"Intel server strategy crashes as Xeon roadmap changes".
The Inquirer makes for fun reading but it's not the most reliable rag. I'm not sure many people on here have said much about Intel's roadmap being predictable. No CPU manf really has a rock solid roadmap.
Just keep on repeating to yourself that Intel will get to 65 nm first, by 6 months or more, and then will get to 45 nm first by 1 year or more, and then may be the only private semiconductor manufacturer to have a 32 nm fab.
Anything smaller than that (<22 nm) than that may require a capital investment (subsidy) from the Federal government.
LAUNCH OF SERVER CHIP DELAYED UNTIL MID-2006 TO BOLSTER QUALITY CONTROL
By Dean Takahashi
Mercury News
Hurting its prospects in the market for high-end server chips, Intel said Monday that it has delayed the launch of its newest Itanium microprocessor until the middle of 2006.
The world's biggest chip maker said the chip, code-named Montecito, had to be delayed from early 2006 until mid-2006 to ensure it fixes quality problems.
``I don't see this as a crushing delay, but it certainly is embarrassing,'' said Jonathan Eunice, an analyst at market analyst firm Illuminata. ``It sounds like a four- to five-month delay, and that's not a disastrous day for Intel.''
Just keep on repeating to yourself that Intel will get to 65 nm first, by 6 months or more, and then will get to 45 nm first by 1 year or more, ...
Crappier, slower CPUs made with a smaller process. Wow, the "megahertz myth" gets replaced by the "nanometer myth". That's progress for you.
Onlooker: Yes, Intel has a lot of experience tearing up their roadmaps and making new ones. So instead of having no roadmap, we'll have lots of wrong ones. That's progress for you.
As for the Inquirer, they've been more reliable than Intel's press releases...
(sigh) It looks like the Apple CPU Maker Curse has already taken hold at Intel...
Crappier, slower CPUs made with a smaller process. Wow, the "megahertz myth" gets replaced by the "nanometer myth". That's progress for you.
It's killed every good CPU and ISA in the past. Alpha is dead. MIPS Rx000 is dead. HP-UX PA-x000 nearly so. PowerPC tried and failed; all it has now is embedded and mainframes. Moto 68k? Was it a "never tried?"
Spin all you want PPC has been the biggest boondoggle for Apple. They were never able to put 1 ppc vs Intel except for its intro at 500 which ended being 450. There isnt a G5 that can compete with a Intel or AMD so Apple had to use 2 chips and lots of spin on its benches. So bad they had to cram 4 cores into the machine because there just hasnt been advancement on G4 or G5 in years. Intel,Amd anything is better then Stagnating year after year of stagnating PPC. Even now years later apple still is pushing the 2.0 G5. Apple is frozen solid with PPC and has been for years.
They were never able to put 1 ppc vs Intel except for its intro at 500 which ended being 450.
The PPC history is much, much older than that. And are you talking about MHz only?
Quote:
Originally posted by Aurora
So bad they had to cram 4 cores into the machine because there just hasnt been advancement on G4 or G5 in years. Intel,Amd anything is better then Stagnating year after year of stagnating PPC
So Intel's progress from 3.06GHz to 3.2GHz isn't stagnation, but the G5s progress from 2.0GHz to 2.7GHz in the same period is?
Fastest G4 1.67 from Apple and it is smoked by the otherside not even close.
Fastest single G5 is a 2.1 G5 iMac and its PPC can be hammered by the otherside in any game just go see barefeats.
Dual G5s can match exceed in many pro apps but here we are comparing 2 G5s with 1 Intel/AMD, In Gaming those dual G5s get hammered by the single AMD's & Intels. Blame coding, software etc but the gaming side on Mac has been weak and consumers which make up 95% of the market are gamers also.
Apple is achieving these speeds by throwing a refrigerator on top of these ppc chips. AMD for example is running FX57s with only a fan. Same with the new dual core stuff they have. PPC isnt all that, Apple Software division has saved Apples hardware division for years the way i see it.
Fastest G4 1.67 from Apple and it is smoked by the otherside not even close.
Fastest single G5 is a 2.1 G5 iMac and its PPC can be hammered by the otherside in any game just go see barefeats.
Fastest G4 is at 1.67GHz and the fastest single core G5 is at 2.7GHz. When comparing against AMD and Intel it doesn't matter what Apple do with their models.
Don't you find strange that it has the two extremities? It looks like choice to me, not failure.
At one point in time, IBM produced mass volumes of 604e and 750 PPC processors at 250 nm before Intel was able to produce Pentia on its 250 nm process. They even got to 200 nm before them too. That's also when (604e era) when PPC performed pretty compared to Pentium and Pentium Pro CPUs. Well, once Intel matured their 250 nm fab and got to 180 nm, it was all over for PPC as a PC market alternative. They tried. They failed. They simply couldn't capture enough of the PC market to justify continuing fab development faster than Intel.
It will take a very long slow turning of the ship for the situation to reverse. Intel misperforms for several years straight, while its competitor performs perfectly for several years. One chip architecture won't turn the tide. It has to multiple ones.
Just always remember the nuances of Moore's "2nd Law". The next generation fab will afford the semi manufacturer to build chips with twice as many transistors for approximately the same cost, or the same chip with the same number of transistors at approximately half the cost. (Going to 300 mm wafers, and in the future maybe even 450 mm wafers, further reduces costs per chip).
The kick in the butt though is that it costs twice much money to develop that next-gen fab. That means the manufacturer has to sell twice as many chips as before to make money.
If a company isn't keeping up with volume with opening markets, it doesn't have a chance, and they are indeed left with servicing niches of the semiconductor market like Freescale and IBM are, like PA Semi will. Intel is the big dog in the most profittable market. They have the money to move forward. IBM, maybe, they have been living off of a mainframe monopoly for a very long time now. AMD has to rely on IBM for semi tech. Sony-Toshiba may be big enough combined to support <65 nm development. Going forward, only Intel has market and money to go to 45 nm alone, maybe even 32 nm.
We're nearing the end of the line with CMOS. Who's going to have enough market to support a 20 billion $ fab for 22 nm in 2012+? The only one I can think of is the US Federal government.
As for Montecito delayed due to quality issues, don't sweat it. Montecito is a freaking 1.7 billion, that's billion, transistor processor. That's audacious. That's impossible. Yet Intel thinks it can fab it, sell it at reasonable cost, and make money from it. It's no surprise they are having fab issues with it.
At one point in time, IBM produced mass volumes of 604e and 750 PPC processors at 250 before Intel was able to produce Pentia on its 25 nm process. They even got to 200 nm before them too. That's also when (604e era) when PPC performed pretty compared to Pentium and Pentium Pro CPUs. Well, once Intel matured their 250 nm fab and got to 180 nm, it was all over for PPC as a PC market alternative. They tried. They failed. They simply couldn't capture enough of the PC market to justify continuing fab development faster than Intel.
It will take a very long slow turning of the ship for the situation to reverse. Intel misperforms for several years straight, while its competitor performs perfectly for several years. One chip architecture won't turn the tide. It has to multiple ones.
Just always remember the nuances of Moore's "2nd Law". The next generation fab will afford the semi manufacturer to build chips with twice as many transistors for approximately the same cost, or the same chip with the same number of transistors at approximately half the cost. (Going to 300 mm wafers, and in the future maybe even 450 mm wafers, further reduces costs per chip).
The kick in the butt though is that it costs twice much money to develop that next-gen fab. That means the manufacturer has to sell twice as many chips as before to make money.
If a company isn't keeping up with volume with opening markets, it doesn't have chance, and they are indeed left with servicing niches of the semiconductor market like Freescale and IBM are, like PA Semi will. Intel is the big dog in the most profittable market. They have the money to move forward. IBM, maybe, they been living off of a mainframe monopoly for a very long time now. AMD has to rely on IBM for semi tech. Sony-Toshiba may be big enough combine to support 65 nm develop. Going forward, only Intel has market and money to go to 45 nm alone, maybe even 32 nm.
We're nearing the end of the line with CMOS. Who's going to have enough market to support a 20 billion $ fab for 22 nm in 2012+? The only one I can think of is the US Federal government
And explains, with some sophistication, why Apple is going Intel.
As we go forward, PPC's problems are only compounded by a lack of interest, investment, guts...motivation... IBM/Motorola don't have a big enough PPC market. Apple will have a big enough Intel market! It's the OS. Apple's time has come...
The G5 Quad is going to be a nice Workstation. I'm going to buy it the moment the 7800GT does live.
But it's funny that a good to 'average' PC will destroy it in games.
The sooner we move to Intel chips the better for overall performance in serious and game apps.
The boost Apple can get from Gamers to the tower line and their other machines will be significant if they play their cards right with the Intel transition. I'd like to see a range of mini-towers below the 'dual' workstations.
THE K10 LOOKS TO BE dead, or at least really really delayed.
...
Like the horrendously complex eight issue K9 before it, I would bet good money that it will slip beneath the waves without a hiccup.
Either way, if you were expecting the K10 in 2007, don't, and maybe not in 2008 either. To make up for it, there is a new chip called K8L to slot in the middle, but little is known about that, as yet. ...
Comments
Originally posted by cubist
http://www.theinquirer.net/?article=27192
"Intel server strategy crashes as Xeon roadmap changes".
The Inquirer makes for fun reading but it's not the most reliable rag. I'm not sure many people on here have said much about Intel's roadmap being predictable. No CPU manf really has a rock solid roadmap.
Anything smaller than that (<22 nm) than that may require a capital investment (subsidy) from the Federal government.
LAUNCH OF SERVER CHIP DELAYED UNTIL MID-2006 TO BOLSTER QUALITY CONTROL
By Dean Takahashi
Mercury News
Hurting its prospects in the market for high-end server chips, Intel said Monday that it has delayed the launch of its newest Itanium microprocessor until the middle of 2006.
The world's biggest chip maker said the chip, code-named Montecito, had to be delayed from early 2006 until mid-2006 to ensure it fixes quality problems.
``I don't see this as a crushing delay, but it certainly is embarrassing,'' said Jonathan Eunice, an analyst at market analyst firm Illuminata. ``It sounds like a four- to five-month delay, and that's not a disastrous day for Intel.''
Originally posted by THT
Just keep on repeating to yourself that Intel will get to 65 nm first, by 6 months or more, and then will get to 45 nm first by 1 year or more, ...
Crappier, slower CPUs made with a smaller process. Wow, the "megahertz myth" gets replaced by the "nanometer myth". That's progress for you.
Onlooker: Yes, Intel has a lot of experience tearing up their roadmaps and making new ones. So instead of having no roadmap, we'll have lots of wrong ones. That's progress for you.
As for the Inquirer, they've been more reliable than Intel's press releases...
(sigh) It looks like the Apple CPU Maker Curse has already taken hold at Intel...
Originally posted by cubist
Crappier, slower CPUs made with a smaller process. Wow, the "megahertz myth" gets replaced by the "nanometer myth". That's progress for you.
It's killed every good CPU and ISA in the past. Alpha is dead. MIPS Rx000 is dead. HP-UX PA-x000 nearly so. PowerPC tried and failed; all it has now is embedded and mainframes. Moto 68k? Was it a "never tried?"
It isn't a myth.
Originally posted by THT
PowerPC tried and failed; all it has now is embedded and mainframes.
Don't you find strange that it has the two extremities? It looks like choice to me, not failure.
Originally posted by Aurora
They were never able to put 1 ppc vs Intel except for its intro at 500 which ended being 450.
The PPC history is much, much older than that. And are you talking about MHz only?
Originally posted by Aurora
So bad they had to cram 4 cores into the machine because there just hasnt been advancement on G4 or G5 in years. Intel,Amd anything is better then Stagnating year after year of stagnating PPC
So Intel's progress from 3.06GHz to 3.2GHz isn't stagnation, but the G5s progress from 2.0GHz to 2.7GHz in the same period is?
Fastest single G5 is a 2.1 G5 iMac and its PPC can be hammered by the otherside in any game just go see barefeats.
Dual G5s can match exceed in many pro apps but here we are comparing 2 G5s with 1 Intel/AMD, In Gaming those dual G5s get hammered by the single AMD's & Intels. Blame coding, software etc but the gaming side on Mac has been weak and consumers which make up 95% of the market are gamers also.
Apple is achieving these speeds by throwing a refrigerator on top of these ppc chips. AMD for example is running FX57s with only a fan. Same with the new dual core stuff they have. PPC isnt all that, Apple Software division has saved Apples hardware division for years the way i see it.
Originally posted by Aurora
Fastest G4 1.67 from Apple and it is smoked by the otherside not even close.
Fastest single G5 is a 2.1 G5 iMac and its PPC can be hammered by the otherside in any game just go see barefeats.
Fastest G4 is at 1.67GHz and the fastest single core G5 is at 2.7GHz. When comparing against AMD and Intel it doesn't matter what Apple do with their models.
Originally posted by PB
Don't you find strange that it has the two extremities? It looks like choice to me, not failure.
At one point in time, IBM produced mass volumes of 604e and 750 PPC processors at 250 nm before Intel was able to produce Pentia on its 250 nm process. They even got to 200 nm before them too. That's also when (604e era) when PPC performed pretty compared to Pentium and Pentium Pro CPUs. Well, once Intel matured their 250 nm fab and got to 180 nm, it was all over for PPC as a PC market alternative. They tried. They failed. They simply couldn't capture enough of the PC market to justify continuing fab development faster than Intel.
It will take a very long slow turning of the ship for the situation to reverse. Intel misperforms for several years straight, while its competitor performs perfectly for several years. One chip architecture won't turn the tide. It has to multiple ones.
Just always remember the nuances of Moore's "2nd Law". The next generation fab will afford the semi manufacturer to build chips with twice as many transistors for approximately the same cost, or the same chip with the same number of transistors at approximately half the cost. (Going to 300 mm wafers, and in the future maybe even 450 mm wafers, further reduces costs per chip).
The kick in the butt though is that it costs twice much money to develop that next-gen fab. That means the manufacturer has to sell twice as many chips as before to make money.
If a company isn't keeping up with volume with opening markets, it doesn't have a chance, and they are indeed left with servicing niches of the semiconductor market like Freescale and IBM are, like PA Semi will. Intel is the big dog in the most profittable market. They have the money to move forward. IBM, maybe, they have been living off of a mainframe monopoly for a very long time now. AMD has to rely on IBM for semi tech. Sony-Toshiba may be big enough combined to support <65 nm development. Going forward, only Intel has market and money to go to 45 nm alone, maybe even 32 nm.
We're nearing the end of the line with CMOS. Who's going to have enough market to support a 20 billion $ fab for 22 nm in 2012+? The only one I can think of is the US Federal government.
As for Montecito delayed due to quality issues, don't sweat it. Montecito is a freaking 1.7 billion, that's billion, transistor processor. That's audacious. That's impossible. Yet Intel thinks it can fab it, sell it at reasonable cost, and make money from it. It's no surprise they are having fab issues with it.
[edit: grammer. grrr....]
Originally posted by THT
...
.... Montecito is a freaking 1.7 billion, that's billion, transistor processor. ...
Even at that, Motorola and IBM trump its' 4 -5 month delay. Sometimes by years in the case of Motorola
At one point in time, IBM produced mass volumes of 604e and 750 PPC processors at 250 before Intel was able to produce Pentia on its 25 nm process. They even got to 200 nm before them too. That's also when (604e era) when PPC performed pretty compared to Pentium and Pentium Pro CPUs. Well, once Intel matured their 250 nm fab and got to 180 nm, it was all over for PPC as a PC market alternative. They tried. They failed. They simply couldn't capture enough of the PC market to justify continuing fab development faster than Intel.
It will take a very long slow turning of the ship for the situation to reverse. Intel misperforms for several years straight, while its competitor performs perfectly for several years. One chip architecture won't turn the tide. It has to multiple ones.
Just always remember the nuances of Moore's "2nd Law". The next generation fab will afford the semi manufacturer to build chips with twice as many transistors for approximately the same cost, or the same chip with the same number of transistors at approximately half the cost. (Going to 300 mm wafers, and in the future maybe even 450 mm wafers, further reduces costs per chip).
The kick in the butt though is that it costs twice much money to develop that next-gen fab. That means the manufacturer has to sell twice as many chips as before to make money.
If a company isn't keeping up with volume with opening markets, it doesn't have chance, and they are indeed left with servicing niches of the semiconductor market like Freescale and IBM are, like PA Semi will. Intel is the big dog in the most profittable market. They have the money to move forward. IBM, maybe, they been living off of a mainframe monopoly for a very long time now. AMD has to rely on IBM for semi tech. Sony-Toshiba may be big enough combine to support 65 nm develop. Going forward, only Intel has market and money to go to 45 nm alone, maybe even 32 nm.
We're nearing the end of the line with CMOS. Who's going to have enough market to support a 20 billion $ fab for 22 nm in 2012+? The only one I can think of is the US Federal government
Mandatory reading for all those PPC fan boys.
Lemon Bon Bon
And explains, with some sophistication, why Apple is going Intel.
As we go forward, PPC's problems are only compounded by a lack of interest, investment, guts...motivation... IBM/Motorola don't have a big enough PPC market. Apple will have a big enough Intel market!
The G5 Quad is going to be a nice Workstation. I'm going to buy it the moment the 7800GT does live.
But it's funny that a good to 'average' PC will destroy it in games.
The sooner we move to Intel chips the better for overall performance in serious and game apps.
The boost Apple can get from Gamers to the tower line and their other machines will be significant if they play their cards right with the Intel transition. I'd like to see a range of mini-towers below the 'dual' workstations.
And a goddamn monitor bundle. Too much to ask?
lemon bon bon
AMD's K10 is delayed or dead
THE K10 LOOKS TO BE dead, or at least really really delayed.
...
Like the horrendously complex eight issue K9 before it, I would bet good money that it will slip beneath the waves without a hiccup.
Either way, if you were expecting the K10 in 2007, don't, and maybe not in 2008 either. To make up for it, there is a new chip called K8L to slot in the middle, but little is known about that, as yet. ...