The buzzword is customer-centric innovation
Even Michael Dell is following Hector Ruiz's "customer-centric innovation" doctrine. This must be a big change for Dell.
There is no doubt that going forward, AMD will dominate the computing landscape and AMD64* will become the industry standard, not only because AMD64 is vastly superior than IA32, but because AMD will grow the ecosystem and allow others to prosper within the AMD64 food chain. The news that SUN Sparc and AMD Opteron will share a common socket is huge and has profound implications to the enterprise. Imagine that you can buy a single SUN server that runs SPARC and AMD64 at native speed the same time. AMD's ccHT will be the fabric that interconnects different types of compute cycles, like CISCO routers that connect computers over the internet. AMD's Torrenza will combine and multiply the forces of its partners.
* AMD64 is not just about the AMD64 instruction set, which Intel has partially copied, it is more about Direct Connect Architecture.
posted by Sharikou, Ph. D. @ 7/10/2006 11:36:00 AM
29 Comments:
Wondering if you saw this article.
Seems the pricewar may be benefiting Intel, more than AMD.
Another story seems to have Intel still very close to Dell, releasing two new chips.
Also, I know you don't like TG Daily, but they have an interseting article about future plans to combat the SUN system.
Have a good one.
"Wondering if you saw this article.
Seems the pricewar may be benefiting Intel, more than AMD."
Uhhhh... OK? In what way? Notebooks yea, desktop, nope! Intel's already stuffed the channels with P4's that nobody within thier right mind(i.e. intelligent consumers) would want, no surpise there, they gotta move it somehow.
"AMD still holds the lion's share of the desktop market. AMD accounted for 73 percent of all retail desktops in the United States, while Intel accounted for only 26.8 percent. But it's a mixed blessing, Bhavnani noted. AMD in May was dominant in the machines that cost $750. The chipmaker accounted for 84.5 percent of desktops costing up to $500 and 88.9 percent of desktops priced between $500 and $750."
"Another story seems to have Intel still very close to Dell, releasing two new chips."
Dell just signed a deal with AMD to not only sell servers with Opterons but are planned on doing desktops and notebooks too. Dell and Intel's marriage made in hell days are long over with now they have signed on the bottom line with AMD, that's a fact.
"Also, I know you don't like TG Daily, but they have an interseting article about future plans to combat the SUN system."
By that time(2010) K10 will be released, the K10 will have possibly be octocore, Hypertransport 4.0 or higher will be implemented and Opterons using HTX/Co-processors with Torrenza technology. With a almost seemingly limitless bus like Hypertransport, having so many cores to a processor and the Torrenza technology to specialize a serversetup for specific tasks due to the HTX/co-processors, Intel has ALOT of planning to do even hope to combat AMD and Sun's plans on the server front. They're already behind if they are merely planning 32 cores to have a hope in hell to combat that platform. Flexability of the architecture is the key here, it's going to take more than sheer brute strength of multiple cores as Intel likes to tout in that little article to best what thier opposition has planned.
Torrenza is a literal paradigm shift in server computing, it's going to radically change the server market and every one in the industry will want to be involved, be assured of that. By that time, once Torrenza is wholly adopted and co-processors from other companies are also in the market, Intel will have no choice but to conform and adopt Torrenza in or become a dinosaur, bound for extinction, it's that simple.
Maybe Dell sees AMD's new
generation of Turion aka Bulldozer will be more competitive in 07, hence Michael Dell is hedging his bet.
Way to be selective...
You left this out, "By contrast, Intel accounted for 52.1 percent of desktops costing $750 to $999 and a whopping 91.7 percent of those starting at $1,250 that month."
"Dell just signed a deal with AMD" No way, are you serious, its only been on every website for the last month. The point of the link was to point out that Intel and Dell are still close, still giving Dell preferable treatment (try and win them back), even in the midst of AMD.
"Torrenza is a literal paradigm shift in server computing," I am not familiar with it so I can not comment on it, but I can comment on this "bound for extinction, it's that simple.", nothing is that simple, AMD has to weather the next 4 quarters with out a new chip, and with out the performance crown (desktops/mobile anyways). They have no answer for Conroe until quarter 4 2007. 65nm is slow moving as well.
I trully believe that Intel has an ace up their sleev, in that Conroe is capable of running 1333 FSB, and clock to almost 3.5Ghz on air, they probably will do that when the FX64 comes out, but they might not have to... all in all AMD is in trouble, and bad.
"Maybe Dell sees AMD's new
generation of Turion aka Bulldozer will be more competitive in 07, hence Michael Dell is hedging his bet."
Looks like you have brain washed by reading all those crappy analysts..
All Mike sees is a nice lucerative product line with AMD processors now and in the future..and I strongly believe that he is probably crying that ,why he didnt go AMD 3 years back.. bottomline all Mike cares about is his $$
I also believe it's best for both Sun & AMD if T1/Opteron could work side-by-side by sharing the same socket. The only regret is that it's a bit too late (2008). Two yearsis half a lifetime for a microarchitecture! Sun should've planned this out in 2003 and realized it as soon as socket F comes out.
My other doubt is that these two are different processors. How do you have a single system image when one executes UntraSparc and the other x86 instruction sets? Which operating system would allow you do this, is it even possible?
"Way to be selective..."
Take your own advice, Intel ain't doing near as hot as you make them out to be, that is why I included that snippet.
"No way, are you serious, its only been on every website for the last month. The point of the link was to point out that Intel and Dell are still close, still giving Dell preferable treatment (try and win them back), even in the midst of AMD."
There is no merit in hoping for sweetheart deals from Paul anymore as their kickbacks and cuts are already equivlent to thier current wholesale price! It's a matter of moving inferior inventory, moving old P4 stock! That is why Dell is going AMD because EVERYONE is getting basically wholesale price to remove Intel's mountains of P4's, they AREN'T going to make much if any off that inventory! Also, if Dell and Intel were still so buddy buddy AMD wouldn't even be doing deals with Micheal Dell because Intel would yank on his chain the moment he even looks at AMD. Bottom line: the rules have changed, the market has changed and Dell knows it.
"Torrenza is a literal paradigm shift in server computing," I am not familiar with it so I can not comment on it, but I can comment on this "bound for extinction, it's that simple.", nothing is that simple, AMD has to weather the next 4 quarters with out a new chip, and with out the performance crown (desktops/mobile anyways). They have no answer for Conroe until quarter 4 2007. 65nm is slow moving as well."
What you fail to understand is that Torrenza and it's co-processor technology as been stated is going to allow for SPECIALIZATION of server tasks, this is paramount for big iron servers! As far I know Intel has nothing of this sort in it's plans and that could be a huge disaster in a market they are slowly losing out to AMD.
As far as the next 4 quarters Intel MAY have the crown, they may loose it too with Rev. G on 65nm too. Until both AMD and Intel have products on the market this IS ALL CONJECTURE and nobody knows for sure what is going to happen. Also, AMD's 65nm is right on track with mature yields, 65nm AM2 processors will be here in December. Above all, go research and go read up some more as you are horridly uniformed on the matter it seems, Google is your friend.
"I trully believe that Intel has an ace up their sleev, in that Conroe is capable of running 1333 FSB, and clock to almost 3.5Ghz on air, they probably will do that when the FX64 comes out, but they might not have to... all in all AMD is in trouble, and bad."
1333 mhz on a alternating one-way outdated FSB design vs. a 1GHz serial(up and down commuicating) Hypertransport design. Oooookay. Sure Intel can crank out a 3.5 ghz clockable ENGINEERING SAMPLE out of a development fab, big deal. So could AMD or IBM or Texas Instruments. Seriously you doth protest too much, go read up on AMD's announcements of K8L and it's capabilities before you pull more out of your crack thanks.
To me, it seems like intel is the one still on it's heels. Intel is not "way ahead" in the 65mn development, they are just quicker to rush out. If AMD didn't mind 40% yeilds, they could 'probably' release them now. Seems like Intel released 65nm to "prove they are better" (RE-action). But all said and done, it's all conjecture until AMD does. (unless someone here is an insider...doubtful)
Other than that, AMD is the one working on platforms and not just "fast cpus" - at least ones that look beneficial to people/companys and not just intel pocketbooks
"How do you have a single system image when one executes UntraSparc and the other x86 instruction sets? Which operating system would allow you do this, is it even possible?"
Using virtualization you could have different OS's running, as well as the Sparc can be configured as a Co-Processor or vice versa.
I analyzed the 32-Core CPU from Craptel here.
Whats up with the 32Core in 2010 BS Intel story, reminds me of the 10GHz chip that they were supposed to release..ha ha.
I am sure that more BS Intel stories will follow soon to keep up the media focus and fool people.
But Intel marketing should realise that it isnt the 1990s anymore when people brought such bullsh*t !!
Ok, not trying to bash AMD or make Intel look like roses.
"What you fail to understand is that Torrenza and it's co-processor technology as been stated is going to allow for SPECIALIZATION of server tasks, this is paramount for big iron servers!"
Like I had said I don't know anything about Torrenza, so thank you for explaining it.
"Seriously you doth protest too much, go read up on AMD's announcements of K8L and it's capabilities before you pull more out of your crack thanks."
In regards to K8L, four quarters away, Conroe 4 days, and Intel is supposed to be shifting to 45nm around that time. AMD is going to be playing catchup for a little while.
People who like AMD always bash Intel for their lack of an internal memory controller, yet they are able to achieve performance without it, let me be specific, desktop and mobile platforms. If and when they decide to incorperate an IMC and a form of HT they will be doing very well.
"Until both AMD and Intel have products on the market this IS ALL CONJECTURE and nobody knows for sure what is going to happen."
You are exactly right.
Sorry for bouncing around, just trying to respond to all your points.
"Above all, go research and go read up some more as you are horridly uniformed on the matter it seems, Google is your friend."
I should have been more specific, FX processors will not go to 65nm until late next year, unless I am wrong and very well could be, but here is a link.
Again not trying to bash, just looking at whats out there.
"Intel is not "way ahead" in the 65mn development, they are just quicker to rush out. If AMD didn't mind 40% yeilds, they could 'probably' release them now."
AMD's Henri Richard has already stated that Intel moves to new processes a year before AMD, but their yields don't even get close to AMD's for another 6-7 months, which means Intel loses HUGE every time they move processes while AMD savors and matures their process's before fully ramping it; meaning cheaper CPU's for everybody.
Check this page:
http://www.amdcompare.com/techoutlook/
Check the line 'More Than Two Cores'. And then, check the availability date.
Check the last line in the first section:
Greater than 8-way Computing -> 2008.
This is the Cadiz and the Zamora.
They are 32-way, quad-cores chips!
AMD 128 cores in 2008 vs Intel 128 cores in 2009/2010.
It's too late for Intel...
"HyperTransport technology was invented at AMD with contributions from industry partners and is managed and licensed by the HyperTransport Technology Consortium, a Texas non-profit corporation."
"The OPMA (Open Platform Management Architecture) specification is just one more example of how AMD innovates within industry standards to provide real solutions to real problems. Open standards puts the power of choice back in the hands of customers."
"Torrenza is the realisation of its long-term plan to create an open and extensible microarchitecture. These announcements represent AMD’s future – and a view into global cooperation we are well on our way into creating" - AMD corporate vice-president and chief technology officer Phil Hester.
"Chartered's already strong capabilities, enhanced with the speed, accuracy and agility of APM, will help Chartered attain a rapid AMD64 product production ramp with high yields and high quality." - Gary Heerssen, senior vice president, Corporate Manufacturing Group, AMD.
"We are extremely pleased to enter into a strategic relationship
with AMD and to bring them on as an investor." - Atiq Raza, chairman and chief executive officer, Raza Microelectronics, Inc.
SemIndia Inc., a consortium of Indian technologists, has agreed to lead the investment of $3 billion in a wafer fab in India to make chips using technology licensed from AMD. "AMD is acknowledged as a technology leader in the industry, and is the ideal partner for India's foray into semiconductor manufacturing. We are excited to see the commitment demonstrated by AMD and SemIndia, and the Government of India is backing this initiative solidly.” - Thiru Dayanidhi Maran, Minister for Communications and Information Technology in the Indian federal government.
"As a customer-centric company, we embrace and promote a collaborative, connected business model." - Marty Seyer, corporate vice president and general manager, Microprocessor Business Unit, Computation Products Group, AMD.
In 3 words:
AMD = Sharing, Cooperation, Partnership.
INTEL = Me, Myself and I.
Customers don't like getting bent over just to add RAM to their systems. This is what is happening with with Intel's low volume FB-DIMM.
Dell's latest FB-DIMM, (i.e. Full Bendover DIMM), at 32GB specs out to:
32GB 533MHz (8x4GB), Dual Ranked DIMMs [add $29,679 or $785/month1]
Call it $30K.
If you wanted to build a 32GB Opteron system, you could pay $4560, call it $5K for 16 2GB DDR400 ECC DIMMs.
With the rest of the money, $25K, you could a 4P (four physical processor) system with 4 Opteron 8XX series processors. Giving you eight cores.
And still you'd have money left over. And this is just considering the additional RAM for the Dell, not even the Dell system itself!
This is just one data point, an important one, that shows how Intel is just a rip-off. There is no VALUE proposition. Just an endless rape of the customer who is forced to buy half-baked crap at top price.
It is no secret that an 8 core 32GB Opteron system will completely frag a 4 core 32GB Intel system.
Smart customers do the math and buy AMD.
"AMD's Henri Richard has already stated that Intel moves to new processes a year before AMD, but their yields don't even get close to AMD's for another 6-7 months, which means Intel loses HUGE every time they move processes while AMD savors and matures their process's before fully ramping it; meaning cheaper CPU's for everybody."
Nothing like an unbiased, objective speculation about a competitor's yield performance - I don't suppose you have any ACTUAL DATA?
If Intel said their yields are 20% better than AMD would you believe them?
Suppose for example yields are 2X worse and you get ~2X the #/die wafer wouldn't that be about breakeven? (This is not quite correct as I'm ignoring testing costs and the additional technology cost for the next gen wafer processing).
Are you saying Intel's yields are >2X worse than previous technology for the first 18 months of the ramp? (The one year they are ahead + the 6-7 months to "catch up" to AMD)
i can't wait to upgrade to K8L in Q1 of 2007. nuff said.
"Ok, not trying to bash AMD or make Intel look like roses."
That's fine, I have no problem with that but I also give credit where it's due and so far, the only credit I will give to Intel is finally (FINALLY) getting back in the game and getting competitive with AMD and helping lowering prices. Personally I feel the NGMA (basically an upgraded PIII/Pentium Pro) technology is a huge step up from the P4, but it's still a ways from what the K8 is evolving into.(K8L)
"Like I had said I don't know anything about Torrenza, so thank you for explaining it."
I will further expand here:
Co-processors of the Torrenza technology works like this, instead of plugging in another Opteron into a 4 socket board,(or higher) one of them can be a specilized co-processor like a GPU from Nvidia or something like Sun's SPARC processor or whatever ends up out on the market. This will allow a server to be specilized to a specific task such as a data processing server, a graphics rendering powerhouse and so forth all thanks to whatever kind of co-processor an IT or system builder plugs into the board. It's all seamless too, hardware based, no special software will be necessary either from what I have read. The performance gains from a server that is more specialized in this manner are immense.
"In regards to K8L, four quarters away, Conroe 4 days, and Intel is supposed to be shifting to 45nm around that time. AMD is going to be playing catchup for a little while."
At the latest K8L will be out in 1H of 2007. Rev. G at 65nm could be tweaked before then to provide enough performance to surpass Conroe, Woodcrest and Merom during this timeframe.
Here is also another certainty, Intel's NGMA architecture will ONLY be able to cover about 30% of Intel's product mix at the end of the year. It won't be until next year until Intel is completely transitioned and until then they must clear thier inventories of inferior P4 products at a HUGE loss. On top of that, slashing thier bottom line with Core 2's pricing is also asking for serious trouble. Such drastic actions to regain marketshare back from AMD will have some dire consequences eventually for Intel financially, that you can bank on too. (Mr. Sharikou has already expanded on this viewpoint in detail)
FWIW, Intel's 90nm Prescott's sucked in power consumption and heat dispersing compaired to AMD's 130nm processors. IMHO AMD will not need 45nm to compete with Intel while they are at 300mm wafers and at 65nm if history repeats itself as the 2nd generation of SOI and has proven it's worth in AMD's K8 line at 90nm.(low wattage, highly overclockable, low heat) As long as thier 65nm tech rolls out as well as thier 90nm did (in which yields are doing quite well from what I read) AMD will remain quite competitive with Intel. The specialized APM software of thier FABs play a huge and crucial key to thier quick ramping and high yields, which is vitally important against Intel having more FAB capacity.(albiet thier slower ramping time and lower yields)
"People who like AMD always bash Intel for their lack of an internal memory controller, yet they are able to achieve performance without it, let me be specific, desktop and mobile platforms. If and when they decide to incorperate an IMC and a form of HT they will be doing very well."
Yes, but at what cost? Cache thrashing is a very real problem with Core2. AM2's potential with DDR2 has yet to be realized and seen until low latency 1ghz DDR2 is out on the market. The true power of ccHT(cache coherent Hypertransport) in the 3.0 spec has yet to be seen as well. Sadly Intel will never adopt HT as thier corporate mindset is not willing to accept anything that AMD has been involved with. (ccHT is unlike regular HT, it also has to be liscenced and the liscenee pays a fee to the HT Consortium to use it)
Until Intel get's CSI working as a viable alternative to HT,(2009 at the latest?) their architecture will continue to be bottlenecked by the aging FSB, especially with Woodcrest's massive need for bandwidth between it's cores, this is a very observable fact.(and why Opteron still rules the 4P and up arena, 2p is a toss up IMHO)
"I should have been more specific, FX processors will not go to 65nm until late next year, unless I am wrong and very well could be, but here is a link."
I don't read anything that says to the contray of AMD's announcement of 65nm processors, that means all of thier products(that means FX) across the board will be transitioned to 65nm in December.
"Again not trying to bash, just looking at whats out there."
Same here, while I may be a loyal AMD customer, I see the whole picture too for good or for the bad for either Intel or AMD. Right now, Intel has a good architecture considering thier lackluster P4 line and I'm sure it will do well for them for the time being. However thier success I feel will be short lived as AMD isn't sitting on their laurels either and have some big things coming in the future. If you think AMD is going to let Intel keep that crown long... Well let's just say Intel won't get to keep it long enough to give it a good spit shine. ;^) AMD knows they are in for a long fight, they will do everything to keep that momentum and the ground they took from Intel that they have worked very hard for in the past few years.
AMD = Sharing, Cooperation, Partnership.
INTEL = Me, Myself and I.
wow ....(speechless)
just short reminder on what intel did that benefit the whole industry inclusive AMD and the users inclusive you and me
PCI, USB, AC97. if you hate intel so much, just try throw away those things that has that.
"AMD = Sharing, Cooperation, Partnership.
INTEL = Me, Myself and I.
wow ....(speechless)
just short reminder on what intel did that benefit the whole industry inclusive AMD and the users inclusive you and me
PCI, USB, AC97. if you hate intel so much, just try throw away those things that has that."
While Intel's early innovations are indeed with merit, however they as a company have since long lost that focus and innovative spirit. No one can dispute thier achievements, but no one cannot deny thier unscruplous(and possibly unlawful) business tactics either. Thanks to the boldness of AMD, this long running scenrio in the industry is finally coming to light and Intel could pay dearly for it.
The fact they do all they can legally (and illegially) to establish and keep thier monopoly over the past decades is a crime to thier competition and the rest of the industry. How much of those crimes will remain to be seen in the anti-trust case. So far things don't bode well for them as they are already in hot water abroad.(the Intel KK office raids in Japan for starters...)
Thanks for the insight and not being biased as many AMD people here are.
"The Athlon 64 FX-64 will be AMD’s desktop flagship processor until Q3’07 when the Athlon 64 FX-66 arrives. This part will have a 125-watt TDP as well and presumably be clocked at 3.2 GHz unless AMD has some other trick. It will be based on the 90nm Windsor core and not the upcoming 65nm Brisbane core."
Maybe I am miss reading this here, but it seems the FX will be 90nm, not trying to push the point, but rather clarify what I am reading.
"At the latest K8L will be out in 1H of 2007."
Yeah I read that this morning an feel pretty smart right now, the news and speculation changes every day...:)
"(Mr. Sharikou has already expanded on this viewpoint in detail)"
I have to admit anything he says I have to take with a grain of salt, its very hard to get around some of the biassed opinions he shows, but all in all I don't want to read that everything is "hunki dori for Intel" which is why I like to read people like yourselfs comments on the matters at hand.
Thanks.
"At the latest K8L will be out in 1H of 2007."
I have to say many Intel fanbois have very low IQ. Read this new again
http://www.digitimes.com/mobos/a20060710A1001.html
It stated that K8L on socket AM3 won't be available until 2008, it went further to say K8L is compatible with AM2 socket. Did that mean K8L delayed?
Go read AMD's CTO's June 1 presentation, on page 12 of his slide, it was clearly stated K8L server and desktops will be introduced in mid-2007. And I bet that's part of the reason DELL goes AMD.
"I have to say many Intel fanbois have very low IQ."
Then I must say that many AMD fanbois (You) are completely ingulfed in the hopes and dreams that AMD can rebound against Intel that you would fail to look at the big picture and realize that AMD is definitely in trouble, you sound like a f#$%ing soccer fan whos team is down 4 goals with a minute to play begging for some one to cause a penalty or some sh!t, I am by no mean saying that AMD can not and will not reclaim the crown, but in a few days the desktop crown will be handed over.
Ok I went and read it AGAIN, I must be a complete moron because it does say "that K8L on socket AM3 won't be available until 2008, it went further to say K8L is compatible with AM2 socket." and then if you get past the AMD bliss you must have been feeling you would have read on and read the part about "While AMD is expected to not make any changes to the micro-architecture of its CPUs in the next 18 months or even longer" K8L IS A NEW MICRO-ARCHITECTURE, isn't it.
"AMD's CTO's June 1" vs an article written July 10, you of all people know things can change dramatically in a months time.
It would be nice if you didn't slam your readers, I like Intel but AMD has alot of advantages especially in servers and I can say that without feeling like I'm cheating on my girlfriend, can you say something nice about Intel without feeling like your cheating on your girlfriend?
Keep blogging man.
While AMD is expected to not make any changes to the micro-architecture of its CPUs in the next 18 months or even longer" K8L IS A NEW MICRO-ARCHITECTURE, isn't it.
Micro-architecture is Intel's favourite word. For AMD, it's architecture. K8L has some enhancement on the execution core, but its major strength is where Intel is totally lacking: scalability and enterprise I/O requirements. AMD is at a completely different level from Intel. Look at SUN's new servers introduced today, they are simply impossible to build with Intel's outdated technology.
K8L was scheduled to be launched in mid-2007. AMD will pull it in earlier. Just wait and see.
"K8L was scheduled to be launched in mid-2007. AMD will pull it in earlier. Just wait and see."
this is why i can't wait to upgrade next year when k8l will be out. hopefully amd will price k8l competitively. i'm sure by the time k8l is out, conroe will drop in price alot. i guess we'll have to wait and see.
yield must be excellent for amd ot pull k8l up to an earlier release. it's amazing how efficient amd operates with 5 times less capacity than intel. this is the kind of company that you want to invest in.
it's all about servers and laptops...
while I would i agree AMD has a strong offering at the server end, but since when AMD is good at laptop??
If you do not know, just go and google at the AMD attempt in entering mobile. the first attempt was a total failure where chips didn't 'like' each other; and the current attempt is unproven yet.
while you might cry that Intel platform is a marketing stuff, it is actually not as simple as that.
re: Sparc and AMD64 running concurrently in the same system - I don't see it. Sparc is big-endian, AMD64 is little-endian, so data sharing would be mostly impossible. Obviously code sharing would be impossible, so the only way to run them would be to partition the machine into separate halves. What would be the point?
Is there anything that Sparc does well, better than AMD64? Again, what's the point?
The fact they do all they can legally (and illegially) to establish and keep thier monopoly over the past decades is a crime to thier competition and the rest of the industry. How much of those crimes will remain to be seen in the anti-trust case. So far things don't bode well for them as they are already in hot water abroad.(the Intel KKK office raids in Japan for starters...)
WOW!! i knew intel had some nazi working for them.... but goin all the way to KKK... aint thats something
Post a Comment
<< Home