server market for Xeon 5000's.
Please! Please! I need to see some benches on a K8L engineering sample. I saw early benches on Conroe/Merom and they were outstanding. But I have yet to see 1 bench.
With Intel's scalability (AMD has never had a mainstream product over 2.6 GHz or so) where they start at low clocks and just move up the ladder as they refine the process, even a long-in-the-tooth mainstream 3.8 Ghz Conroe can still be competitive in 12-18 months.
K8L is currently vaporware...
You don't seem to understand the definiton of vaporware. When Anand previewed Conroe earlier this year it was six months from release. Vaporware would include both Tejas and Whitefield which were talked about but then canceled.
You don't get to have two different definitions for Intel and AMD. AMD has stated that they will demo K8L in the last quarter. If you consider that vaporware then you have to consider everything that Intel has not yet released vaporware including Kentsfield. You would also have to consider almost everything Intel talked about during IDF as vapor initiatives. For example, their version of Torrenza won't even be defined for another 12-18 whereas you can buy Torrenza products today.
Scientia - I do not disagree with you. Declaring an intent to build or do something is just hot air. Until we see the technology and it does what it's manufacturer claimed it would...it's vapor for me. That holds good for Intel and AMD. If it's not real...or it's not going to result in a commercially viable product then it's just positioning and posturing as far as I'm concerned.
No double standards on that for me.
scientia you are just another delusional amd fan boi who cant come to grips with his favorite company losing the performance crown. Rest easy fan boy its going to be a long year before we hear anything good coming out of amd.
I like how you say AMD will be at 45 nm six months after intel....yep i see how easily amd shifted to 65nm by that track record they will have no problem at all. Whatever helps you sleep better at nite buddy.
K8L is vaporware if i dont see a benchmark or even an engineering sample or demo anything tangible actually....its vaporware we mite as well be comparing it to nehlam or gelsher for all i care and no that does not hold true for kentfield since they have been benchmarked all over just look past your amd zone website for the info you will see it all over.
All AMD has rite now is empty promises and people buying X2 in the hopes that K8L will be as good as they say it will be. I like how every AMD fanboi say that k8l will trounce kentfield without a single piece of evidence to back up their claim but when conroe was benchmarked they all cried foul.
BINGO!!!!!
I don't know what you see, but I see K8L quad at MS high tech summit. Maybe the vapor your seeing is that escaping your bong.
Got a link...I don't have time right now to look so if they've demoed something give me the link and I'll pop it up.
Is this what you're talking about:
http://www.theinquirer.net/default.aspx?article=34742
That's a good sign. Now what I'd love to see is this demoed and some benchmarks. But I guess we'll have to wait. Let me repeat myself once again:
K8L in theory should slam Conroe. But with the claimed performance increase from Intel on Penryn things might get interesting from a price/performance perspective. We won't know until we see K8L benchmarks vs Conroe and we won't know about Penryn till next year. So keep your hat on and wait for the ride.
K8L has to slam Conroe and hold its own against Penryn because AMD doesn't have the wherewithall to redesign their µ-arch as often as Intel does. Such is the situation we find ourselves in now. Remember Intel was hurting with Netbust, it had to come to market first with a new µ-arch. AMD had the advantage of being able to compete a little longer with an older chip, Intel did not.
Randy - true true. It's a funny world. If you go further back one could even say Intel is the one who stayed with their older architecture (Netburst) and skipped a new introduction cycle in 2002-2003 thereby trying to live off their old technology much longer.
But I agree, post Nehalem the gloves are off from Intel. AMD will be playing perennial catch up with their own µ architecture. Couple that with this whole 1-2 punch of µ-arch and new process technology every alternate year is quite a clever idea IMHO. But again, it requires consistent execution...something Intel seems to lose every few years as they become complacent I suspect.
"I don't know what you see, but I see K8L quad at MS high tech summit. Maybe the vapor your seeing is that escaping your bong."
Nice now let me see you put that big wafer in your pc case and run windows on it. Sumone says k8L is vapor ware and the comeback is a preproduction picture of a wafer........pathetic.
Maybe you should get high on the vapors of YOUR bong cuz its going to be a long year for you amd fanbois.
Your second point atleast is more logical than your first one,it makes an interesting argument instead of cheering for a picture of a wafer.
Interesting posts. I don't mind being called a fanboy because if you have no foundation for your arguments you might as well start with an ad hominem attack.
However, there are several factual corrections that should be made.
If you've read my blog you know that I've said that Intel has the lead on the desktop for E6600 and up; AMD currently has nothing to compete. So, apparently, I'm either being called a fanboy because you haven't read my blog and are basing your statement on nothing or perhaps you feel that I should be saying that Intel is ahead on every Conroe processor.
Secondly, I do not expect K8L to trounce Conroe. I am expecting K8L to be a bit ahead in terms of FP/SSE but probably slower per clock in terms of Integer. Kentsfield is a different story because of the dual die design. I'm not exactly sure how someone could argue this point as the advantage of Yonah and Conroe over Smithfield and Presler should be obvious. As far as I know, Intel will release a true native quad design and this would remove Kentsfield's deficiencies.
The argument about a K8L demo is ludicrous. AMD simply does not leak its chips to be benchmarked before release. AMD has stated multiple times that they will demo K8L in the last quarter of this year. There is no reason to expect any benchmarks before the demo.
Vaporware is when you say that you are building hardware that never gets released, like Tejas and Whitefield. A paper launch is when you say that hardware is available and it is not. For example, upon release Intel's chips are not usually available for a month or more. AMD's are available the same day as release. Generally though it is considered acceptable if chips are available at the end of the month of release even if they are not available the same day.
The argument about how often the architecture can be upgraded is not quite right. AMD is moving to a modular design. But, this won't affect core releases until 2008.
BTW, did you fail to notice that Intel didn't mention either an onboard memory controller or CSI?
Scientia my apologies to you, i have not read the blog that you speak of but you provide a very logical argument...therefore i take my fanboi comment back.
I dont mind people supporting their company afterall everyone has a lil bias but it does piss me off when one company is held at a higher or different standard than the others.
However i believe that for the price every intel conroe chip is better than or equal to its similarly priced counterpart and not just the E6600 and above. The only place amd has left is chips under $180 dollars which will go for the mainstream people but not the computer enthusiast.
The only chip by amd that has good performance and value is the 5000+ but ever since the release of the price cuts you can not find those chips on the market unless you buy a prebuilt system from a manufacturer.
Scientia - I need to apologise. I should have been more explicit instead of just throwing a casual "vaporware" comment. My point remains, until I see the technology actually functioning, making the specific assumption that performance will be as hypothesized is not a good use of time. I too believe K8L will give Core 2 and Core 2 Quad a run for their money. But until I see it happen I don't make absolute claims. Same for Penryn on 45nm...I'm not saying it will win. However, it will probably be a competitive part. But we have to wait and see, I'm not taking Otellini's word for it.
You discuss about CPUs but you all forget about the rest of PC.
AMD will be soon performance king in terms of integrated chipset with graphics. You also forget about upcoming HT 3.0 and DDR3 which gives AMD another huge advantage in terms of overall system performance.
I deal with things i can get in the near future and today, not fantasize about things in the long term. Intel will pwn all with their laser silicon technology.....see how pointless that is.
Though HT 3.0 and DDR3 are benefits but i will believe it when i see the benchmarks for it. I have already seen AMD go from ddr1 to ddr2 and that was lackluster at best.
I have noticed that AMD fans tend to focus more about the products AMD WILL provide compared to the products they ARE providing. Everyone is getting AM2 because it MITE or MITE NOT be compatible with k8l kinda like how reverse hyperthreading was supposed to come along and pwn all.
I dont remember any tech person telling people to buy prescott becuase it will be compatible with conroe. Deal with stuff thats out today and let the slide shows in trade forums deal with future technology.
I deal with things i can get in the near future and today
Now that Conroe is out.
Though HT 3.0 and DDR3 are benefits but i will believe it when i see the benchmarks for it. I have already seen AMD go from ddr1 to ddr2 and that was lackluster at best.
HT 3.0 has no real effect on the desktop; this is for server systems greater than 4-way. However, it does include some things like power saving mode which would help notebooks.
I have noticed that AMD fans tend to focus more about the products AMD WILL provide
Then you haven't been paying attention. When Anand previewed Conroe, he said it was six months away. K8L should be available in six months.
compared to the products they ARE providing.
This sounds like a sliding scale. Before Conroe was released it was what Intel will be releasing soon. Now that Conroe is out it is what is available now. I assume when K8L is released it will be back to what will be available soon so that Penryn can be counted.
Everyone is getting AM2 because it MITE or MITE NOT be compatible with k8l
No. AMD has guaranteed that K8L quad and dual core will work in the same socket.
I dont remember any tech person telling people to buy prescott becuase it will be compatible with conroe.
Funny because I do recall people who went out and bought "Prescott compatible" socket 478 motherboards and then discovered that Prescott wouldn't work.
Deal with stuff thats out today and let the slide shows in trade forums deal with future technology.
Right after an Intel release. However, if AMD has just released then go with what will be available soon.
Folks - this is an open discussion but it is completely fair to give AMD and Intel the same latitude...or lack thereof when giving them credit for future developments. Where I draw the line is on product performance because that is a huge unknown. However, will AMD get to 65nm...absolutely! Will Intel introduce 45nm. Sure. Stuff like this is absolutely fair to analyse into any hypothesis. But it doesn't make sense to assume product X will outperform product Y when neither of them have been benchmarked publicly.
thx
"Now that Conroe is out."
Correction before conroe i would recommend the X2 because it was clearly the superior processor some people have the ability to recognize performance when they see it instead of making excuses or coming up with future products.
"HT 3.0 has no real effect on the desktop; this is for server systems greater than 4-way. However, it does include some things like power saving mode which would help notebooks."
I could care less, it was one of the AMD fans touting its benefits in the first place. I will read up about it when i actually see it implemented and importance in the market place.
"This sounds like a sliding scale. Before Conroe was released it was what Intel will be releasing soon. Now that Conroe is out it is what is available now. I assume when K8L is released it will be back to what will be available soon so that Penryn can be counted."
Nope ever since conroe was first benchmarked every AMD fan including yourself discredited it only to be shut up by its performance when it was actually released. Your ASSumption is wrong and even if someone DOES do that wouldnt it be kinda hypocritical for you to be pissed off about that after all they learned the technique from AMD fan bois oh mr. "great intel's NEW conroe beat amd's three year old k8".
"No. AMD has guaranteed that K8L quad and dual core will work in the same socket."
Yes it mite be the same socket but explain how ddr3 and AM3 come into play if you are using the same motherboard that you use for your current AM2 system?
"Funny because I do recall people who went out and bought "Prescott compatible" socket 478 motherboards and then discovered that Prescott wouldn't work."
You are a liar!!! No one was expecting conroe to perform the way it did infact everyone was in denial till it was actually released and its performance credited. Prescott was the joke of the tech industry and the only people who purchased that were people in the mainstream that did not know the difference NOT PEOPLE LOOKING TO UPGRADE TO CONROE. Even intel die hard fans didnt know what to expect with conroe after the netburst fiasco for you to even make that statement makes you either ignorant or a liar.
"Then you haven't been paying attention. When Anand previewed Conroe, he said it was six months away. K8L should be available in six months."
Hahaahah man just when i thought you were smart you pull statements like this. ANAND DID A PREVIEW OF CONROE BECAUSE HE GOT THE CHANCE TO PHYSICALLY SEE THE MACHINE RUN AND BENCHMARKED (albeit it was intel setup but it was something tangilble nevertheless). What do you want anand to do now write an article about a preproduction inquirer picture and run benchmarks on it. K8l WILL NOT BE HERE IN SIX MONTHS FOR THE DESKTOP MARKET!!! Go update your information and dont tell me we are talking about servers now since we have been talking about desktops all this time.
Post a Comment