The quads from Intel provide four physical cores per socket. That is the definition of a quad in this context.
Well, yeah, and if I say the context is the motherboard, then I can define a four-socket board holding single-core CPUs to be a quad core chip. It would be equally ludicrous in the context of chip design.
... the FSB issue would cripple performance. Amazingly, it wasn't the case - they sold in droves
I take it you haven't looked at proper SMP benchmarks. Sitting on the same FSB sucked rather badly for SMP scalability. What sells is a function of marketing and perception. When it comes to managing that, Intel rules. That's why you have to search hard to find any real-world SMP benchmarks being done by the various bought-and-paid-for review websites.
Customers just want to fit 4 cores in one socket. That's all that matters.
Note that the post was about the number of cores/threads in the Niagra chip design. In terms of chip design, the circuitry on the silicon is what matters, not how you package, integrate, or market it. Moreover, it does matter to a customer if marketing speak fobs him with two dualcore chips on a cracker instead of an integrated four core design.
Performance does not scale purely with the number of cores, it also matters how efficiently the cores can share resources and intercommunicate. Things like accessing shared memory and inter-process communication are an important part of real world applications. Just try to run a heavy duty threaded server benchmark with a lot of IPC on a faux "Quad" and compare how it scales relative to a true Dual core design, and you'll be lucky to get 1.5-1.6 times the performance instead of twice.
Along with having more cores than the quads from Intel and AMD...
What quad from Intel/AMD? Intel is selling two dual cores on a cracker. The "quad" bit is just marketing, the actual silicon chips are pure dual core designs that have to talk across the front side bus just as in a two-socket server. And AMD has so far only been previewing their quads, you can't buy them yet.
There has not to be any reason for their use of multiple frequencies, more than that the inventors believe that it is important.
How about economic viability? We are talking about a commercial product. There is a strong incentive to build something with a good price/performance ratio. Experimenting around to find conditions that work well is not hard, so it is safe to assume that that was done.
Don't be silly. Pure heat energy provides energy quanta on the order of k*T (Boltzmann's constant times the temperature in Kelvins). Only when this is close to bond energies will materials break down through a purely thermal (no catalysts, radicals, or whatever) process. With typical bond energies on the order of 1 eV, you'll see that the required temperature for a purely thermal dissociation process is way higher than can be found in systems were with lots of chemistry going on, such as biological tissue.
Lai and Singh's work have not been surpressed: rather it has shown to be hard to replicate.
How can you possibly interpret the actions that have been taken surrounding Lai and Singh's work (see the article I linked) as anything other than surpression? And their work has been replicated: http://www.springerlink.com/content/21np1etlm5pj5u 5g/, in a simpler system even.
And there is another explanation (...) it is only the heat that gives the result.
If that were the case, why use multiple specific frequencies instead of just an un-tuned microwave source or Ohmic heating? Besides, there are strong indications that an unusual mechanism is at play. Lai and Singh have found that the DNA effects are reduced when using an iron chelate (removes iron from tissue) first. Maybe iron services as an intermediate or catalyst for free radical formation under EM excitation. It is not implausible because such heavy elements have a large spin-orbit coupling.
Clearly, the specific properties of the microwaves used (not just the heating) cause specific chemical changes in the plastic and rubber compound. Essentially, polymer chains are being broken up. But we as consumers are told not to worry about microwaves from cell phones, WiFi, or base stations.
Why? Well, the photon energy (Planck's constant times the microwave frequency) is too low to cause chemical changes. All microwaves are supposed to do to tissue is a small amount of heating. Yet clearly, other processes that damage polymers can go on as well. DNA is a polymer, and breaks in DNA induced by EM fields is exactly what Lai and Singh at Washington state university have found. Their work has been surpressed: http://www.washington.edu/alumni/columns/march05/w akeupcall01.html.
So you want completely synthetic benchmarks instead of testing the applications that their readers actually use.
No, not completely. Application benchmarks are valuable for those that use that application. But as a measure of general system performance they are limited. You need a to combine multiple application benchmarks in order to get some idea of system performance. And individual applications can be (and have been) skewed in favor of particular hardware in a manner that is not easy to detect (being closed source). Then there is the possibility of bias in picking one application and not the other.
I would prefer to see vendors of important applications benchmark their own product on various types of hardware and post the results on their product website. That way, heavy users of that application can inform themselves by pulling information from a source far less likely to be biased to one or the other hardware vendor.
The rot is far deeper. This article vastly understates the problem: there are so many levers manufacturers can pull in order to influence or bias reviews, payola is only the start of it. Development of corrupt benchmark software used by the review sites can be bought, biased compilers (Intel compiler) generate some of the code being benched, advertisement money can be withheld or expanded, early or free samples can be provided or denied.
The review sites, in turn, can do a lot to make review seem fair while applying a subtle bias. They can limit themselves to certain benchmarks, (de)emphasize or arbitrarily weigh some results, frame the the article, or spin the conclusion.
It is not hard to see this in action. Take the pervasive and saturating Core 2 hype on all sites, last year, for example. Many sites were running the same biased selection of benchmarks. Nearly all sites avoided 64-bit benchmarks.
I would like to see a bootable Linux benchmark CD that runs stock GCC compiled code in 32 and 64-bit mode and provides various workload, scalability, and throughput tests. Something that is open and runs precisely the same code on all machines. Something anyone can pop in his own PC or laptop. But then, even if that were to exist, would the sites start to report that benchmark in their reviews?
Sigh, yet another general statement without supporting evidence. I think your post is a sly bit of astroturfing for NVidia. ATI has had WHQL ( http://en.wikipedia.org/wiki/WHQL_Testing ) certified new driver releases for years now. NVidia has only recently been able to get their new releases WHQL qualified. Sure, there is more to drivers, but it indicates that ATI has had a solid development, testing, and qualification regimen in place for a long time.
It is not a store. It is not spam. What you see here, ladies and gentlemen, is a desperate attempt to surpress a dissenting scientific viewpoint that is well corroborated by observation. Consider: why do these people feel compelled to resort smearing in order to stifle a scientific debate?
IOW, he's sticking to his ideas that were tenable 40 years ago, even though they haven't been for the past 20 years.
Now now, Arp is basing his view on observation. Nature does not lie. That the consensus is still otherwise does not mean his views are untenable. In the end, experiment and observation will win out, but it can take a while just like it took a while for the heliocentric model to be accepted.
I can somewhat imagine this being a bit of a possibility in the medical field, which is a large financial business if anything. But cosmology?
Indeed. This has the people being censored bemused as well. They mostly think the vested interest being defended is the large amount of prestige and funding tied up in current the positions that the mainstream have staked out and their research programs. I think the problem is more fundamental: cosmology may not be that relevant, but it is based on physics which certainly can be very relevant. There are good reasons to believe that a proper cosmology will require fundamental revisions to physics, relativity theory in particular. See for example the observations mentioned in this online book http://surf.de.uu.net/bookland/sci/farce/farce_toc .html.
A few more things, by the sound of it. The first thing you should note is that the peer review system is very effective at filtering information. This makes it suited to both its official intent, which is to improve the quality of discourse, as well as to censorship. You seem to assume it is the former, but that is just an assumption about the intent and integrity of those holding editorial positions and key chairs.
Well hidden from university students and the general public?
Not really. We're talking about a fairly large number of scientists. The problem is the peer review system: they have trouble getting their observations and views published. The general public, in turn, is not knowledgable enough to make an independent judgment, and as such relies on the scientific consensus which is moderated by the peer review system.
Current cosmology is anything but settled. The following interesting documentary shows the perspective of astrophysicists and cosmologists that believe the mainstream view is flawedhttp://www.mininova.org/tor/360930. There definitely are quite a few observations that do not fit the mainstream cosmology.
The cosmologies described here are based on the inference that the universe is expanding in a manner proportional to the observed roughly constant redshift-to-distance ratio (Hubble constant). The idea is that as space is stretched, the wavelength of light is stretched along with it, as it transverses that space.
Essentially, current mainstream cosmology is likely to be complete bunk, because it is predicated on one particular ill-founded interpretation of redshift.
For mobile, AMD has gone a different route for now, they have reworked the K8 for extremely low power: http://www.theinquirer.net/default.aspx?article=39 894. The two cores and memory controller get independent voltage planes. And the cores can clock up and down independently. It makes good sense: for mobile, low power is crucial.
Many of the high-end features (double FPU units, hypertransport interconnects, and so on) of the Barcelona design are not required for a laptop, and add power draw caused by static leakage, even when not in use. In due time, though, AMD will no doubt rework the K10/Barcelona core into a mobile design. Probably they will release a moderately power mobile Barcelona version before that, for high-end workstation type laptops.
Indeed so. Anyone having bought or buying an AM2/AM+2 desktop motherboard can drop in Phenom processors. When you have a performance AMD 4x4 (1,207-pin Socket F) board with FX processors, you can drop in the new quad core FX chips as well. Similarly, when you have a DDR2 Opteron server/big-iron, you can also upgrade.
That makes the current AMD platforms attractive: you can buy a cheap Athlon X2 chip to get good performance now, and later upgrade to a Phenom chip and get excellent performance and four-way multiprocessing. I plan to wait with my upgrade until the price comes down a bit.
According to Intel engineers though, communication between the chips was never a bottleneck
What a load of crap. For quite a few applications, it definitely is a bottleneck. If you have single-threaded tasks that sit happily on their own processor and do not intercommunicate, then, yeah, it does not matter much what connectivity the cores and dies have. But in the real world, multi-threading and SMP tasks do need to intercommunicate, often heavily so. Also, processes will often migrate from one core to the next because the core it was running on before is in use. At that moment, fast inter-core synchronization of the caches is very helpful.
On-chip connectivity can be much broader and lower-latency than off-chip connectivity. The two-dual-core in one package "quad cores" of Intel have to talk via the off-package north bridge. As you can see from the AMD Barcelona/K10/10h snapshot, the cores live together on a single piece of silicon.
The space between the the cores is a very broad crossbar, allowing fast inter-core synchronization/cache-coherency. The uniform block at the edge of the chip, outside the cores, is the L3 cache shared by all four cores. Each core has its own L1 and L2 cache. This design is nicely symmetric: each core has equivalent resources. It should do very well on heavy-duty symmetric multiprocessing applications.
This story is a classic example of selling your research by pretending it may have a practical application. Do something with a quantum system, and sell it by saying it has a potential quantum computing application. Do something on a nanoscopic scale, and call it nanotechnology. Do something with a semiconductor, and say it is for future chips.
What these researchers have done is pretty standard fare: the nitrogen vacancy defect in diamond is photo-active center that can and has been studied extensively by optical excitation. The unpaired electron spins of these centres can be manipulated via microwaves. With a low concentration and tight focus, you can study individual centers. Some of these will have and adjacent carbon-13 instead of the more abundant carbon-12 neighbor. Its magnetic moment can be observed through its interaction with the electron spin. Cute, but nothing what so fucking ever to do with any kind of practical application.
Science has become colonized by hypesters, marketeers, and slick talking band wagon jumpers. All in pursuit of that next bit of funding and fame.
Slashdot Mirror
Well, yeah, and if I say the context is the motherboard, then I can define a four-socket board holding single-core CPUs to be a quad core chip. It would be equally ludicrous in the context of chip design.
I take it you haven't looked at proper SMP benchmarks. Sitting on the same FSB sucked rather badly for SMP scalability. What sells is a function of marketing and perception. When it comes to managing that, Intel rules. That's why you have to search hard to find any real-world SMP benchmarks being done by the various bought-and-paid-for review websites.
Note that the post was about the number of cores/threads in the Niagra chip design. In terms of chip design, the circuitry on the silicon is what matters, not how you package, integrate, or market it. Moreover, it does matter to a customer if marketing speak fobs him with two dualcore chips on a cracker instead of an integrated four core design.
Performance does not scale purely with the number of cores, it also matters how efficiently the cores can share resources and intercommunicate. Things like accessing shared memory and inter-process communication are an important part of real world applications. Just try to run a heavy duty threaded server benchmark with a lot of IPC on a faux "Quad" and compare how it scales relative to a true Dual core design, and you'll be lucky to get 1.5-1.6 times the performance instead of twice.
Don't be silly. Pure heat energy provides energy quanta on the order of k*T (Boltzmann's constant times the temperature in Kelvins). Only when this is close to bond energies will materials break down through a purely thermal (no catalysts, radicals, or whatever) process. With typical bond energies on the order of 1 eV, you'll see that the required temperature for a purely thermal dissociation process is way higher than can be found in systems were with lots of chemistry going on, such as biological tissue.
How can you possibly interpret the actions that have been taken surrounding Lai and Singh's work (see the article I linked) as anything other than surpression? And their work has been replicated: http://www.springerlink.com/content/21np1etlm5pj5u 5g/, in a simpler system even.
If that were the case, why use multiple specific frequencies instead of just an un-tuned microwave source or Ohmic heating? Besides, there are strong indications that an unusual mechanism is at play. Lai and Singh have found that the DNA effects are reduced when using an iron chelate (removes iron from tissue) first. Maybe iron services as an intermediate or catalyst for free radical formation under EM excitation. It is not implausible because such heavy elements have a large spin-orbit coupling.
Clearly, the specific properties of the microwaves used (not just the heating) cause specific chemical changes in the plastic and rubber compound. Essentially, polymer chains are being broken up. But we as consumers are told not to worry about microwaves from cell phones, WiFi, or base stations.
Why? Well, the photon energy (Planck's constant times the microwave frequency) is too low to cause chemical changes. All microwaves are supposed to do to tissue is a small amount of heating. Yet clearly, other processes that damage polymers can go on as well. DNA is a polymer, and breaks in DNA induced by EM fields is exactly what Lai and Singh at Washington state university have found. Their work has been surpressed: http://www.washington.edu/alumni/columns/march05/w akeupcall01.html.
No, not completely. Application benchmarks are valuable for those that use that application. But as a measure of general system performance they are limited. You need a to combine multiple application benchmarks in order to get some idea of system performance. And individual applications can be (and have been) skewed in favor of particular hardware in a manner that is not easy to detect (being closed source). Then there is the possibility of bias in picking one application and not the other.
I would prefer to see vendors of important applications benchmark their own product on various types of hardware and post the results on their product website. That way, heavy users of that application can inform themselves by pulling information from a source far less likely to be biased to one or the other hardware vendor.
The rot is far deeper. This article vastly understates the problem: there are so many levers manufacturers can pull in order to influence or bias reviews, payola is only the start of it. Development of corrupt benchmark software used by the review sites can be bought, biased compilers (Intel compiler) generate some of the code being benched, advertisement money can be withheld or expanded, early or free samples can be provided or denied.
The review sites, in turn, can do a lot to make review seem fair while applying a subtle bias. They can limit themselves to certain benchmarks, (de)emphasize or arbitrarily weigh some results, frame the the article, or spin the conclusion.
It is not hard to see this in action. Take the pervasive and saturating Core 2 hype on all sites, last year, for example. Many sites were running the same biased selection of benchmarks. Nearly all sites avoided 64-bit benchmarks.
I would like to see a bootable Linux benchmark CD that runs stock GCC compiled code in 32 and 64-bit mode and provides various workload, scalability, and throughput tests. Something that is open and runs precisely the same code on all machines. Something anyone can pop in his own PC or laptop. But then, even if that were to exist, would the sites start to report that benchmark in their reviews?
Sigh, yet another general statement without supporting evidence. I think your post is a sly bit of astroturfing for NVidia. ATI has had WHQL ( http://en.wikipedia.org/wiki/WHQL_Testing ) certified new driver releases for years now. NVidia has only recently been able to get their new releases WHQL qualified. Sure, there is more to drivers, but it indicates that ATI has had a solid development, testing, and qualification regimen in place for a long time.
It is not a store. It is not spam. What you see here, ladies and gentlemen, is a desperate attempt to surpress a dissenting scientific viewpoint that is well corroborated by observation. Consider: why do these people feel compelled to resort smearing in order to stifle a scientific debate?
A few more things, by the sound of it. The first thing you should note is that the peer review system is very effective at filtering information. This makes it suited to both its official intent, which is to improve the quality of discourse, as well as to censorship. You seem to assume it is the former, but that is just an assumption about the intent and integrity of those holding editorial positions and key chairs.
Secondly, editorial systems have been thoroughly corrupted before. For an example, read this book http://www.amazon.com/Into-Buzzsaw-Leading-Journal ists-Expose/dp/1591022304.
Current cosmology is anything but settled. The following interesting documentary shows the perspective of astrophysicists and cosmologists that believe the mainstream view is flawedhttp://www.mininova.org/tor/360930. There definitely are quite a few observations that do not fit the mainstream cosmology.
The cosmologies described here are based on the inference that the universe is expanding in a manner proportional to the observed roughly constant redshift-to-distance ratio (Hubble constant). The idea is that as space is stretched, the wavelength of light is stretched along with it, as it transverses that space.
The problem with all these mainstream cosmologies is that observations have been made that require rather different (non-cosmological) mechanisms for redshift to exist. Halton Arp has made and detailed these observations, and the surrounding controversy http://www.amazon.com/Seeing-Red-Redshifts-Cosmolo gy-Academic/dp/0968368905. Paul Mermet is another astrophysicist that has studied the matter http://www.newtonphysics.on.ca/HUBBLE/Hubble.html.
Essentially, current mainstream cosmology is likely to be complete bunk, because it is predicated on one particular ill-founded interpretation of redshift.
For mobile, AMD has gone a different route for now, they have reworked the K8 for extremely low power: http://www.theinquirer.net/default.aspx?article=39 894. The two cores and memory controller get independent voltage planes. And the cores can clock up and down independently. It makes good sense: for mobile, low power is crucial.
Many of the high-end features (double FPU units, hypertransport interconnects, and so on) of the Barcelona design are not required for a laptop, and add power draw caused by static leakage, even when not in use. In due time, though, AMD will no doubt rework the K10/Barcelona core into a mobile design. Probably they will release a moderately power mobile Barcelona version before that, for high-end workstation type laptops.
Indeed so. Anyone having bought or buying an AM2/AM+2 desktop motherboard can drop in Phenom processors. When you have a performance AMD 4x4 (1,207-pin Socket F) board with FX processors, you can drop in the new quad core FX chips as well. Similarly, when you have a DDR2 Opteron server/big-iron, you can also upgrade.
That makes the current AMD platforms attractive: you can buy a cheap Athlon X2 chip to get good performance now, and later upgrade to a Phenom chip and get excellent performance and four-way multiprocessing. I plan to wait with my upgrade until the price comes down a bit.
On-chip connectivity can be much broader and lower-latency than off-chip connectivity. The two-dual-core in one package "quad cores" of Intel have to talk via the off-package north bridge. As you can see from the AMD Barcelona/K10/10h snapshot, the cores live together on a single piece of silicon.
The space between the the cores is a very broad crossbar, allowing fast inter-core synchronization/cache-coherency. The uniform block at the edge of the chip, outside the cores, is the L3 cache shared by all four cores. Each core has its own L1 and L2 cache. This design is nicely symmetric: each core has equivalent resources. It should do very well on heavy-duty symmetric multiprocessing applications.
This story is a classic example of selling your research by pretending it may have a practical application. Do something with a quantum system, and sell it by saying it has a potential quantum computing application. Do something on a nanoscopic scale, and call it nanotechnology. Do something with a semiconductor, and say it is for future chips.
What these researchers have done is pretty standard fare: the nitrogen vacancy defect in diamond is photo-active center that can and has been studied extensively by optical excitation. The unpaired electron spins of these centres can be manipulated via microwaves. With a low concentration and tight focus, you can study individual centers. Some of these will have and adjacent carbon-13 instead of the more abundant carbon-12 neighbor. Its magnetic moment can be observed through its interaction with the electron spin. Cute, but nothing what so fucking ever to do with any kind of practical application.
Science has become colonized by hypesters, marketeers, and slick talking band wagon jumpers. All in pursuit of that next bit of funding and fame.