I'm not so concerned with clock as I am with cores. I think an Itanium core is about 25 million transistors. So a 16 core Itanium with 32MB of on chip cache seems like it could happen. Sweet.
Is it much harder to translate directly to 32 nm, or is this just a case of making more money by slowing development down?
It's much harder, doubly as expensive, and takes twice as long to develop.
Moore's so-called 2nd law for CMOS fabs is that the cost to develop them doubles every node. And remember that the so-called 1st law is that chip density doubles every node, with each node advance occuring every 18-24 months.
It is purely an economic equation. Essentially, in order to recoup the cost of developing a next generation fab, that fab has to sell twice as many chips over the lifespan of the fab. If the market can't support that, then there is no point in developing the fab.
Keep in mind that Intel is shipping 65 nm CPUs right now. Everyone else is 6 months behind them. Poor Freescale is a full node behind Intel. AMD won't be shipping until 2H 06. IBM likely in 2H 06 as well. Freescale won't have a 65 nm fab, but will share one like they are now with the 90 nm fab, which incidently, we haven't seen the fruits yet. They won't be jumping any node advances because it costs too much, takes too much time and don't have the market to support such a development.
The cost of developing a 65 nm? Probably $3 to $5 billion. The cost of developing a 45 nm? Probably $5 to $10 billion. The cost of moving directly from 65 nm to 32 nm? Probably $15 billion, 4+ years, and in the meanwhile, the 65 nm chip performance will stagnate for 2 years, leaving you at a horrible performance and market disadvantage for those two years. No company can survive that.
The other problem? AMD's 2005 revenue was about $5.5 billion. Intel's 2005 revenue was about $33 billion. This is why Intel is always first to the next nodes. They have the money and the market to get there while other companies have to extend the life cycles of their fabs, must form investment collaborations, and have no choice but to fall behind.
I'm not so concerned with clock as I am with cores. I think an Itanium core is about 25 million transistors. So a 16 core Itanium with 32MB of on chip cache seems like it could happen. Sweet.
Itanium needs more cache. So it would be more like 4 cores with 48 MB of cache. A 3 billion transistor chip. If Intel can ship Montecito, a dual-core 2-way MT Itanium chip with 24 MB L3 in Summer of 2006, then I'd have more confidence in Intel shipping a 4 core 2-way MT Itanium chip in 2007, but who knows.
Intel has a 4 core Xeon chip with 16 MB L3 (Tigerton or resurrected Whitfield) and Montecito in the plans, both of them 1+ billion transistor processors. Sort of crazy, can't imagine the yields being good, but we'll see.
Itanium needs more cache. So it would be more like 4 cores with 48 MB of cache. A 3 billion transistor chip. If Intel can ship Montecito, a dual-core 2-way MT Itanium chip with 24 MB L3 in Summer of 2006, then I'd have more confidence in Intel shipping a 4 core 2-way MT Itanium chip in 2007, but who knows.
Intel has a 4 core Xeon chip with 16 MB L3 (Tigerton or resurrected Whitfield) and Montecito in the plans, both of them 1+ billion transistor processors. Sort of crazy, can't imagine the yields being good, but we'll see.
Tigerton is kind of a step back, and bummer though. Whitfield had an On Die Memory Controller, and now Tigerton does not. Now no one anticipates intel to have an O.D.M.C until 2009. This is the kind of thing that got Apple behind in it's expectations from IBM, and lead their relationship into trouble. First you have it, then you don't. I'm looking at stuff like this as a reason for Apple to use less expensive AMD quad cored Opteron MP (8 cores total) processors in the Xserve. and Pro Macintosh sooner rather than 2009. I mean why not?
Tigerton is kind of a step back, and bummer though. Whitfield had an On Die Memory Controller, and now Tigerton does not. Now no one anticipates intel to have an O.D.M.C until 2009. This is the kind of thing that got Apple behind in it's expectations from IBM, and lead their relationship into trouble. First you have it, then you don't. I'm looking at stuff like this as a reason for Apple to use less expensive AMD quad cored Opteron MP (8 cores total) processors in the Xserve. and Pro Macintosh sooner rather than 2009. I mean why not?
Having an on-chip memory controller isn't the be-all-and-end-all of chip design.
Having an on-chip memory controller isn't the be-all-and-end-all of chip design.
#1) You missed the point.
#2) I didn't say it was, but it is the direction intel is driving to, and if they are not getting it done in time they are falling behind.
#3) = #2) is the point.
[edit] #4AMD has been using ODMC for quite some time, and if I recall correctly. it was the first ODMC based AMD processors that put AMD ahead of intel in processor performance. They have not fallen behind intel sense.
[edit] #4AMD has been using ODMC for quite some time, and if I recall correctly. it was the first ODMC based AMD processors that put AMD ahead of intel in processor performance. They have not fallen behind intel sense.
Maybe not in the desktop market but AMD is behind in the notebook market. I think the later is the future of the money making.
They may be "behind" but that depends on how you define what's "forward".
If you define forward as low-power good-performance, then Intel is the king. If you define forward as 64-bit, low-power, good-performance, then AMD is the king.
Intel has the edge now because they just released their dual-core mobile processor. But AMD is expected to release their dual-core, 64-bit processor soon and then things kinda change.
So true, but I was really only considering, and mentioning AMD for XServes, and Pro Macintoshes. The Laptop processor tech lead will likely remain in the hands of intel.
Although I don't see how the laptop market could effect, or cross into the Workstation, or Server market in anyway. There is totally different reasoning behind the purchase of each for professionals. .
What is the 'name' for that AMD cpu that you are referring to, Gene Clean?
Yeah Gene, WTF is that clever AMD processor you refer to on the portable front? Turion64 (Project Taylor) with Yamato platform?
Current AMD Turion competed with Intel Pentium M. Now Intel Core Solo and Core Duo are out and have far better performance for about the same TDP.
The dualcore variant of AMD Turion64 (Taylor and its 30W TDP, succeeded by Trinidad with its 60W TDP, both targeted for 2006) will still be manufactured with a 90 nm process.
In this timeframe for laptops, and as pointed by bcwake, Intel will have after Yonah: Merom in mid-2006 and Gilo in early 2007 @ 65 nm, and Penryn @ 45nm in H2'2007.
On the desktop front, the first "K9" quadcore AMD, derivative from Athlon64 X2, will hit market no sooner than 2007 too. As for Intel the same year (Kentsfield, Cloverton and Tigerton).
Why are Fabs located in Israel? Is that a second silicon valley, or great tax advantages, or just coincidence?
[list=1][*]Availability of engineers (that's the silicon valley thing)[*]Government subsidy (bringing high-tech industry to unemployment-stricken Kiryat Gat)[*]Proximity to where the processors are designed (not sure that is a real advantage)[*]Slightly cheaper labor than in the US, but much closer culturally than China.[/list=1]
So we can have funny names for our processors, like Yonah and Merom. I'm hoping for a quad-core 45 nm Yitzhak.
Names like "Banias", "Dotan", "Yonah" and "Merom" are internal development names, just like Tiger or Longhorn.
All the processors designed in Haifa are named after mountain streams in the Galilee (no American or European would call these things rivers). I don't think there's one called Yitzhak.
So you think Apple will announce Merom Books at Paris? I'm going abroad to Japan at end of August, and I'll need a laptop. Or do you guys think I could get one for about the same price in Japan?
With the current yen/dollar rate, I believe that Apple computers are actually cheaper right now in Japan. But as you know the yen rate is always in flux. Can't forecast the price in August.
BTW, you will need to get used the Japanese keyboard if you buy one here (Japan).
Yeah Gene, WTF is that clever AMD processor you refer to on the portable front? Turion64 (Project Taylor) with Yamato platform?
It's called AMD Taylor.
Quote:
Current AMD Turion competed with Intel Pentium M.
And I was talking about a future chip, no?
Quote:
Now Intel Core Solo and Core Duo are out and have far better performance for about the same TDP.
Again, I was talking about the future Turion64.
Quote:
The dualcore variant of AMD Turion64 (Taylor and its 30W TDP, succeeded by Trinidad with its 60W TDP, both targeted for 2006) will still be manufactured with a 90 nm process.
You just forgot to tell us that they will migrate to a 65nm process soon after.
Quote:
AMD Taylor will first appear on IBM's 90nm Silicon on insulator (SOI) process, but will migrate to 65nm, likely with Silicon-Germanium (SiGe) stressed process which was recently achieved through the combined effort of IBM and AMD, with 40% improvement over comparable 65nm processes.
Quote:
On the desktop front, the first "K9" quadcore AMD, derivative from Athlon64 X2, will hit market no sooner than 2007 too. As for Intel the same year (Kentsfield, Cloverton and Tigerton).
Somehow I have a feeling that AMD will still kick Intel's ass in the desktop, and compete very nicely in the notebook arena. Regardless of Intel codenames and such.
With the current yen/dollar rate, I believe that Apple computers are actually cheaper right now in Japan. But as you know the yen rate is always in flux. Can't forecast the price in August.
BTW, you will need to get used the Japanese keyboard if you buy one here (Japan).
Oh shoot! I forgot about the keyboard! What's different about it? It's still QWERTY right?
Comments
do any one see Itaniums will be in XServe, that will be one hell of a server!
Originally posted by Zandros
Is it much harder to translate directly to 32 nm, or is this just a case of making more money by slowing development down?
It's much harder, doubly as expensive, and takes twice as long to develop.
Moore's so-called 2nd law for CMOS fabs is that the cost to develop them doubles every node. And remember that the so-called 1st law is that chip density doubles every node, with each node advance occuring every 18-24 months.
It is purely an economic equation. Essentially, in order to recoup the cost of developing a next generation fab, that fab has to sell twice as many chips over the lifespan of the fab. If the market can't support that, then there is no point in developing the fab.
Keep in mind that Intel is shipping 65 nm CPUs right now. Everyone else is 6 months behind them. Poor Freescale is a full node behind Intel. AMD won't be shipping until 2H 06. IBM likely in 2H 06 as well. Freescale won't have a 65 nm fab, but will share one like they are now with the 90 nm fab, which incidently, we haven't seen the fruits yet. They won't be jumping any node advances because it costs too much, takes too much time and don't have the market to support such a development.
The cost of developing a 65 nm? Probably $3 to $5 billion. The cost of developing a 45 nm? Probably $5 to $10 billion. The cost of moving directly from 65 nm to 32 nm? Probably $15 billion, 4+ years, and in the meanwhile, the 65 nm chip performance will stagnate for 2 years, leaving you at a horrible performance and market disadvantage for those two years. No company can survive that.
The other problem? AMD's 2005 revenue was about $5.5 billion. Intel's 2005 revenue was about $33 billion. This is why Intel is always first to the next nodes. They have the money and the market to get there while other companies have to extend the life cycles of their fabs, must form investment collaborations, and have no choice but to fall behind.
Originally posted by Splinemodel
I'm not so concerned with clock as I am with cores. I think an Itanium core is about 25 million transistors. So a 16 core Itanium with 32MB of on chip cache seems like it could happen. Sweet.
Itanium needs more cache. So it would be more like 4 cores with 48 MB of cache. A 3 billion transistor chip. If Intel can ship Montecito, a dual-core 2-way MT Itanium chip with 24 MB L3 in Summer of 2006, then I'd have more confidence in Intel shipping a 4 core 2-way MT Itanium chip in 2007, but who knows.
Intel has a 4 core Xeon chip with 16 MB L3 (Tigerton or resurrected Whitfield) and Montecito in the plans, both of them 1+ billion transistor processors. Sort of crazy, can't imagine the yields being good, but we'll see.
No wonder Jobs went to Intel.
Next up: Octa cores at 5 GHz w/ SSE Altima*
*Incorporating the altivec instruction set.
Oh but wait, IBM just released their 3 GHz PPC 970mp! Look, it's Freescale, check out their 200 watt 2.2 GHz dual core G4 - awww, isn't it cute?
Originally posted by msantti
Intel seems to be doing better with the 65nm process than they did with 90nm.
Only because expectations were reset at 90 nm.
Originally posted by THT
Itanium needs more cache. So it would be more like 4 cores with 48 MB of cache. A 3 billion transistor chip. If Intel can ship Montecito, a dual-core 2-way MT Itanium chip with 24 MB L3 in Summer of 2006, then I'd have more confidence in Intel shipping a 4 core 2-way MT Itanium chip in 2007, but who knows.
Intel has a 4 core Xeon chip with 16 MB L3 (Tigerton or resurrected Whitfield) and Montecito in the plans, both of them 1+ billion transistor processors. Sort of crazy, can't imagine the yields being good, but we'll see.
Tigerton is kind of a step back, and bummer though. Whitfield had an On Die Memory Controller, and now Tigerton does not. Now no one anticipates intel to have an O.D.M.C until 2009. This is the kind of thing that got Apple behind in it's expectations from IBM, and lead their relationship into trouble. First you have it, then you don't. I'm looking at stuff like this as a reason for Apple to use less expensive AMD quad cored Opteron MP (8 cores total) processors in the Xserve. and Pro Macintosh sooner rather than 2009. I mean why not?
Originally posted by onlooker
Tigerton is kind of a step back, and bummer though. Whitfield had an On Die Memory Controller, and now Tigerton does not. Now no one anticipates intel to have an O.D.M.C until 2009. This is the kind of thing that got Apple behind in it's expectations from IBM, and lead their relationship into trouble. First you have it, then you don't. I'm looking at stuff like this as a reason for Apple to use less expensive AMD quad cored Opteron MP (8 cores total) processors in the Xserve. and Pro Macintosh sooner rather than 2009. I mean why not?
Having an on-chip memory controller isn't the be-all-and-end-all of chip design.
Originally posted by Programmer
Having an on-chip memory controller isn't the be-all-and-end-all of chip design.
#1) You missed the point.
#2) I didn't say it was, but it is the direction intel is driving to, and if they are not getting it done in time they are falling behind.
#3) = #2) is the point.
[edit] #4 AMD has been using ODMC for quite some time, and if I recall correctly. it was the first ODMC based AMD processors that put AMD ahead of intel in processor performance. They have not fallen behind intel sense.
Originally posted by onlooker
[edit] #4 AMD has been using ODMC for quite some time, and if I recall correctly. it was the first ODMC based AMD processors that put AMD ahead of intel in processor performance. They have not fallen behind intel sense.
Maybe not in the desktop market but AMD is behind in the notebook market. I think the later is the future of the money making.
If you define forward as low-power good-performance, then Intel is the king. If you define forward as 64-bit, low-power, good-performance, then AMD is the king.
Intel has the edge now because they just released their dual-core mobile processor. But AMD is expected to release their dual-core, 64-bit processor soon and then things kinda change.
Although I don't see how the laptop market could effect, or cross into the Workstation, or Server market in anyway. There is totally different reasoning behind the purchase of each for professionals. .
Originally posted by NordicMan
What is the 'name' for that AMD cpu that you are referring to, Gene Clean?
Yeah Gene, WTF is that clever AMD processor you refer to on the portable front? Turion64 (Project Taylor) with Yamato platform?
Current AMD Turion competed with Intel Pentium M. Now Intel Core Solo and Core Duo are out and have far better performance for about the same TDP.
The dualcore variant of AMD Turion64 (Taylor and its 30W TDP, succeeded by Trinidad with its 60W TDP, both targeted for 2006) will still be manufactured with a 90 nm process.
In this timeframe for laptops, and as pointed by bcwake, Intel will have after Yonah: Merom in mid-2006 and Gilo in early 2007 @ 65 nm, and Penryn @ 45nm in H2'2007.
On the desktop front, the first "K9" quadcore AMD, derivative from Athlon64 X2, will hit market no sooner than 2007 too. As for Intel the same year (Kentsfield, Cloverton and Tigerton).
Originally posted by aplnub
Why are Fabs located in Israel? Is that a second silicon valley, or great tax advantages, or just coincidence?
[list=1][*]Availability of engineers (that's the silicon valley thing)[*]Government subsidy (bringing high-tech industry to unemployment-stricken Kiryat Gat)[*]Proximity to where the processors are designed (not sure that is a real advantage)[*]Slightly cheaper labor than in the US, but much closer culturally than China.[/list=1]
Originally posted by synp
Proximity to where the processors are designed (not sure that is a real advantage)
It is. It means design and manufacturing teams can work together very closely.
Originally posted by BRussell
So we can have funny names for our processors, like Yonah and Merom. I'm hoping for a quad-core 45 nm Yitzhak.
Names like "Banias", "Dotan", "Yonah" and "Merom" are internal development names, just like Tiger or Longhorn.
All the processors designed in Haifa are named after mountain streams in the Galilee (no American or European would call these things rivers). I don't think there's one called Yitzhak.
Originally posted by Hyde
So you think Apple will announce Merom Books at Paris? I'm going abroad to Japan at end of August, and I'll need a laptop. Or do you guys think I could get one for about the same price in Japan?
With the current yen/dollar rate, I believe that Apple computers are actually cheaper right now in Japan. But as you know the yen rate is always in flux. Can't forecast the price in August.
BTW, you will need to get used the Japanese keyboard if you buy one here (Japan).
Originally posted by Cosmos 1999
Yeah Gene, WTF is that clever AMD processor you refer to on the portable front? Turion64 (Project Taylor) with Yamato platform?
It's called AMD Taylor.
Current AMD Turion competed with Intel Pentium M.
And I was talking about a future chip, no?
Now Intel Core Solo and Core Duo are out and have far better performance for about the same TDP.
Again, I was talking about the future Turion64.
The dualcore variant of AMD Turion64 (Taylor and its 30W TDP, succeeded by Trinidad with its 60W TDP, both targeted for 2006) will still be manufactured with a 90 nm process.
You just forgot to tell us that they will migrate to a 65nm process soon after.
AMD Taylor will first appear on IBM's 90nm Silicon on insulator (SOI) process, but will migrate to 65nm, likely with Silicon-Germanium (SiGe) stressed process which was recently achieved through the combined effort of IBM and AMD, with 40% improvement over comparable 65nm processes.
On the desktop front, the first "K9" quadcore AMD, derivative from Athlon64 X2, will hit market no sooner than 2007 too. As for Intel the same year (Kentsfield, Cloverton and Tigerton).
Somehow I have a feeling that AMD will still kick Intel's ass in the desktop, and compete very nicely in the notebook arena. Regardless of Intel codenames and such.
Originally posted by BeigeUser
With the current yen/dollar rate, I believe that Apple computers are actually cheaper right now in Japan. But as you know the yen rate is always in flux. Can't forecast the price in August.
BTW, you will need to get used the Japanese keyboard if you buy one here (Japan).
Oh shoot! I forgot about the keyboard!