Having merged with ATI, AMD now has all the IP it needs to do such a device. AMD in fact has a so-called "Fusion" development program http://en.wikipedia.org/wiki/AMD_Fusion to do just that. The Silicon on Insulator (SOI) process technology that AMD (unlike Intel) uses for CPUs has low leakage and is well suited for low-power devices. So in the short term, while the GPU and CPU are still separate, good reductions in power consumption can already be had by switching the production of the GPU/chipset of mobile devices to SOI.
They have a good chance. For one, their market share is rather higher than you make it out to be: about 20% of the 80x86 market vs Intel's 80%. Also, the computer manufacturers have an interest in keeping the competition between Intel and AMD alive. Unless they behave irrationally, they will help AMD to fully break the monopoly.
But the main thing that is pending for AMD is the antitrust lawsuit. Assuming there will be a just judgment, which is not a given with the US justice system led by the likes of Alberto Gonzales, a multi-billion dollar compensation for anti-competitive practices will fall to AMD. They have enough debt financing to last until then.
Indeed, let's wait for the benchmarks. I would like some more real-world and 64-bit benchmarks: most recent reviews seems to have studiously avoided those in favor of synthetic 32-bit only benchmarks that are not very representative and are easily skewed with processor-specific optimizations.
And I'm not sure going to 45nm process will allow Intel to step back ahead. It seems process improvements have been yielding diminishing results in performance related areas. Transistor density will go up, though, so Intel can compensate by adding more cache. Also, AMD's process technology is a little advanced than Intel's at the same feature size: Intel does not do Silicon on Insulator, dual stress liners, and a few other things.
When it comes to multi-processing scalability, AMD's Barcelone/10h/Phenm single-die four core with hypertransport inter-chip interconnects will do far better than the two-die four core shared-bus Intel chips. Also, both the old and new AMD architecture will do relatively better on 64-bit code than the Intel Core 2 architecture: Intel's micro-op fusion does not work work in 64-bit, and their 64-bit extensions are a relatively recent add on to the old Core architecture. The FPU power of the new 10h architecture will be excellent as well.
On the other hand, Intel chips will remain very competitive on integer code, cache-happy benchmarks, particularly when run in 32-bit mode. Also, the SSE4 extensions of the upcoming 45nm Intel ships will help for encoding/decoding and some rendering applications, provided that the software has been properly optimized to take advantage of them.
The "Cold Fusion" field has seen many more experimental successes: detection of neutrons, tritium, helium, transmutations of heavier elements, non-natural-abundance isotope ratios, detection of ionizing radiation. The best place to visit for an overview of the field is http://www.lenr-canr.org/.
Though the experiments are remarkable, no concensus on the theory has emerged yet. Nuclear reactions are clearly happening, but it is doubtful that it is conventional fusion, that is, nuclei moving fast enough to surmount their mutual Coulombic repulsion. Something seems to be screening or catalysing the reactions.
I refer to the phenomena observed during cold fusion research. This does include excess heat, reproduced many times, in several labs. But there is more: low levels of neutrons, production of tritium, production of helium, and other isotopes. Easy to measure at very high sensitivity using mass spectroscopy.
However, it is imprudent to suppose that all of these phenomena are necessarily the result of conventional fusion reactions. The Coulomb barrier is too high, and the results are freaky enough to suggest something not predicted by conventional theory. It seems probable that there is a screening mechanism at play, analogous to muon-catalyzed fusion.
In science, experiment trumps theory and preconception. Take the trouble to look at the experimental results first. They're pretty clear cut. Then look at blockage, ridicule and spin surrounding the results. You'll notice that it goes rather far beyond the normal civilized scientific discourse. That, and the fact that most physicists would like novelty, as you rightly state, suggests protection of interests outside of physics.
Smolin is right in his critique. But the cause lies deeper. Academic physics (and other branches of science) are controlled. Part of the control involves funding innocuous research such as the mathematical acrobatics of string theory. However, when research gets close to upsetting an applecart, public ridicule, editorial control, and career sabotage is deployed.
For example, we all have been told that cold fusion is bunk, and that quantum theory is so well-tested that it is virtually proven. Well, think again. The cold fusion phenomena are real and have been replicated hundreds of times. And there is a new theory (GUT CQM) that outperforms standard quantum mechanics on the only frontier that really counts: falsifyable and quantitatively precise predictions of observations and phenomena.
The story makes clear why a statistic such as average income or wealth is bogus. It means nothing for the vast majority of people. A far better measure would be median income or wealth.
But then, government statistics are there to mislead you instead of inform you. To see what else is bogus, look here: http://www.shadowstats.com/cgi-bin/sgs/.
Quite so. It is highly suspect that the main stream media have immediately and massively covered this story, and that the security apparatus is feeding them with novel little factoids on a continual basis. Keeps it festering. Putin aint no saint, but this media campaign reminds me of 9/11, with Islamism, Osama and the hijackers being fingered, pictures and all, before the day was over.
The reality is for WiFi, mobile phones or similar technology to cause the problems that are often suggested would require a significant new way for EM fields to interact with matter that has gone completely unnoticed till now. This would require that parts of the standard model that have been experimentaly proved to unprecidented levels of precision are also plain wrong. It just is not happening fokes.
Sorry, you are misguided. Yes, the photon energy is too small to directly break bonds or cause a chemical reaction. But there is ample room for indirect mechanisms. For example, the spin direction of the unpaired electron of a free radical can easily be changed through low-frequency low-power EM fields, and several chemical processes such as recombination of free-radical pairs depend on spin direction (because of the Pauli exclusion principle).
The review fails to mention that, unlike AMD's current mobile Turion CPUs, the upcoming Yonahs will not run 64-bit code. What is Intel thinking? With 64-bit OS and software support increasingly available, who will want to invest a lot in such a laptop? Yet dual-core laptops are supposed to be high-end, and, being a more expensive investment, ought to last longer.
But what exactly was unethical about lab workers also being donors in the first place?
Yes, most confusing. What is so unethical about donating a few of many eggs for your own research? And rather sexist too, IMO. No-one would be up in arms about a male researcher blowing a wad so as to research a few sperms. And imagine the waste involved in that! All those millions of poor chanceless wasted spermlets.
And it is all in such a good and long tradition. Back around 1670-1680, the inventor of the optical microscope, Antony van Leeuwenhoek, discovered sperm cells by, you guessed it, examining his own sperm.
Umm, that's good and all for the Athlon XP and 64 series, but could you please tell me how an Opteron 148 compares to a 175 or a 240?
People that buy servers are supposed to have enough clue to not require a rating system. Besides, Opteron specs have more relevant dimensions. For large server deployments, performance per Watt is often the relevant metric. Also, the number of cHT ports determines in a how-many-way SMP system you can use the Opteron.
The first digit of the Opteron model number indicates the number of coherent HyperTransport (cHT) ports. "1" for 1-socket systems with zero cHT ports per CPU, "2" for 2 sockets and 1 port, "8" for 4-8 sockets and 3 ports. Note that when the Horus cHT bridging and caching chip is released next week, Opteron single-memory-image boxes will scale up to 32-64 sockets. Put dual-core Opterons in the sockets and you'll have a 64-128-way box that will decimate price/performance in the big-iron space.
The review calculates price/performance based on the price of the CPUs instead of total system cost. A useless measure, since a CPU on its own cannot do anything useful. It also hides the added system costs for CPUs that consume a lot of power: larger PSU, more cooling and noise reduction measures. And then there are the additional platform costs for CPUs that only work with particular chipsets or expensive motherboards. Never mind the increment to your electricity bill.
What this smells like is yet another bullshit metric invented by the Intel marketing department. One wonders how much these review sites get paid for prostituting themselves.
Yeah, Intel is stupid. Their engineers didn't realize that their obviously-old netburst architecture couldn't compete against amd. How idiots they were.
Aside from the feeble attempt at sarcasm, the above does raise a serious question: were their engineers really that dumb? No, but they were not given a voice
Intel's troubles were caused by hubris and marketing-driven decisions on the part of management. They went for the NetBurst/P4 design because MHz sells. They ignored warnings by the process specialists about the risks of such a high-frequency design. They assumed they had the monopoly power to move the market over to Itanium, and thus did not start planning an 80x86 design beyond NetBurst/P4 until AMD scared the shit out of them with Opteron/K8.
It could have been much worse, even. The Pentium-M was a semi-skunkworks project by their Israeli design team. It is an evolution of the old PIII architecture. Without it, they would have been trounced in the mobile space, and would have had nothing to plug the gap between now and their wholly new architecture which is expected late 2006.
The AMD chips have got more memory bandwidth, so they can keep both cpus fed with data reasonably well.
Not just that. The AMD dualcore chips have an on-chip connection between the cores: both cores share a crossbar fronting the memory controllers and have the on-chip equivalent of a coherent HyperTransport connection. So, you see, the AMD design is in fact a real dual-core design. The current Intel dual-cores, on the other hand, share nothing on-chip.
Today, AMD opened Fab 36. It is a 300mm Fab that will top out at twice the capacity of their current 200mm Fab 30. It will ramp over the next two years. That means they'll be able to serve half of the market. Up from 18% or so, now. In addition, they have arrangements for outsourcing production to Chartered Semi, if there is even more demand to cope with.
Add to that superior CPU products in the desktop, mobile, and server space as well as OEM's that seem to have lost their fear of Intel retalition, and what we have here is a revolution that will unseat Intel from its monopoly position.
The method referenced is called "magnetic domain-wall logic". It works by propagating domain walls in small ferromagnetic wires. They have demonstrated various logic elements. Unfortunately, the domain-wall propagation is powered by immersing the circuit into a rotating or alternating magnetic field. This requires bulky external coils, rotating the IC itself inside a permanent magnet, or rotating a permanent magnet around the IC.
In the demonstrations, field frequency is measured in the tens of Hertz. I could not find a reference to the propagation velocity, but one wire width per cycle would be a reasonable guess. In short, horribly slow.
People complaining about Lua memory management performance and multi-tasking scalability are behind the times.
The 5.1 development version of Lua, due for final release late 2005, includes an incremental garbage collector and has been enhanced by Mike Pall with optional patches that allow for the efficient multitasking of large numbers of scripts (http://lua-users.org/wiki/ResumableVmPatch and http://luajit.luaforge.net/coco.html). Note that those patches fit a niche similar to stackless Python. Running scripts in different threads was already supported in prior Lua versions with only a small amount of C-side infrastructure.
In addition, Mike has recently released a JIT (http://luajit.luaforge.net/) extension for even better performance.
Something sure does not add up, so here is my guess as to what is going on.
Apple is making most of its money with IPods. But those run the risk of eventually loosing out to converged devices, specifically mobile phones that also play music. So, Apple would be wise to react by selling mobile phones that are ITunes ready.
But they lack the technology. Intel has it, but has not been able to sell their phone chips. In part because of bad execution: big company ills. In part because people are weary of a monopolist. So Intel is in need of a good and loyal brand to push their mobile tech.
It is therefore likely that the computing deal is only stage one in a wider-ranging cooperation where Intel provides the technology and the money, and Apple the design and the cool.
Slashdot Mirror
Having merged with ATI, AMD now has all the IP it needs to do such a device. AMD in fact has a so-called "Fusion" development program http://en.wikipedia.org/wiki/AMD_Fusion to do just that. The Silicon on Insulator (SOI) process technology that AMD (unlike Intel) uses for CPUs has low leakage and is well suited for low-power devices. So in the short term, while the GPU and CPU are still separate, good reductions in power consumption can already be had by switching the production of the GPU/chipset of mobile devices to SOI.
There is nothing weird about it if you know who owns the media and how they operate. Disney, Viacom, Timewarner, News Corp, Bertelsmann and GE own more than 90% of US media outlets. See http://en.wikipedia.org/wiki/Concentration_of_medi a_ownership for details. To get an inside view of how the media handle stories that are unwelcome to the establishment, I can recommend the following book: http://www.amazon.com/Into-Buzzsaw-Leading-Journal ists-Expose/dp/1591022304
Greg Palast is one of the few good reporters out there still willing to probe further and deeper. An interesting book of his, detailing the election, war, and oil machinations is "Armed Madhouse": http://www.gregpalast.com/madhouse/index.php/about /. For a quick summary, see this talk he gave in Portland: http://video.google.com/videoplay?docid=-680222806 2297352475.
They have a good chance. For one, their market share is rather higher than you make it out to be: about 20% of the 80x86 market vs Intel's 80%. Also, the computer manufacturers have an interest in keeping the competition between Intel and AMD alive. Unless they behave irrationally, they will help AMD to fully break the monopoly.
But the main thing that is pending for AMD is the antitrust lawsuit. Assuming there will be a just judgment, which is not a given with the US justice system led by the likes of Alberto Gonzales, a multi-billion dollar compensation for anti-competitive practices will fall to AMD. They have enough debt financing to last until then.
Indeed, let's wait for the benchmarks. I would like some more real-world and 64-bit benchmarks: most recent reviews seems to have studiously avoided those in favor of synthetic 32-bit only benchmarks that are not very representative and are easily skewed with processor-specific optimizations.
And I'm not sure going to 45nm process will allow Intel to step back ahead. It seems process improvements have been yielding diminishing results in performance related areas. Transistor density will go up, though, so Intel can compensate by adding more cache. Also, AMD's process technology is a little advanced than Intel's at the same feature size: Intel does not do Silicon on Insulator, dual stress liners, and a few other things.
When it comes to multi-processing scalability, AMD's Barcelone/10h/Phenm single-die four core with hypertransport inter-chip interconnects will do far better than the two-die four core shared-bus Intel chips. Also, both the old and new AMD architecture will do relatively better on 64-bit code than the Intel Core 2 architecture: Intel's micro-op fusion does not work work in 64-bit, and their 64-bit extensions are a relatively recent add on to the old Core architecture. The FPU power of the new 10h architecture will be excellent as well. On the other hand, Intel chips will remain very competitive on integer code, cache-happy benchmarks, particularly when run in 32-bit mode. Also, the SSE4 extensions of the upcoming 45nm Intel ships will help for encoding/decoding and some rendering applications, provided that the software has been properly optimized to take advantage of them.
For an video/documentary outlining the status of the "Cold Fusion" field, see the following over on Google video: http://video.google.com/videoplay?docid=6426393169 641611451&q=COLD+FUSION&hl=en
The "Cold Fusion" field has seen many more experimental successes: detection of neutrons, tritium, helium, transmutations of heavier elements, non-natural-abundance isotope ratios, detection of ionizing radiation. The best place to visit for an overview of the field is http://www.lenr-canr.org/.
Though the experiments are remarkable, no concensus on the theory has emerged yet. Nuclear reactions are clearly happening, but it is doubtful that it is conventional fusion, that is, nuclei moving fast enough to surmount their mutual Coulombic repulsion. Something seems to be screening or catalysing the reactions.
I refer to the phenomena observed during cold fusion research. This does include excess heat, reproduced many times, in several labs. But there is more: low levels of neutrons, production of tritium, production of helium, and other isotopes. Easy to measure at very high sensitivity using mass spectroscopy. However, it is imprudent to suppose that all of these phenomena are necessarily the result of conventional fusion reactions. The Coulomb barrier is too high, and the results are freaky enough to suggest something not predicted by conventional theory. It seems probable that there is a screening mechanism at play, analogous to muon-catalyzed fusion.
In science, experiment trumps theory and preconception. Take the trouble to look at the experimental results first. They're pretty clear cut. Then look at blockage, ridicule and spin surrounding the results. You'll notice that it goes rather far beyond the normal civilized scientific discourse. That, and the fact that most physicists would like novelty, as you rightly state, suggests protection of interests outside of physics.
Smolin is right in his critique. But the cause lies deeper. Academic physics (and other branches of science) are controlled. Part of the control involves funding innocuous research such as the mathematical acrobatics of string theory. However, when research gets close to upsetting an applecart, public ridicule, editorial control, and career sabotage is deployed.
For example, we all have been told that cold fusion is bunk, and that quantum theory is so well-tested that it is virtually proven. Well, think again. The cold fusion phenomena are real and have been replicated hundreds of times. And there is a new theory (GUT CQM) that outperforms standard quantum mechanics on the only frontier that really counts: falsifyable and quantitatively precise predictions of observations and phenomena.
The story makes clear why a statistic such as average income or wealth is bogus. It means nothing for the vast majority of people. A far better measure would be median income or wealth.
But then, government statistics are there to mislead you instead of inform you. To see what else is bogus, look here: http://www.shadowstats.com/cgi-bin/sgs/.
Quite so. It is highly suspect that the main stream media have immediately and massively covered this story, and that the security apparatus is feeding them with novel little factoids on a continual basis. Keeps it festering. Putin aint no saint, but this media campaign reminds me of 9/11, with Islamism, Osama and the hijackers being fingered, pictures and all, before the day was over.
The review fails to mention that, unlike AMD's current mobile Turion CPUs, the upcoming Yonahs will not run 64-bit code. What is Intel thinking? With 64-bit OS and software support increasingly available, who will want to invest a lot in such a laptop? Yet dual-core laptops are supposed to be high-end, and, being a more expensive investment, ought to last longer.
Yes, most confusing. What is so unethical about donating a few of many eggs for your own research? And rather sexist too, IMO. No-one would be up in arms about a male researcher blowing a wad so as to research a few sperms. And imagine the waste involved in that! All those millions of poor chanceless wasted spermlets.
And it is all in such a good and long tradition. Back around 1670-1680, the inventor of the optical microscope, Antony van Leeuwenhoek, discovered sperm cells by, you guessed it, examining his own sperm.
People that buy servers are supposed to have enough clue to not require a rating system. Besides, Opteron specs have more relevant dimensions. For large server deployments, performance per Watt is often the relevant metric. Also, the number of cHT ports determines in a how-many-way SMP system you can use the Opteron.
The first digit of the Opteron model number indicates the number of coherent HyperTransport (cHT) ports. "1" for 1-socket systems with zero cHT ports per CPU, "2" for 2 sockets and 1 port, "8" for 4-8 sockets and 3 ports. Note that when the Horus cHT bridging and caching chip is released next week, Opteron single-memory-image boxes will scale up to 32-64 sockets. Put dual-core Opterons in the sockets and you'll have a 64-128-way box that will decimate price/performance in the big-iron space.
Some simple checks can prevent this sillyness from perpetuating. Bob Park's "What's New" column http://www.bobpark.org/ is an amusing and up to date reference for this kind of thing. Here is what he has to say about the "Integrity Research Institute" (the name alone should have raised a red flag): http://www.searchum.umd.edu/search?q=%22integrity+ research+institute%22&site=&btnG=Search+UM&output= xml_no_dtd&sort=date%3AD%3AL%3Ad1&ie=UTF-8&client= UMCP&oe=UTF-8&proxystylesheet=UMCP
The review calculates price/performance based on the price of the CPUs instead of total system cost. A useless measure, since a CPU on its own cannot do anything useful. It also hides the added system costs for CPUs that consume a lot of power: larger PSU, more cooling and noise reduction measures. And then there are the additional platform costs for CPUs that only work with particular chipsets or expensive motherboards. Never mind the increment to your electricity bill.
What this smells like is yet another bullshit metric invented by the Intel marketing department. One wonders how much these review sites get paid for prostituting themselves.
Aside from the feeble attempt at sarcasm, the above does raise a serious question: were their engineers really that dumb? No, but they were not given a voice
Intel's troubles were caused by hubris and marketing-driven decisions on the part of management. They went for the NetBurst/P4 design because MHz sells. They ignored warnings by the process specialists about the risks of such a high-frequency design. They assumed they had the monopoly power to move the market over to Itanium, and thus did not start planning an 80x86 design beyond NetBurst/P4 until AMD scared the shit out of them with Opteron/K8.
It could have been much worse, even. The Pentium-M was a semi-skunkworks project by their Israeli design team. It is an evolution of the old PIII architecture. Without it, they would have been trounced in the mobile space, and would have had nothing to plug the gap between now and their wholly new architecture which is expected late 2006.
Not just that. The AMD dualcore chips have an on-chip connection between the cores: both cores share a crossbar fronting the memory controllers and have the on-chip equivalent of a coherent HyperTransport connection. So, you see, the AMD design is in fact a real dual-core design. The current Intel dual-cores, on the other hand, share nothing on-chip.
Today, AMD opened Fab 36. It is a 300mm Fab that will top out at twice the capacity of their current 200mm Fab 30. It will ramp over the next two years. That means they'll be able to serve half of the market. Up from 18% or so, now. In addition, they have arrangements for outsourcing production to Chartered Semi, if there is even more demand to cope with.
Add to that superior CPU products in the desktop, mobile, and server space as well as OEM's that seem to have lost their fear of Intel retalition, and what we have here is a revolution that will unseat Intel from its monopoly position.
The method referenced is called "magnetic domain-wall logic". It works by propagating domain walls in small ferromagnetic wires. They have demonstrated various logic elements. Unfortunately, the domain-wall propagation is powered by immersing the circuit into a rotating or alternating magnetic field. This requires bulky external coils, rotating the IC itself inside a permanent magnet, or rotating a permanent magnet around the IC.
In the demonstrations, field frequency is measured in the tens of Hertz. I could not find a reference to the propagation velocity, but one wire width per cycle would be a reasonable guess. In short, horribly slow.
People complaining about Lua memory management performance and multi-tasking scalability are behind the times.
The 5.1 development version of Lua, due for final release late 2005, includes an incremental garbage collector and has been enhanced by Mike Pall with optional patches that allow for the efficient multitasking of large numbers of scripts (http://lua-users.org/wiki/ResumableVmPatch and http://luajit.luaforge.net/coco.html). Note that those patches fit a niche similar to stackless Python. Running scripts in different threads was already supported in prior Lua versions with only a small amount of C-side infrastructure.
In addition, Mike has recently released a JIT (http://luajit.luaforge.net/) extension for even better performance.
Something sure does not add up, so here is my guess as to what is going on.
Apple is making most of its money with IPods. But those run the risk of eventually loosing out to converged devices, specifically mobile phones that also play music. So, Apple would be wise to react by selling mobile phones that are ITunes ready.
But they lack the technology. Intel has it, but has not been able to sell their phone chips. In part because of bad execution: big company ills. In part because people are weary of a monopolist. So Intel is in need of a good and loyal brand to push their mobile tech.
It is therefore likely that the computing deal is only stage one in a wider-ranging cooperation where Intel provides the technology and the money, and Apple the design and the cool.