Your arguments seems convincing, but why Apple did not choose the cell architecture ?. If my memory is correct, Apple considered this choice.
IBM had a different market in mind, the benefits of cell are yet to be proven. I will say it again the shift to Intel was as much about the other chips as it was about the CPU. So what would Apple have at the end of the day with Cell, a great processor that is difficult to program for at best, IBM will sell you the tools but forget GCC support, so either Apple pays IBM to distribute their compilier in Xtools, or they roll their own which would take great efforts and time. They also would still have the problem of all of the other chips, Intel makes much more than CPUs, and they make them in packages designed for various target markets. Let cell prove itself in the market and then let Intel steal the idea, and make their version and while Apple is waiting they get to use all of those other support chips and Macs will not be any different than anything else on the market. This problem does not require much brain effort to figure out why Apple did what they did.
Your arguments seems convincing, but why Apple did not choose the cell architecture ?. If my memory is correct, Apple considered this choice.
We don't know what Intel's plans look like, but Apple will have gotten a better look at them. Intel also rules in terms of process technology, and Apple can be assured that Intel will build chips that are extremely competitive with the bulk of the laptop and desktop PC markets (duh!). IBM is interested in other things.
This doesn't mean that future Intel chips won't be Cell-like in nature, however. The challenges of what to do with more transistors and how to deal with the realities of physics are invariant across companies.
If Intel does not, for whatever reason, supply the processors that Apple requires there is a simple, if expensive solution. Apple could buy AMD and for the first time ever be in total control of its chip future.
Apple has a Market Cap of about $50 billion against AMD about $10 Billion US Dollars. Apple could easily afford a cash and script offer for AMD and still have several billion dollars in the bank and little to no debt.
They could also very quickly switch processors over to AMD.
I am not saying it is a good idea, but it could be the ultimate backup plan.
Your arguments seems convincing, but why Apple did not choose the cell architecture ?. If my memory is correct, Apple considered this choice.
Single threaded integer performance on Cell would be worse than any of the contemporary Intel offerings. Merom cores also appear to be less wattage than contemporary PPU cores.
Also, Intel will have a special, dedicated low-wattage 65-nm fab in 2007, in addition to their high performance fabs. 1.5 to 2 GHz Merom-derivatives on that process will probably be <5 Watts TDP (while the high-perf fabs produce 2 to 3 GHz Merom-derivatives). That means computers with no fans, and we know who is a big fan of that.
I wouldn't be surprised if this low-watt 65nm fab from Intel will be unique to the industry.
Is it the Apple CPU Maker Curse? It's beginning to look like it!
Not likely
Quote:
This should worry folks at Dell, HP and SGI, as they're most vulnerable to Intel's shortcomings.
Dell only need compare their DL380 sales versus the DL385(AMD) to see that despite the AMD product being faster the Intel sales dwarf AMDs.
Quote:
"If Intel is ever to reclaim the performance lead from AMD, it must make the transition to an on-chip memory controller," he wrote
Another juicy quote from Nathan Brockwood who I've never ready anything particularly "insightful" from.
Correct me if I'm wrong and don't hesitate to slam me. I'm a big boy and I can take it. Most of the reviews I've seen comparing Opterons to Xeons show that the Opty's have an advantage with memory latency versus the Xeon but in actual throughput numbers that advantage disappears.
So it seems to me that if you have an app that is latency sensitive like a database then the Opteron system should benchmark better but if you have an app that craves massive bandwidth then you should see most of the performance delta disappear.
If latency is the only problem seperating an external controller from an ondie controller then Intel likely knows that the migration from DDRII to FB-DIMM will ameliorate the latency issue since theoretically the serialized nature and design of FB-DIMM minimizes latency.
I think it's foolish for anyone to claim that Intel cannot reclaim a performance lead unless they've seen Merom running in a lab. Wanna laugh...go read some Analyst posts from 4 years ago and see how accurate they were. They would have told you that we'd be using 5Ghz procs by now.
That's FUD. Opterons destroy Xeons in any benchmark you name. Dell doesn't sell any AMD machines, but it isn't because the Intels are better, it's because Intel basically pays them not to sell AMD. Anyone who thinks there's a level playing field out there has his head in the sand.
Intel's processors are technically inferior to AMD's - this is an established fact. Intel's business practices are questionable if not blatantly unethical, anticompetitive and monopolistic - again, an established fact.
Intel may have cooked benchmarks that say that in some particular cases Xeons are faster. And they claim that they haven't done anything that other monop- I mean companies have done in the past. So what? It's all a bunch of crap. Give it up.
And guess what? None of that matters. Apple wants to join the mainstream, to take the 'hardware superiority' issue off the table.
Anyone who cares about raw performance hasn't been buying Macs anyway, and they won't buy them when they're Intel-based either.
That's FUD. Opterons destroy Xeons in any benchmark you name
How about revenue generated
Quote:
Anyone who cares about raw performance hasn't been buying Macs anyway, and they won't buy them when they're Intel-based either.
Evidently they aren't buying Opteron's either in large enough quantities. My point isn't that the Opteron isn't a superior chip it's that the collective response from many companies is "Meh"
So by that benchmark, Windows is better than OS X and Dell makes better computers than, say, Apple?
'Cause if you look at revenue..
Quote:
Evidently they aren't buying Opteron's either in large enough quantities. My point isn't that the Opteron isn't a superior chip it's that the collective response from many companies is "Meh"
I don't know about Opterons, but some companies aren't saying "Meh".
That's FUD. Opterons destroy Xeons in any benchmark you name. Dell doesn't sell any AMD machines, but it isn't because the Intels are better, it's because Intel basically pays them not to sell AMD. Anyone who thinks there's a level playing field out there has his head in the sand.
Intel's processors are technically inferior to AMD's - this is an established fact. Intel's business practices are questionable if not blatantly unethical, anticompetitive and monopolistic - again, an established fact.
Intel may have cooked benchmarks that say that in some particular cases Xeons are faster. And they claim that they haven't done anything that other monop- I mean companies have done in the past. So what? It's all a bunch of crap. Give it up.
And guess what? None of that matters. Apple wants to join the mainstream, to take the 'hardware superiority' issue off the table.
Anyone who cares about raw performance hasn't been buying Macs anyway, and they won't buy them when they're Intel-based either.
Relax, there is one underlying truth, Apple could have chosen the best CPU, and if that were the case Cell may have been the winner, in switiching to x86 Apple could have also chosen AMD, they did not. As we speak, Intel is shifting the line-up and pulling better performers further ahead to market. Intel has and is making support chips that integrate and compliment nicely. The switch to Intel was made in the face of Cell and AMD dominance in performance areas. Remember what was said performance per watt, Intel will lead, in fact Intel will dominate. Currently no, but look at how much the land scape has changed in the latter half of '05, '06 will only accelerate this.
Relax, there is one underlying truth, Apple could have chosen the best CPU, and if that were the case Cell may have been the winner, in switiching to x86 Apple could have also chosen AMD, they did not. As we speak, Intel is shifting the line-up and pulling better performers further ahead to market. Intel has and is making support chips that integrate and compliment nicely. The switch to Intel was made in the face of Cell and AMD dominance in performance areas. Remember what was said performance per watt, Intel will lead, in fact Intel will dominate. Currently no, but look at how much the land scape has changed in the latter half of '05, '06 will only accelerate this.
"Best CPU" is a multi-dimensional problem and without specifying all of the criteria used to evaluate it, you cannot determine which is the best. Apple is very much concerned with low power laptop (and compact machine) processors, and they chose the only company that is certainly going to be:
a) In posession of the best process technology in the world going forward.
b) Building low power high performance processors appropriate for use in laptops and compact machines.
c) Building higher performance desktop processors.
d) Building server level processors.
e) Able to compete with the largest PC processor maker in the world (i.e. Intel).
We aren't privvy to Intel's private roadmap. I'm quite sure that Apple knew about the various Intel roadmap changes publicly announced in the last six months, and they knew it well before they made their decision. They are also planning for beyond 2007... well beyond. Everyone (well, almost everyone at any rate) criticizing the Intel decision is doing so without knowing everything that the Apple decision makers know.
I don't know about Opterons, but some companies aren't saying "Meh".
Trust me I want to see more AMD sales. It keeps Intel honest about pricing their processors. I'm just saying that it's more difficult than many imagine to move someone off an Intel server to an AMD even if they know the AMD is faster.
Thus is we think about why Apple chose Intel over AMD. This Intel preference is going to have more of a cachet than AMD right now for many people.
That makes sense in retail. Consumers are looking for the best deal possible.
Businesses are a little different. They tend to identify what they need on a more granular level and then hammer you on price. They're much harder to get off of certain products.
Man all day it's DL360, DL380, X336, x346 (HP and IBM servers) I wish I was moving more Opteron based computers but AMD is going to have to maintain a price/performance lead for a bit longer before IT gains the type of trust that means more server boxen flying out the door.
If latency is the only problem seperating an external controller from an ondie controller then Intel likely knows that the migration from DDRII to FB-DIMM will ameliorate the latency issue since theoretically the serialized nature and design of FB-DIMM minimizes latency.
... The Bensley platform, to be released by Intel in 1Q06 will be a landmark product launch. Bensley has quite a few technical improvements to look forward to: dual front side buses, a snoop filter and the first implementation of FB-DIMMs. Intel is also beginning to deliver on their promise to provide platforms, not just chips. By any standards, Bensley will be a competitive server and workstation offering and the first good product based on IntelÕs dual core MPUs. Performance wise, Intel should be closely matched with AMD, assuming that there are no radical shake ups in the roadmaps
We'll see if the benchmarks bare that out in part 2.
Quote:
I think it's foolish for anyone to claim that Intel cannot reclaim a performance lead unless they've seen Merom running in a lab. Wanna laugh...go read some Analyst posts from 4 years ago and see how accurate they were. They would have told you that we'd be using 5Ghz procs by now.
It's all about the fabs, baby. If it wasn't for Intel's fab power, they'd be in even worse shape with Prescott bombing and all. It will be fab power that will bring them back to supremacy again if the new microarchitecture is marginally close to AMD K8 performance. Dothan is already marginally close. Merom, if better IPC, will push them over the edge.
In a time when CPUs are moving to multi-cores and huge caches, I can't imagine how Intel's fab power can't but be the one and only true advantage. Even the Itanium has the potential to become mainstream in 4 or 5 years when multi-core and cache heavy (20+ MB) processing becomes the norm for CPUs.
Anyways, remember what Jobs said about bulding computers they want to build. Somehow, I don't think it is a gigantic box containing the fastest processor on the planet (though I think Intel will provide that with Merom-derivatives anyways, and Itanium even). It's going to be a work of industrial design art with the processor in a confined space where performance/watt is at a premium. I think we're all in agreement about that.
Intel's roadmap looks like a map of Arlington, Va. And they keep ripping it up and changing it.
And yes, as THT says, Apple went to Intel to get low-power processors and chipset integration - not high performance - so this is mostly irrelevant. The above is just an interesting article on the original topic.
Comments
Originally posted by Powerdoc
Your arguments seems convincing, but why Apple did not choose the cell architecture ?. If my memory is correct, Apple considered this choice.
IBM had a different market in mind, the benefits of cell are yet to be proven. I will say it again the shift to Intel was as much about the other chips as it was about the CPU. So what would Apple have at the end of the day with Cell, a great processor that is difficult to program for at best, IBM will sell you the tools but forget GCC support, so either Apple pays IBM to distribute their compilier in Xtools, or they roll their own which would take great efforts and time. They also would still have the problem of all of the other chips, Intel makes much more than CPUs, and they make them in packages designed for various target markets. Let cell prove itself in the market and then let Intel steal the idea, and make their version and while Apple is waiting they get to use all of those other support chips and Macs will not be any different than anything else on the market. This problem does not require much brain effort to figure out why Apple did what they did.
Originally posted by Powerdoc
Your arguments seems convincing, but why Apple did not choose the cell architecture ?. If my memory is correct, Apple considered this choice.
We don't know what Intel's plans look like, but Apple will have gotten a better look at them. Intel also rules in terms of process technology, and Apple can be assured that Intel will build chips that are extremely competitive with the bulk of the laptop and desktop PC markets (duh!). IBM is interested in other things.
This doesn't mean that future Intel chips won't be Cell-like in nature, however. The challenges of what to do with more transistors and how to deal with the realities of physics are invariant across companies.
Apple has a Market Cap of about $50 billion against AMD about $10 Billion US Dollars. Apple could easily afford a cash and script offer for AMD and still have several billion dollars in the bank and little to no debt.
They could also very quickly switch processors over to AMD.
I am not saying it is a good idea, but it could be the ultimate backup plan.
Originally posted by Powerdoc
Your arguments seems convincing, but why Apple did not choose the cell architecture ?. If my memory is correct, Apple considered this choice.
Single threaded integer performance on Cell would be worse than any of the contemporary Intel offerings. Merom cores also appear to be less wattage than contemporary PPU cores.
Also, Intel will have a special, dedicated low-wattage 65-nm fab in 2007, in addition to their high performance fabs. 1.5 to 2 GHz Merom-derivatives on that process will probably be <5 Watts TDP (while the high-perf fabs produce 2 to 3 GHz Merom-derivatives). That means computers with no fans, and we know who is a big fan of that.
I wouldn't be surprised if this low-watt 65nm fab from Intel will be unique to the industry.
http://www.bsc.es/projects/deepcompu..._for_cell.html
I'll stick with a game processor for my console and a desktop processor for my desktop, though.
http://www.theregister.co.uk/2005/10...tel_xeon_2009/
Is it the Apple CPU Maker Curse? It's beginning to look like it!
http://www.forbes.com/markets/2005/1...rtner=yahootix
Originally posted by cubist
"Second-class Intel to trail AMD for years" - The Register
http://www.theregister.co.uk/2005/10...tel_xeon_2009/
Is it the Apple CPU Maker Curse? It's beginning to look like it!
Not likely
This should worry folks at Dell, HP and SGI, as they're most vulnerable to Intel's shortcomings.
Dell only need compare their DL380 sales versus the DL385(AMD) to see that despite the AMD product being faster the Intel sales dwarf AMDs.
"If Intel is ever to reclaim the performance lead from AMD, it must make the transition to an on-chip memory controller," he wrote
Another juicy quote from Nathan Brockwood who I've never ready anything particularly "insightful" from.
Correct me if I'm wrong and don't hesitate to slam me. I'm a big boy and I can take it. Most of the reviews I've seen comparing Opterons to Xeons show that the Opty's have an advantage with memory latency versus the Xeon but in actual throughput numbers that advantage disappears.
So it seems to me that if you have an app that is latency sensitive like a database then the Opteron system should benchmark better but if you have an app that craves massive bandwidth then you should see most of the performance delta disappear.
If latency is the only problem seperating an external controller from an ondie controller then Intel likely knows that the migration from DDRII to FB-DIMM will ameliorate the latency issue since theoretically the serialized nature and design of FB-DIMM minimizes latency.
I think it's foolish for anyone to claim that Intel cannot reclaim a performance lead unless they've seen Merom running in a lab. Wanna laugh...go read some Analyst posts from 4 years ago and see how accurate they were. They would have told you that we'd be using 5Ghz procs by now.
Intel's processors are technically inferior to AMD's - this is an established fact. Intel's business practices are questionable if not blatantly unethical, anticompetitive and monopolistic - again, an established fact.
Intel may have cooked benchmarks that say that in some particular cases Xeons are faster. And they claim that they haven't done anything that other monop- I mean companies have done in the past. So what? It's all a bunch of crap. Give it up.
And guess what? None of that matters. Apple wants to join the mainstream, to take the 'hardware superiority' issue off the table.
Anyone who cares about raw performance hasn't been buying Macs anyway, and they won't buy them when they're Intel-based either.
Originally posted by cubist
"Second-class Intel to trail AMD for years" - The Register
http://www.theregister.co.uk/2005/10...tel_xeon_2009/
Is it the Apple CPU Maker Curse? It's beginning to look like it!
In the case you did not realize it, this was the plan! The only guaranteed method to destroy Intel once and for all.
That's FUD. Opterons destroy Xeons in any benchmark you name
How about revenue generated
Anyone who cares about raw performance hasn't been buying Macs anyway, and they won't buy them when they're Intel-based either.
Evidently they aren't buying Opteron's either in large enough quantities. My point isn't that the Opteron isn't a superior chip it's that the collective response from many companies is "Meh"
Originally posted by hmurchison
[B]How about revenue generated
So by that benchmark, Windows is better than OS X and Dell makes better computers than, say, Apple?
'Cause if you look at revenue..
Evidently they aren't buying Opteron's either in large enough quantities. My point isn't that the Opteron isn't a superior chip it's that the collective response from many companies is "Meh"
I don't know about Opterons, but some companies aren't saying "Meh".
Originally posted by cubist
That's FUD. Opterons destroy Xeons in any benchmark you name. Dell doesn't sell any AMD machines, but it isn't because the Intels are better, it's because Intel basically pays them not to sell AMD. Anyone who thinks there's a level playing field out there has his head in the sand.
Intel's processors are technically inferior to AMD's - this is an established fact. Intel's business practices are questionable if not blatantly unethical, anticompetitive and monopolistic - again, an established fact.
Intel may have cooked benchmarks that say that in some particular cases Xeons are faster. And they claim that they haven't done anything that other monop- I mean companies have done in the past. So what? It's all a bunch of crap. Give it up.
And guess what? None of that matters. Apple wants to join the mainstream, to take the 'hardware superiority' issue off the table.
Anyone who cares about raw performance hasn't been buying Macs anyway, and they won't buy them when they're Intel-based either.
Relax, there is one underlying truth, Apple could have chosen the best CPU, and if that were the case Cell may have been the winner, in switiching to x86 Apple could have also chosen AMD, they did not. As we speak, Intel is shifting the line-up and pulling better performers further ahead to market. Intel has and is making support chips that integrate and compliment nicely. The switch to Intel was made in the face of Cell and AMD dominance in performance areas. Remember what was said performance per watt, Intel will lead, in fact Intel will dominate. Currently no, but look at how much the land scape has changed in the latter half of '05, '06 will only accelerate this.
Originally posted by Brendon
Relax, there is one underlying truth, Apple could have chosen the best CPU, and if that were the case Cell may have been the winner, in switiching to x86 Apple could have also chosen AMD, they did not. As we speak, Intel is shifting the line-up and pulling better performers further ahead to market. Intel has and is making support chips that integrate and compliment nicely. The switch to Intel was made in the face of Cell and AMD dominance in performance areas. Remember what was said performance per watt, Intel will lead, in fact Intel will dominate. Currently no, but look at how much the land scape has changed in the latter half of '05, '06 will only accelerate this.
"Best CPU" is a multi-dimensional problem and without specifying all of the criteria used to evaluate it, you cannot determine which is the best. Apple is very much concerned with low power laptop (and compact machine) processors, and they chose the only company that is certainly going to be:
a) In posession of the best process technology in the world going forward.
b) Building low power high performance processors appropriate for use in laptops and compact machines.
c) Building higher performance desktop processors.
d) Building server level processors.
e) Able to compete with the largest PC processor maker in the world (i.e. Intel).
We aren't privvy to Intel's private roadmap. I'm quite sure that Apple knew about the various Intel roadmap changes publicly announced in the last six months, and they knew it well before they made their decision. They are also planning for beyond 2007... well beyond. Everyone (well, almost everyone at any rate) criticizing the Intel decision is doing so without knowing everything that the Apple decision makers know.
I don't know about Opterons, but some companies aren't saying "Meh".
Trust me I want to see more AMD sales. It keeps Intel honest about pricing their processors. I'm just saying that it's more difficult than many imagine to move someone off an Intel server to an AMD even if they know the AMD is faster.
Thus is we think about why Apple chose Intel over AMD. This Intel preference is going to have more of a cachet than AMD right now for many people.
Originally posted by hmurchison
How about revenue generated
Not that I'm bashing Intel, or saying I think AMD would be better. I don't. But you may find that the revenue table has turned, and is still turning.
AMD vs. Intel Revenue
Businesses are a little different. They tend to identify what they need on a more granular level and then hammer you on price. They're much harder to get off of certain products.
Man all day it's DL360, DL380, X336, x346 (HP and IBM servers) I wish I was moving more Opteron based computers but AMD is going to have to maintain a price/performance lead for a bit longer before IT gains the type of trust that means more server boxen flying out the door.
Originally posted by hmurchison
If latency is the only problem seperating an external controller from an ondie controller then Intel likely knows that the migration from DDRII to FB-DIMM will ameliorate the latency issue since theoretically the serialized nature and design of FB-DIMM minimizes latency.
Hey, lookee:
A Preview of Intel's Bensley Platform (Part I)
... The Bensley platform, to be released by Intel in 1Q06 will be a landmark product launch. Bensley has quite a few technical improvements to look forward to: dual front side buses, a snoop filter and the first implementation of FB-DIMMs. Intel is also beginning to deliver on their promise to provide platforms, not just chips. By any standards, Bensley will be a competitive server and workstation offering and the first good product based on IntelÕs dual core MPUs. Performance wise, Intel should be closely matched with AMD, assuming that there are no radical shake ups in the roadmaps
We'll see if the benchmarks bare that out in part 2.
I think it's foolish for anyone to claim that Intel cannot reclaim a performance lead unless they've seen Merom running in a lab. Wanna laugh...go read some Analyst posts from 4 years ago and see how accurate they were. They would have told you that we'd be using 5Ghz procs by now.
It's all about the fabs, baby. If it wasn't for Intel's fab power, they'd be in even worse shape with Prescott bombing and all. It will be fab power that will bring them back to supremacy again if the new microarchitecture is marginally close to AMD K8 performance. Dothan is already marginally close. Merom, if better IPC, will push them over the edge.
In a time when CPUs are moving to multi-cores and huge caches, I can't imagine how Intel's fab power can't but be the one and only true advantage. Even the Itanium has the potential to become mainstream in 4 or 5 years when multi-core and cache heavy (20+ MB) processing becomes the norm for CPUs.
Anyways, remember what Jobs said about bulding computers they want to build. Somehow, I don't think it is a gigantic box containing the fastest processor on the planet (though I think Intel will provide that with Merom-derivatives anyways, and Itanium even). It's going to be a work of industrial design art with the processor in a confined space where performance/watt is at a premium. I think we're all in agreement about that.
http://www.cooltechzone.com/index.ph...k=view&id=1962
Intel's roadmap looks like a map of Arlington, Va. And they keep ripping it up and changing it.
And yes, as THT says, Apple went to Intel to get low-power processors and chipset integration - not high performance - so this is mostly irrelevant. The above is just an interesting article on the original topic.