It is so sad what a good portion of Mac community has become - mindless Intel fanboys. Just with one simple keynote speech by Jobs.
This chip is getting laughed at across the x86 world...
Would it be possible to provide a link or 2 that supports your statement that "This chip is getting laughed at across the x86 world."? Hapless as I am I couldn't find anyone disparaging this chip, but my internet skills are quite lacking.
Would it be possible to provide a link or 2 that supports your statement that "This chip is getting laughed at across the x86 world."? Hapless as I am I couldn't find anyone disparaging this chip, but my internet skills are quite lacking.
I bet, he will have some troubles to back up his claims. The Yonah is a good chip, much better than the old G4 BTW.
That's said, I will consider updating my powerbook only when meroms chips will be avalaible. At equal clockspeed, the merom is supposed (nobody tested it yet) to be 30 % faster, and less power hungry.
So it would seem most likely in Apple laptops we'll be seeing single core parts or "L"/"U" part dual cores with the possibility of "T" parts on the high end. These clock speeds aren't really much better than what we're seeing with the G4s but the faster busses and, I would expect, better IPC would compensate greatly for that.
As an interesting comparison though, the 90nm PowerPC 970FX "desktop" CPU consumes about 25W max. The PowerPC 970FX low-power CPU consumes less than 14W at 1.4GHz and 16W at 1.6GHz. Yes, these are not dual core parts but they do not appear too shabby by comparison. I am not sure if the low power 970FX is 65nm or 90nm.
I may be one of the few who feels this way but after looking at the numbers I don't buy Apple's reasoning for switching to Intel. I feel they've switched for reasons other than performance per watt.
Would it be possible to provide a link or 2 that supports your statement that "This chip is getting laughed at across the x86 world."? Hapless as I am I couldn't find anyone disparaging this chip, but my internet skills are quite lacking.
This chip certainly hasn't been met without criticism. From what I've seen around the web it's about 50-50 for and against although with all the misinformation thrown in there it's hard to tell how many would hold the positions they do given the facts or if they set aside bias. Certainly the AMD fans aren't seeing this as much of a threat but there is no doubt it will best the Turion for now at least.
Yonah does appear to be a good CPU, Intel needed a good CPU with all the other overpriced/underpowered P4 variants they have been releasing in recent times. I think it's too much to call it a great CPU though, maybe Merom will do better but maybe it will also have better competition.
So it would seem most likely in Apple laptops we'll be seeing single core parts or "L"/"U" part dual cores with the possibility of "T" parts on the high end. These clock speeds aren't really much better than what we're seeing with the G4s but the faster busses and, I would expect, better IPC would compensate greatly for that.
As an interesting comparison though, the 90nm PowerPC 970FX "desktop" CPU consumes about 25W max. The PowerPC 970FX low-power CPU consumes less than 14W at 1.4GHz and 16W at 1.6GHz. Yes, these are not dual core parts but they do not appear too shabby by comparison. I am not sure if the low power 970FX is 65nm or 90nm.
I may be one of the few who feels this way but after looking at the numbers I don't buy Apple's reasoning for switching to Intel. I feel they've switched for reasons other than performance per watt.
I think that the main reason for switching was costs. The IBM roadmap was not that impressive, and IBM wanted Apple to participate at the R&D of the new chips. The PPC in the PC world (desktop or laptop) is a niche market, IBM did not wanted to support the cost of the R&D alone, and wanted some financial help from Apple.
In the contrary, there is no need for Apple to participate at the R&D of Intel chips. More Intel is ready to help Apple, via compilators, and has many other higt tech to offer. Apple love to use new tech in it's computers, at the difference of many PC builders who are very conservative.
Now the impressive Intel chips are not the Yonah, but rather the Merom and other conroe. I bet that we will have to wait until 2007 to see real stunning chips.
I think that the main reason for switching was costs. The IBM roadmap was not that impressive, and IBM wanted Apple to participate at the R&D of the new chips. The PPC in the PC world (desktop or laptop) is a niche market, IBM did not wanted to support the cost of the R&D alone, and wanted some financial help from Apple.
In the contrary, there is no need for Apple to participate at the R&D of Intel chips. More Intel is ready to help Apple, via compilators, and has many other higt tech to offer. Apple love to use new tech in it's computers, at the difference of many PC builders who are very conservative.
Now the impressive Intel chips are not the Yonah, but rather the Merom and other conroe. I bet that we will have to wait until 2007 to see real stunning chips.
I agree, in costs on many fronts. Intel will be a good partner for Apple. They already have their road map, and son't need Apple to design better chips. In the area of watts and performance, remember that Intel was at the bottom of that pile. Marginal performance and big watts ala net burst. Now Intel has Yonah which is designed for lap top computing and it is running neck and neck with the big boys. Pretty good flip, low power good performance. Now look at the lowly mini, put a dual core Yonah part in there and suddenly it is kicking some major ass. A mini with Balls! G4 what? Yes for the high end a Merom part will be great. You are correct in that Apple did not switch for performance per watt, they switched for the whole package, lower costs, better tech selection, less supply issues, and performance per watt. The fact that Intel is making all of the chips means that they can integrate them to save money enhance performance save power.
So it would seem most likely in Apple laptops we'll be seeing single core parts or "L"/"U" part dual cores with the possibility of "T" parts on the high end. These clock speeds aren't really much better than what we're seeing with the G4s but the faster busses and, I would expect, better IPC would compensate greatly for that.
The simple reality is that Intel will offer a 31 Watt TDP 65nm 2 GHz Yonah in Q1 06 while Freescale will be delivering ~40 Watt max 90nm 1.7 GHz 8641D in "Summer" of 06, and AMD will be delivering a ~30 Watt 65nm Turion in 2H 06.
The dual-core G4 is simply a non-starter against Yonah and Turion chips. It'll come at lower GHz, be more expensive, higher wattage, and lower performance.
Quote:
As an interesting comparison though, the 90nm PowerPC 970FX "desktop" CPU consumes about 25W max. The PowerPC 970FX low-power CPU consumes less than 14W at 1.4GHz and 16W at 1.6GHz. Yes, these are not dual core parts but they do not appear too shabby by comparison. I am not sure if the low power 970FX is 65nm or 90nm.
The low power 970fx parts are 90nm parts. A 1.6 GHz 970fx "power optimized" part is 21 Watts max. Very nice. The huge problem is that Intel will have a dual core Yonah at 1.66 GHz running 24 Watts TDP in Q1 06.
Other big problem is that the 2 GHz 970fx "power optimized" part is 50 Watts max while a 2 GHz Yonah is at 31 Watts TDP. That's almost 4x the integer performance per watt.
IBM also won't be producing 65 nm parts until Summer 06 as well.
Quote:
I may be one of the few who feels this way but after looking at the numbers I don't buy Apple's reasoning for switching to Intel. I feel they've switched for reasons other than performance per watt.
Yeah sure undoubtedly. I'm sure they switched for multiple reasons. But it was the correct decision, and Intel will be producing some tremendously low power processors by 2007 and some huge processors in 2008. By 2007 they'll have a dedicated ultra-low power 65 fab (P1265) which could produce a <1 Watt max 1 to 1.4 GHz Yonah.
So, there is some truth to the performance/watt reasons.
Yeah sure undoubtedly. I'm sure they switched for multiple reasons. But it was the correct decision, and Intel will be producing some tremendously low power processors by 2007 and some huge processors in 2008. By 2007 they'll have a dedicated ultra-low power 65 fab (P1265) which could produce a <1 Watt max 1 to 1.4 GHz Yonah.
So, there is some truth to the performance/watt reasons.
Heh, well I guess I just have less faith in Intel's ability to keep to their roadmap as they never have managed in the past. They have never significantly outperformed their competition in performance and I'd have trouble believing they are suddenly going to manage it now. I'm sure their roadmaps didn't show them significantly behind AMD on the desktop as they have been for quite a long time now. AMD and IBM certainly aren't going to just sit still waiting for Intel to catch up performance-wise.
I'm not claiming IBM could have worked miracles here but as far as roadmaps go, it seems Apple just believed Intel's hype instead of IBM's hype (or just plain didn't care).
Certainly though, the cost arguement would seem to make sense. Intel as a supplier certainly seems to be the cheaper and more reliable path.
Keep in mind Yonah is a stop-gap until Merom. Yonah is simply a extension of the very mature Dothan microarchitecture. Merom is the first chip of Intel's next generation micorachitecture and is rumored to clock similarly to Yonah but offer over 50% greater performance per watt. Merom is also a 64bit processor while Yonah is not.
Memrom should be here by this time next year and Yonah will be relegated to iBooks and Mac Minis.
Heh, well I guess I just have less faith in Intel's ability to keep to their roadmap as they never have managed in the past. They have never significantly outperformed their competition in performance and I'd have trouble believing they are suddenly going to manage it now.
Please explain to me why Intel is the dominant semiconductor manufacturer in the world? (And dominant by a very very very wide margin).
It's not luck.
Quote:
I'm sure their roadmaps didn't show them significantly behind AMD on the desktop as they have been for quite a long time now. AMD and IBM certainly aren't going to just sit still waiting for Intel to catch up performance-wise.
Prescott bombed true, and AMD is ahead for desktop performance. But it's really beside the point. It's first and foremost about the fabs. I'd almost say nothing else matters. The design just needs to be good enough or competitive. The rest, the fabs will take care of.
Intel is the only company that will have the money and the market to move on to 45 nm and 32 nm on its own. While there may be 2 or 3 other consortiums able to get to 45 nm (likely a year behind Intel), there may be no other company able to afford to get to 32 nm. Perhaps Samsung if they can own flash, but Intel will be competing there too.
You're talking about reducing the TDP by a factor of TEN. That's not going to happen. Maybe a 1.4GHz XScale under 1W, though.
Yeah a factor of ten. I was thinking I was being generous with 1 W. Intel's P1265 has demonstrated a reduction in transistor leakage by a factor of 1000, a mere 3 orders of magnitude. That's huge. It'll also translate into x86 handtops.
Please explain to me why Intel is the dominant semiconductor manufacturer in the world? (And dominant by a very very very wide margin).
It's not luck.
Actually I'd argue it was to an extent luck. Whoever IBM picked for the original PC would most likely be in Intel's position today. It could have easily been Motorola or Zilog at least if not a handful of other companies. No point digging up the past but anyone who remembers the 8080 or 8086 knows Intel were never leading performance-wise, even then complete chipsets and reliable production were their biggest assets. They have at least been consistent in that regard.
Quote:
Originally posted by THT Prescott bombed true, and AMD is ahead for desktop performance. But it's really beside the point. It's first and foremost about the fabs. I'd almost say nothing else matters. The design just needs to be good enough or competitive. The rest, the fabs will take care of.
Intel is the only company that will have the money and the market to move on to 45 nm and 32 nm on its own. While there may be 2 or 3 other consortiums able to get to 45 nm (likely a year behind Intel), there may be no other company able to afford to get to 32 nm. Perhaps Samsung if they can own flash, but Intel will be competing there too. [/B]
Certainly that's where the move to Intel makes the most sense assuming they don't run into problems with these transitions. I just wish we didn't have to move to the x86 architechure to get it. Last I'd heard about 30% of the transistors on a x86 die were overheads for keeping compatibility with that architecture. No idea where I heard that statement though so I have no way of backing it up. The programmer/hardware designer side of me (Yes, I did actually study the field) sees the PPC as ultimately a promising architecture for the long term.
I'm not going to be silly about it though, this is ultimately the path Apple has chosen. I intend to buy the 12"/13" Intel Powerbook when it's released as a transition machine (my iBook is due for an update anyway) and upgrade my G5 desktop to an Intel based Mac in another 2-3 years as needs require.
Actually I'd argue it was to an extent luck. Whoever IBM picked for the original PC would most likely be in Intel's position today. It could have easily been Motorola or Zilog at least if not a handful of other companies. No point digging up the past but anyone who remembers the 8080 or 8086 knows Intel were never leading performance-wise, even then complete chipsets and reliable production were their biggest assets. They have at least been consistent in that regard.
Not quite. There were many x86 CPU vendors in the past, many x86 CPU vendors now, and there will be x86 vendors in the future. If Intel has never managed its roadmap well or has never significantly outperformed their x86 competition, their x86 competition would be in much better shape today and Intel at much lower marketshare and perhaps even out of business.
There is a reason why those other x86 clone vendors were destroyed by Intel. Intel came up with good designs and maintained a good roadmap. These weren't small fry companies either. IBM/Cyrix tried. Texas Instruments tried. AMD is still chugging along. Transmeta was supposed take over the mobile x86 world. Via, Winchip, C&T, IDT, NEC, NextGen. There is a much longer list of x86 vendors than PPC vendors.
Quote:
Last I'd heard about 30% of the transistors on a x86 die were overheads for keeping compatibility with that architecture.
Um. You really heard wrong. Perhaps it was some comparison between the Pentium Pro and RISC processors in 1995 and the so-called overhead of translated x86 instructions to RISC-like micro-ops. Today, 10 years later, it's irrelevant. I'm not sure it was really relevant in 1995. In fact, it looks like it wasn't since the Pentium Pro and its successors (P6 microarchitecture CPUs) effectively killed PA-RISC, MIPS, Alpha, SPARC and PowerPC (except for Apple) in the workstation market.
Not quite. There were many x86 CPU vendors in the past, many x86 CPU vendors now, and there will be x86 vendors in the future. If Intel has never managed its roadmap well or has never significantly outperformed their x86 competition, their x86 competition would be in much better shape today and Intel at much lower marketshare and perhaps even out of business.
If you are talking only about the x86 market then it's not much of a surprise that Intel knows the archtechture best. There have always been more promising CPU families than x86 though but Intel has had a lot more cash to spend on R&D. It has gone through some truly asounding transformations and I'm sure will again but as the saying goes, "you can?t make a silk purse from a sow?s ear". Thankfully today the CPU architecture doesn't mean as much as it used to.
Quote:
Originally posted by THT
Um. You really heard wrong. Perhaps it was some comparison between the Pentium Pro and RISC processors in 1995 and the so-called overhead of translated x86 instructions to RISC-like micro-ops. Today, 10 years later, it's irrelevant. I'm not sure it was really relevant in 1995. In fact, it looks like it wasn't since the Pentium Pro and its successors (P6 microarchitecture CPUs) effectively killed PA-RISC, MIPS, Alpha, SPARC and PowerPC (except for Apple) in the workstation market.
Thinking about it, what you say there would make a lot of sense as the number of transistors required for translation would be relatively static so starting at 30% and going down every generation since would make it rather insignificant these days. That said, the PA-RISC, Alpha and SPARC were largely killed by their creators, Intel did not have much to do with it directly. Intel certainly didn't beat the likes of the Alpha in performance until long after the Alphas were canned. The amount of funding Intel has been able to put into things has increased the x86 lifespan too. On top of that, the amount of effort that has gone into optimising compilers for the x86 architecture is also significant. There's no denying Intel has done what they needed to do.
No point digging up the past but anyone who remembers the 8080 or 8086 knows Intel were never leading performance-wise, even then complete chipsets and reliable production were their biggest assets. They have at least been consistent in that regard.
Never underestimate the value of being able to consistently deliver. It means a lot and very few companies have been able to follow Intel in that regard. Their manufacturing has been so strong as to deliver them better margins than their competition and the ability to be massively competitive on price. That way even when they don't necessarily have the best design you'll always have a lot of trouble displacing them.
Full credit to them, they do exactly what they need to. It is a credit to them that such a large company turned themselves around so quickly after the screw up that was Prescott too.
People should also keep in mind Yonah is still just a laptop chip. It's an improvement but it was still designed as a laptop chip.
Pretty impressive roadmap and developments at intel.
Plans are underway for 45 nm processors and up to 8-core cpu's.
Go intel!
That site says that only the dual core Yonah will be available in early 2006. Does that mean that whatever Apple gives us at Macworld (intel wise) will use this chip?
Comments
Originally posted by Roadmap
It is so sad what a good portion of Mac community has become - mindless Intel fanboys. Just with one simple keynote speech by Jobs.
This chip is getting laughed at across the x86 world...
Wanker!
Originally posted by Roadmap
It is so sad what a good portion of Mac community has become - mindless Intel fanboys. Just with one simple keynote speech by Jobs.
This chip is getting laughed at across the x86 world...
Would it be possible to provide a link or 2 that supports your statement that "This chip is getting laughed at across the x86 world."? Hapless as I am I couldn't find anyone disparaging this chip, but my internet skills are quite lacking.
Originally posted by rickag
Would it be possible to provide a link or 2 that supports your statement that "This chip is getting laughed at across the x86 world."? Hapless as I am I couldn't find anyone disparaging this chip, but my internet skills are quite lacking.
I bet, he will have some troubles to back up his claims. The Yonah is a good chip, much better than the old G4 BTW.
That's said, I will consider updating my powerbook only when meroms chips will be avalaible. At equal clockspeed, the merom is supposed (nobody tested it yet) to be 30 % faster, and less power hungry.
As an interesting comparison though, the 90nm PowerPC 970FX "desktop" CPU consumes about 25W max. The PowerPC 970FX low-power CPU consumes less than 14W at 1.4GHz and 16W at 1.6GHz. Yes, these are not dual core parts but they do not appear too shabby by comparison. I am not sure if the low power 970FX is 65nm or 90nm.
I may be one of the few who feels this way but after looking at the numbers I don't buy Apple's reasoning for switching to Intel. I feel they've switched for reasons other than performance per watt.
Originally posted by rickag
Would it be possible to provide a link or 2 that supports your statement that "This chip is getting laughed at across the x86 world."? Hapless as I am I couldn't find anyone disparaging this chip, but my internet skills are quite lacking.
This chip certainly hasn't been met without criticism. From what I've seen around the web it's about 50-50 for and against although with all the misinformation thrown in there it's hard to tell how many would hold the positions they do given the facts or if they set aside bias. Certainly the AMD fans aren't seeing this as much of a threat but there is no doubt it will best the Turion for now at least.
Yonah does appear to be a good CPU, Intel needed a good CPU with all the other overpriced/underpowered P4 variants they have been releasing in recent times. I think it's too much to call it a great CPU though, maybe Merom will do better but maybe it will also have better competition.
Originally posted by MattB
So it would seem most likely in Apple laptops we'll be seeing single core parts or "L"/"U" part dual cores with the possibility of "T" parts on the high end. These clock speeds aren't really much better than what we're seeing with the G4s but the faster busses and, I would expect, better IPC would compensate greatly for that.
As an interesting comparison though, the 90nm PowerPC 970FX "desktop" CPU consumes about 25W max. The PowerPC 970FX low-power CPU consumes less than 14W at 1.4GHz and 16W at 1.6GHz. Yes, these are not dual core parts but they do not appear too shabby by comparison. I am not sure if the low power 970FX is 65nm or 90nm.
I may be one of the few who feels this way but after looking at the numbers I don't buy Apple's reasoning for switching to Intel. I feel they've switched for reasons other than performance per watt.
I think that the main reason for switching was costs. The IBM roadmap was not that impressive, and IBM wanted Apple to participate at the R&D of the new chips. The PPC in the PC world (desktop or laptop) is a niche market, IBM did not wanted to support the cost of the R&D alone, and wanted some financial help from Apple.
In the contrary, there is no need for Apple to participate at the R&D of Intel chips. More Intel is ready to help Apple, via compilators, and has many other higt tech to offer. Apple love to use new tech in it's computers, at the difference of many PC builders who are very conservative.
Now the impressive Intel chips are not the Yonah, but rather the Merom and other conroe. I bet that we will have to wait until 2007 to see real stunning chips.
Originally posted by Powerdoc
I think that the main reason for switching was costs. The IBM roadmap was not that impressive, and IBM wanted Apple to participate at the R&D of the new chips. The PPC in the PC world (desktop or laptop) is a niche market, IBM did not wanted to support the cost of the R&D alone, and wanted some financial help from Apple.
In the contrary, there is no need for Apple to participate at the R&D of Intel chips. More Intel is ready to help Apple, via compilators, and has many other higt tech to offer. Apple love to use new tech in it's computers, at the difference of many PC builders who are very conservative.
Now the impressive Intel chips are not the Yonah, but rather the Merom and other conroe. I bet that we will have to wait until 2007 to see real stunning chips.
I agree, in costs on many fronts. Intel will be a good partner for Apple. They already have their road map, and son't need Apple to design better chips. In the area of watts and performance, remember that Intel was at the bottom of that pile. Marginal performance and big watts ala net burst. Now Intel has Yonah which is designed for lap top computing and it is running neck and neck with the big boys. Pretty good flip, low power good performance. Now look at the lowly mini, put a dual core Yonah part in there and suddenly it is kicking some major ass. A mini with Balls! G4 what? Yes for the high end a Merom part will be great. You are correct in that Apple did not switch for performance per watt, they switched for the whole package, lower costs, better tech selection, less supply issues, and performance per watt. The fact that Intel is making all of the chips means that they can integrate them to save money enhance performance save power.
Originally posted by MattB
So it would seem most likely in Apple laptops we'll be seeing single core parts or "L"/"U" part dual cores with the possibility of "T" parts on the high end. These clock speeds aren't really much better than what we're seeing with the G4s but the faster busses and, I would expect, better IPC would compensate greatly for that.
The simple reality is that Intel will offer a 31 Watt TDP 65nm 2 GHz Yonah in Q1 06 while Freescale will be delivering ~40 Watt max 90nm 1.7 GHz 8641D in "Summer" of 06, and AMD will be delivering a ~30 Watt 65nm Turion in 2H 06.
The dual-core G4 is simply a non-starter against Yonah and Turion chips. It'll come at lower GHz, be more expensive, higher wattage, and lower performance.
As an interesting comparison though, the 90nm PowerPC 970FX "desktop" CPU consumes about 25W max. The PowerPC 970FX low-power CPU consumes less than 14W at 1.4GHz and 16W at 1.6GHz. Yes, these are not dual core parts but they do not appear too shabby by comparison. I am not sure if the low power 970FX is 65nm or 90nm.
The low power 970fx parts are 90nm parts. A 1.6 GHz 970fx "power optimized" part is 21 Watts max. Very nice. The huge problem is that Intel will have a dual core Yonah at 1.66 GHz running 24 Watts TDP in Q1 06.
Other big problem is that the 2 GHz 970fx "power optimized" part is 50 Watts max while a 2 GHz Yonah is at 31 Watts TDP. That's almost 4x the integer performance per watt.
IBM also won't be producing 65 nm parts until Summer 06 as well.
I may be one of the few who feels this way but after looking at the numbers I don't buy Apple's reasoning for switching to Intel. I feel they've switched for reasons other than performance per watt.
Yeah sure undoubtedly. I'm sure they switched for multiple reasons. But it was the correct decision, and Intel will be producing some tremendously low power processors by 2007 and some huge processors in 2008. By 2007 they'll have a dedicated ultra-low power 65 fab (P1265) which could produce a <1 Watt max 1 to 1.4 GHz Yonah.
So, there is some truth to the performance/watt reasons.
Originally posted by THT
By 2007 they'll have a dedicated ultra-low power 65 fab (P1265) which could produce a <1 Watt max 1 to 1.4 GHz Yonah.
So, there is some truth to the performance/watt reasons.
Whoa, less than one watt? That's crazy! That's like, the equivalent to a sleeping Mac of today!
Originally posted by THT
Yeah sure undoubtedly. I'm sure they switched for multiple reasons. But it was the correct decision, and Intel will be producing some tremendously low power processors by 2007 and some huge processors in 2008. By 2007 they'll have a dedicated ultra-low power 65 fab (P1265) which could produce a <1 Watt max 1 to 1.4 GHz Yonah.
So, there is some truth to the performance/watt reasons.
Heh, well I guess I just have less faith in Intel's ability to keep to their roadmap as they never have managed in the past. They have never significantly outperformed their competition in performance and I'd have trouble believing they are suddenly going to manage it now. I'm sure their roadmaps didn't show them significantly behind AMD on the desktop as they have been for quite a long time now. AMD and IBM certainly aren't going to just sit still waiting for Intel to catch up performance-wise.
I'm not claiming IBM could have worked miracles here but as far as roadmaps go, it seems Apple just believed Intel's hype instead of IBM's hype (or just plain didn't care).
Certainly though, the cost arguement would seem to make sense. Intel as a supplier certainly seems to be the cheaper and more reliable path.
Originally posted by THT
By 2007 they'll have a dedicated ultra-low power 65 fab (P1265) which could produce a <1 Watt max 1 to 1.4 GHz Yonah.
You're talking about reducing the TDP by a factor of TEN. That's not going to happen. Maybe a 1.4GHz XScale under 1W, though.
Memrom should be here by this time next year and Yonah will be relegated to iBooks and Mac Minis.
Originally posted by MattB
Heh, well I guess I just have less faith in Intel's ability to keep to their roadmap as they never have managed in the past. They have never significantly outperformed their competition in performance and I'd have trouble believing they are suddenly going to manage it now.
Please explain to me why Intel is the dominant semiconductor manufacturer in the world? (And dominant by a very very very wide margin).
It's not luck.
I'm sure their roadmaps didn't show them significantly behind AMD on the desktop as they have been for quite a long time now. AMD and IBM certainly aren't going to just sit still waiting for Intel to catch up performance-wise.
Prescott bombed true, and AMD is ahead for desktop performance. But it's really beside the point. It's first and foremost about the fabs. I'd almost say nothing else matters. The design just needs to be good enough or competitive. The rest, the fabs will take care of.
Intel is the only company that will have the money and the market to move on to 45 nm and 32 nm on its own. While there may be 2 or 3 other consortiums able to get to 45 nm (likely a year behind Intel), there may be no other company able to afford to get to 32 nm. Perhaps Samsung if they can own flash, but Intel will be competing there too.
Originally posted by wmf
You're talking about reducing the TDP by a factor of TEN. That's not going to happen. Maybe a 1.4GHz XScale under 1W, though.
Yeah a factor of ten. I was thinking I was being generous with 1 W. Intel's P1265 has demonstrated a reduction in transistor leakage by a factor of 1000, a mere 3 orders of magnitude. That's huge. It'll also translate into x86 handtops.
Originally posted by THT
Please explain to me why Intel is the dominant semiconductor manufacturer in the world? (And dominant by a very very very wide margin).
It's not luck.
Actually I'd argue it was to an extent luck. Whoever IBM picked for the original PC would most likely be in Intel's position today. It could have easily been Motorola or Zilog at least if not a handful of other companies. No point digging up the past but anyone who remembers the 8080 or 8086 knows Intel were never leading performance-wise, even then complete chipsets and reliable production were their biggest assets. They have at least been consistent in that regard.
Originally posted by THT Prescott bombed true, and AMD is ahead for desktop performance. But it's really beside the point. It's first and foremost about the fabs. I'd almost say nothing else matters. The design just needs to be good enough or competitive. The rest, the fabs will take care of.
Intel is the only company that will have the money and the market to move on to 45 nm and 32 nm on its own. While there may be 2 or 3 other consortiums able to get to 45 nm (likely a year behind Intel), there may be no other company able to afford to get to 32 nm. Perhaps Samsung if they can own flash, but Intel will be competing there too. [/B]
Certainly that's where the move to Intel makes the most sense assuming they don't run into problems with these transitions. I just wish we didn't have to move to the x86 architechure to get it. Last I'd heard about 30% of the transistors on a x86 die were overheads for keeping compatibility with that architecture. No idea where I heard that statement though so I have no way of backing it up. The programmer/hardware designer side of me (Yes, I did actually study the field) sees the PPC as ultimately a promising architecture for the long term.
I'm not going to be silly about it though, this is ultimately the path Apple has chosen. I intend to buy the 12"/13" Intel Powerbook when it's released as a transition machine (my iBook is due for an update anyway) and upgrade my G5 desktop to an Intel based Mac in another 2-3 years as needs require.
Originally posted by MattB
Actually I'd argue it was to an extent luck. Whoever IBM picked for the original PC would most likely be in Intel's position today. It could have easily been Motorola or Zilog at least if not a handful of other companies. No point digging up the past but anyone who remembers the 8080 or 8086 knows Intel were never leading performance-wise, even then complete chipsets and reliable production were their biggest assets. They have at least been consistent in that regard.
Not quite. There were many x86 CPU vendors in the past, many x86 CPU vendors now, and there will be x86 vendors in the future. If Intel has never managed its roadmap well or has never significantly outperformed their x86 competition, their x86 competition would be in much better shape today and Intel at much lower marketshare and perhaps even out of business.
There is a reason why those other x86 clone vendors were destroyed by Intel. Intel came up with good designs and maintained a good roadmap. These weren't small fry companies either. IBM/Cyrix tried. Texas Instruments tried. AMD is still chugging along. Transmeta was supposed take over the mobile x86 world. Via, Winchip, C&T, IDT, NEC, NextGen. There is a much longer list of x86 vendors than PPC vendors.
Last I'd heard about 30% of the transistors on a x86 die were overheads for keeping compatibility with that architecture.
Um. You really heard wrong. Perhaps it was some comparison between the Pentium Pro and RISC processors in 1995 and the so-called overhead of translated x86 instructions to RISC-like micro-ops. Today, 10 years later, it's irrelevant. I'm not sure it was really relevant in 1995. In fact, it looks like it wasn't since the Pentium Pro and its successors (P6 microarchitecture CPUs) effectively killed PA-RISC, MIPS, Alpha, SPARC and PowerPC (except for Apple) in the workstation market.
Originally posted by THT
Not quite. There were many x86 CPU vendors in the past, many x86 CPU vendors now, and there will be x86 vendors in the future. If Intel has never managed its roadmap well or has never significantly outperformed their x86 competition, their x86 competition would be in much better shape today and Intel at much lower marketshare and perhaps even out of business.
If you are talking only about the x86 market then it's not much of a surprise that Intel knows the archtechture best. There have always been more promising CPU families than x86 though but Intel has had a lot more cash to spend on R&D. It has gone through some truly asounding transformations and I'm sure will again but as the saying goes, "you can?t make a silk purse from a sow?s ear". Thankfully today the CPU architecture doesn't mean as much as it used to.
Originally posted by THT
Um. You really heard wrong. Perhaps it was some comparison between the Pentium Pro and RISC processors in 1995 and the so-called overhead of translated x86 instructions to RISC-like micro-ops. Today, 10 years later, it's irrelevant. I'm not sure it was really relevant in 1995. In fact, it looks like it wasn't since the Pentium Pro and its successors (P6 microarchitecture CPUs) effectively killed PA-RISC, MIPS, Alpha, SPARC and PowerPC (except for Apple) in the workstation market.
Thinking about it, what you say there would make a lot of sense as the number of transistors required for translation would be relatively static so starting at 30% and going down every generation since would make it rather insignificant these days. That said, the PA-RISC, Alpha and SPARC were largely killed by their creators, Intel did not have much to do with it directly. Intel certainly didn't beat the likes of the Alpha in performance until long after the Alphas were canned. The amount of funding Intel has been able to put into things has increased the x86 lifespan too. On top of that, the amount of effort that has gone into optimising compilers for the x86 architecture is also significant. There's no denying Intel has done what they needed to do.
Originally posted by MattB
No point digging up the past but anyone who remembers the 8080 or 8086 knows Intel were never leading performance-wise, even then complete chipsets and reliable production were their biggest assets. They have at least been consistent in that regard.
Never underestimate the value of being able to consistently deliver. It means a lot and very few companies have been able to follow Intel in that regard. Their manufacturing has been so strong as to deliver them better margins than their competition and the ability to be massively competitive on price. That way even when they don't necessarily have the best design you'll always have a lot of trouble displacing them.
Full credit to them, they do exactly what they need to. It is a credit to them that such a large company turned themselves around so quickly after the screw up that was Prescott too.
People should also keep in mind Yonah is still just a laptop chip. It's an improvement but it was still designed as a laptop chip.
http://www.tomshardware.com/cpu/20051203/index.html
Pretty impressive roadmap and developments at intel.
Plans are underway for 45 nm processors and up to 8-core cpu's.
Go intel!
Originally posted by dutch pear
I just cam across this article on Tomshardware:
http://www.tomshardware.com/cpu/20051203/index.html
Pretty impressive roadmap and developments at intel.
Plans are underway for 45 nm processors and up to 8-core cpu's.
Go intel!
That site says that only the dual core Yonah will be available in early 2006. Does that mean that whatever Apple gives us at Macworld (intel wise) will use this chip?