I could still see 30Khz/24V-48V as useful within the data center, even within a server itself. It would simplify wiring and PS count greatly if power could be shipped right down to each board or chip.
But what would be the advantage of 30KHz over DC?
You'd still have to convert it to low voltage DC at the end point, before it enters logic chips and disk drives. It's not practical to make low voltage regulated DC without first making higher-voltage unregulated DC along the way. In other words, you have to go AC(distrib)->DC(unreg)->AC(chopped,pwm)->DC(regula ted). This is true regardless of whether the AC distribution is 30KHz, or 50/60Hz.
If you want to just transform and rectify directly from 30kHz to the final regulated DC with no intervening stages, then you've just created a very difficult problem for yourself. How to regulate it? Integrated circuits need very precise voltages. It's at the chopping stage that precision control/regulation portion of power conversion takes place, typically by pulse-width modulation in response to a feedback loop.
There _are_ ways of regulating without having control of the chopper, e.g. the use of mag-amps, but they are costly and complicated, compared to good old fashioned pulse-width modulation.
So, really, the first thing you'd have to do as the 30kHz aproaches a point of consumption is to rectify it to DC. That being the case, why not just distribute it as DC in the first place, and save youself the rectifying and filtering components?
You almost seem to be arguing for power distribution to be higher frequency AC,
No, I'm not. I've heard others argue for it, but I don't agree.
I beleive there are very good fundamental reasons why distribution should be either DC, or low frequency AC. I'd say DC is best for short range (within the data center), and low frequency is best for long distance/large scale, i.e. national power grids. (That is, of course, opinion. Any statement about what's "best" always is.)
Why not use high frequency (RF) distribution? Well, it radiates. Horribly. No matter how well you shield it, and how well you design it, even a tiny percentage of leakage is a huge RFI problem. Audio frequencies are right out, too, because they tend to interfere with audio equipment.
Another objection to RF power distribution is that the little tiny (if expensive) ferrite tranformers and inductors that work so well at high frequencies, don't scale as well to very large distribution systems, whereas the clunky, heavy low frequency iron core tranformers we all know and hate, _do_ scale well to very large sizes. I'm guessing this problem would be quickly solved if we ever to adopt high frequency distribuiton, but there are plenty of other reasons why high frequency distribution is a bad idea.
One unshakeable advantage of low frequency AC, at least at the national grid level, is that it can be transformed with only electromagnetic components. DC needs to be temporarily chopped into AC whenever you convert it to a different voltage level. Chopping very high voltages is very hard (you can't use "normal" transistors), so it's prohibitively complicated/expensive to convert. You can't use lower voltages at the national grid level, because the I^2R losses would be too high. So, DC isn't practical because of conversion, high frequency AC isn't practical because of radiation, that leaves low frequency AC. QED.
The data center is a different matter. There, DC is practical -- more so than any other option in my opinion.
The slashdot story intro implies that the advantage of DC is that you save a conversion step. Well, maybe you do, maybe you don't, but counting the number of AC-to-DC and DC-to-AC conversions is very misleading.
Converting 50 or 60 Hertz to DC is much more costly and less efficient than converting in either direction at a higher frequency. Low frequency rectification requires large filter capacitors, complex and expensive inrush current limiting, and active power-factor correction.
By doing that front-end work in one place only, preferably from a 3-phase source, you save power and increase reliability. You probably still want multiple 50/60Hz to DC rectifier stages, of course, but now they can be in parallel (for redundancy), rather than each one downstream of the other where a failure of either one will bring down the system.
Just because you're distributing DC to the racks, doesn't mean you don't have to convert it again. It typically gets converted to AC and back to DC at least once, usually twice before it reaches CPU and memory chips. That's equally true in data centers that distribute AC or DC. The fact is, memory and CPU devices want very low DC voltages and very high currents. To make matters worse, not all parts of the system want exactly the same DC voltage, you almost always have to have multiple supply rails. You can't distribute very low voltages, because it would require wires as thick as your arm and they'd still be too resistive and inductive, so instead you distribute the DC at, typically, 48 volts. The subsequent conversion to low DC voltages has to happen via an intermediate AC, but it's a high frequency AC, so it can be done much more efficiently using ferrite magnetic components, active rectification, and often resonant mode filters. This high frequency AC is confined to the internals of a power supply unit, it never travels over wires or between boxes, thus reducing typical high-frequency problems such as RFI.
I haven't mentioned battery-backup (i.e. UPSs). They make the system more complex, but don't change any of the fundamental concerns. Even on a DC distribution system, the UPS system requires it's own additional stages of DC->AC->DC conversion, both while charging (standby) and while discharging (during AC power failure). This is because battery charging has to have a precisely controlled current envelope. And batteries don't discharge at the uniform and well-regulatted voltage that your DC distribution wants. They need regulators, and switchmode regulators (typically DC->AC->DC) are the most efficient choice.
But what I don't get is when you fuse an atom, energy is released, but when you split an atom into two, energy is released as well. How is this not perpetual motion?
IANAP (I am not a physicist), but here's how I understand it. Nature loves middle-weight neuclei. Extremely light neuclei (e.g. Hydrogen) and extremely heavy ones (e.g. Plutonium) are less stable.
For very light elements (e.g. hydrogen), fusion releases energy.
For very heavy elements (e.g. Plutonium), fission releases energy.
In both cases, you release energy by moving towards middle-weight elements. If I recall correctly, Iron has the most stable neucleus of all. The raw materials for fission, such as Uranium and Plutonium, are much heavier than Iron. By breaking up the neuclei into lighter elements, you move closer to the ideal middle-weight stable elements, thus releasing energy. Likewise, the raw materials for fission, such as Hydrogen, are much lighter than Iron. By fusing their nuclei, into heavier elements, you move closer to the ideal middle-weight elements, so you release energy.
There's no perpetual motion involved. You can't get energy back by reversing either type of reaction. For example, you'd have to put energy IN, if you wanted to fission Helium back into Hydrogen, because you'd be moving further away from the ideal middle-weigh neuclei.
So, if someone asks you to invest in their iron-fuelled nuclear power plant, your money is probably best invested elsewhere!
I'll leave it to others to advise you on whether to do it yourself, or how to learn, or how hard or easy it is. I'll just add one thing:
If you _do_ decide to learn to solder, use some form of eye protection, every time. I never took eye protection while soldering seriously, until the day I met a one-eyed technician who would have been a two-eyed technician had he worn safety glasses.
Seriously. Safety glasses are cheap. Wearing them is no hassle. Just do it.
Get a Samsung ML-2151N if you still can. I'm not 100% sure if they are still available.
Very sturdy. Very reliable. Duplex printing. Large paper tray. Talks both PCL or Postscript. Network-enabled, just plug in the 10-baseT and go. Also has USB, (not sure how well that works, never tried it). Also available with wireless network, but at extra cost.
It's an especially good choice for networks with a mixture of Windows/Linux machines.
Doesn't need a special driver for Unix, since it's native postscript. The windows drivers work great too. This is the only printer I've owned that works trouble free on both Linux and Windows, and doesn't need a PhD in driver psychology to get working.
I learned Russian at the age of 40+, having many years earlier wrongly concluded that I had no talent for language learning. What changed? Nothing much, I just developed an interest after visiting the country, which resulted in a higher level of motivation. Motivation is everything.
Music is another example. I plateaued my piano learning at the age of about 12, then gave up altogether (bad teacher too, but my own lack of motivation was a huge part). Then, at the age of 42, I discovered the violin. I've been learning at a great pace ever since. What changed? I heard a performance of the Beethoven violin concerto by Corey Cerovsek, and it got me interested in the instrument. Once again, interest and motivation were the deciding factor, not age.
As for technical knowledge, I learned relatively little in college (EE degree), but have learned a huge amount since leaving school. I learned more when I had a concrete reason to want to learn, rather than the abstract motivations that I had at a younger age.
So for me, the equation for ability to learn is simple... age is irellevant, motivation is everything. (Don't know if I'm typical, though. YMMV.)
In either case, I hope you're right about efficiency gains in a dual core chip. I want a computer I can turn on in the summer.
Just to clarify... I was talking about Montecito. So, (1) You won't get one this summer, (2) you can't afford one on your desktop anyway. Just because Montecito will be low power does not imply that the chip in your PC will be. Not yet anyway.
And, I didn't mean to imply that the efficiency gains have anything to do with the dual-core architecture. Not so. It took heroic effort and some amazing innovation to make Montecito such a low power chip. Eventually, other CPUs will _have_ to follow suit, because we are at or beyond the reasonable limit for per-socket supply delivery and cooling.
Note that cache sizes have fluctuated around 256-512 kb since the P2 days. My P2 and P4 both have 512 kb. I'd be shocked if the reason was something other than that being a sweet spot.
Sorry, I live in a 64-bit world, to the point that I'm quite ignorant of X86 state of the art. I've been blindly (and wrongly) assuming a 64-bit context for this whole conversation.
Your posting reminded me that caches of only 512M still exist! Montecito has 24M between 2 cores. Also, re-reading your posts in the context of 32-bit systems, they now make much more sense to me. X86 die aren't the same huge monsters that I'm used to. No wonder you and I have different views about yield cost tradeoffs -- we live on different parts of the curve.
Unfortunately, most of what I know about Itanium I really can't talk about, other than in very general terms -- assuming I wish to remain employed, that is.:)
The "sweet spot" for cache size is really determined by a race between core performance, and memory latency/bandwidth. Doubling cores doubles the data production/consumption rate. Doubling the frequency also does. The former is less demanding on memory latency and mostly requires more bandwidth. The latter is equally demanding on both. If you double the data production/consumption, and keep the same memory/bus bandwidth, ideally, you'd like to halve the cache miss rate -- but that's pretty unlikely in practice. That's why caches keep growing (at least in the 64-bit world). There is a point of diminuishing returns, because there's an upper limit to both temporal and spacial locality, but we're not quite there yet for Itanium.
When more CPUs start to have integrated memory controllers and point-to-point links instead of multi-drop busses, I predict that cache sizes per core will actually decline for a while, since the memory performance side of the balance will lower the "sweet spot". After that, caches will probably creep slowly upwards again, because no memory or interconnect technology ever scales fast enough to keep up with CPU core performance scaling.
So the $64 question is... When cache sizes per core start to decline or level off, will we see smaller die, or will we see even more cores per die? The way Intel seems to always position Itanium for high-end heavy metal, I expect huge die with more cores, although for X86 they went the other way. Or so I infer from your posting.
Of course, I'm no more thrilled with a 200 watt CPU than anyone else, but that's what you get with a two CPU system anyway.
Not sure which CPU you're referring to here. Is it something in the X86 line? If you're taking a single core power and multiplying by two, then you may be very pleasantly surprised. Montecito, despite having two cores, will have significantly lower total power than its single-core predecessors. (That much I can safely reveal, because it seems to be common knowledge already.)
Having designed systems, I can tell you that difficulties arising from high power-per-socket are very non linear: 200W isn't merely twice as hard to deal with versus 100W. It's easily an order of magnitude more difficult to cool with the same MTBF reliability. Luckily, Intel have realised this, at least in their 64-bit line. Once again, I am too ignorant of the 32-bit world to know the state of the art there.
It's almost as if they've reached some kind of break even point, where the probability of a defect on a dual core die falls to the point where the gains outweigh the losses.
The gains definitely outweigh the losses, or they wouldn't do it. But the gains don't only come from CPU cost-per-core. There are lots of other factors, such as density, power efficiency, potential for core-to-core lockstep, etc.
I have no first-hand knowledge of AMD, but for Itanium, smaller process geometries do not increase yield through smaller die size. As they've shrunk to smaller geometries, they have not shrunk die size at all. All the extra real estate goes into larger caches, and the die size, and thus the (raw) yield, remains about the same. They have improved yield, but it's not through shrinking the die.
They have dramatically improved yield in other ways. As your execution units shrink and cache dominates an ever-increasing percentage of the die area, it becomes easier to use redundancy to make the chip tolerant of defects.
Intel calls it Pellston Technology (I hate marketing speak). And it it this, more than anything, that makes such massive die as Montecito even possible. In the old days, one defect trashed the die. With this sort of technology, most defects are worked around through redundancy. And, if you have too many defects to allow that, you may still salvage the die by selling it at a lower price with a reduced-capacity caches. Most chips shipped to customers have several completely corrected defects.
You're wrong.
Probably. Wouldn't be the first time.
If you take into account the overall _system_ cost, dual-core is definitely far cheaper than dual-socket. System cost also includes cost of board area, power delivery, cooling, etc. OEMs will happily pay well over double for a 2-core vs 1-core because of the savings they will make elsewhere in the system. Dual-core also gives system vendors much more flexibility by allowing the same board design to support twice as wide a range of CPU counts.
I was comparing CPU+package cost only. Having been both a CPU designer and a system designer at different times in my career, I know how to look at it both ways.
Basically, they print two processors onto the same hunk of silicon and it's cheaper to manufacture than two seperate processors because all the other costs like packaging it and testing it stay the same.
Packaging and testing, sure. But overall cost of 1-dual vs 2-single isn't as clear. Big die are expensive -- they require more costly fab techniques, and result in low yield. Beyond a certain size, the loss in yield is just huge.
You can partially recover the lost yield by salvaging some of the failing dual-core die to sell as single-core parts. There are limits to this, though. They make lousy single-core parts for many reasons, including very high leakage power and a larger die that you have to package. To be viable, you need a high dual-core yield.
Bottom line: for equivalent complexity and cache size, I seriously doubt that it's any cheaper to produce one dual-core chip compared to two single-core ones, knowing how sensitive IC economics are to yield.
Proliphix's web site is an example of VERY bad marketing.
At the right price, I'd probably buy one. Even if they don't sell them directly, surely Proliphix's web site ought to give some clue how or where to buy one. What retailers carry them? Who sells them on the internet? How much they cost? Something!
There's a link labelled "DEALERS", but it only describes how to become a dealer, not how to find an existing dealer.
I invested 5 minutes searching for this info, and found nothing. Even a Google search turned up nothing. During those 5 minutes, I stumbled over many competing products (not identical, rather more X-10ish, but still, other people who will gladly take the customers money before the customer ever tracks down how to buy a Proliphix.
Forget all the silly Leonardo and Kate crap, though. That story line was bogus embellishment for a great, otherwise-true story that needed no embelishment. Ignore all of it (except perhaps Kate's tits, which _did_ add something to the movie).
Consider the rest of the movie: The historical facts, and they way they were presented, the special effects, the drama behind the real events (ignoring the silly fictional ones). It was a great movie.
I saw it twice. The first time, I was just blown away by the awesome spectacle I had seen. As I left the theater, overhearing conversations, I was astonished to realize that everyone else leaving the movie were intently discussing the Kate & Leo story, which I'd more or less blocked out as annoying distraction.
Sure, it could have been better. And shorter. All it would have taken to achieve both goals is to leave Leo and Kate on the cutting-room floor. Then it would have been one of the best movies I've ever seen.
How did Cliffhanger somehow not make the bottom 100 list?
Sure, you can always rely on a truly appaulingly bad movie from Stalone. But Cliffhanger is in a class of its own, awful even by Stalone movie standards.
I don't know if there's some quantum lower limit on movie quality, but if there is, this movie defines it.
My LASIK turned out wonderful! My vision is now perfect, absolutely no complications, no dry eye after the first week, no pain, no discomfort, just absolutely magnificent vision. I've had glasses and contact lenses in the past, but they just don't compare to this, for both convenience and quality of vision. My night vision does now include a tiny bit of haloing, but barely enough to be a minor annoyance.
Best $4000 I ever spent!
But I still wouldn't recommend it to my friends. It's all a matter or an individual's risk tolerence. I took the risk knowingly, and am VERY glad I did. But if a friend of mine took the same risk based upon my recommendation, and it turned out badly, I'd feel responsible. So, all I can say is it was the right choice for me, and I'm delighted with the outcome. Not every story I've heard from others has been as positive, though.
Re:HOW MANY shares?
on
Google IPO Swami
·
· Score: 5, Informative
Sorry, I can't help myself. A MAN named Pat Beatty...
Pat isn't too unusual a name for a man. It's short for Patrick. (Go to Ireland, some time).
... married a lady named Beatty Rowell?!?
Oops, my typo! Her name was Betty, not Beatty.
Thanks god she didn't take his surname!
She did take his surname. I wrote her name as "Betty Rowell" in the posting, to keep it amazon.com-friendly, in case anyone was interested in her book. We knew her as "Betty Beatty", but she wrote under the name "Betty Rowell Beatty", and amazon.com only lists her under "Betty Rowell", for some reason.
Back in the 60s and 70s, Pat Beatty and Fritz Johl did similar work with glider wings. Obviously, with much different technology. They flew their prototypes, and raced them competitively. In addition to variable-geometry, they also expeimented with variable-span!
The technology of the day was far less sophisticated than today, but it's an interesting bit of aeronautical history nonetheless.
Although I met Pat Beatty once or twice during the early 80s, I was too young to have seen his variable-geometry and variable-span creations fly, first hand. Most of what I know about them I heard from the old-timers in my flying club, who had been active in gliding competition during the 60s.
Sadly, there seems to be very little surviving literature available on the Beatty and Beatty-Johl designs. Google turns up a few grainy photographs, and articles in ancient editions of Soaring Magazine and Krautkorant (Cape Gliding Club Newsletter), but that's about it.
Pat's wife Beatty Rowell also made significant contributions to aviation, both as a pilot and meteorologist, and wrote the book "Just for the Love of Flying". Time for a re-read, I think.
Slashdot Mirror
Errr... Did you read my whole post?
You're saying that I miss the point, and then you go on to repeat exactly the same point I already made.
Oh well.
I could still see 30Khz/24V-48V as useful within the data center,
a ted). This is true regardless of whether the AC distribution is 30KHz, or 50/60Hz.
even within a server itself. It would simplify wiring and PS count
greatly if power could be shipped right down to each board or
chip.
But what would be the advantage of 30KHz over DC?
You'd still have to convert it to low voltage DC at the end point, before it enters logic chips and disk drives. It's not practical to make low voltage regulated DC without first making higher-voltage unregulated DC along the way. In other words, you have to go AC(distrib)->DC(unreg)->AC(chopped,pwm)->DC(regul
If you want to just transform and rectify directly from 30kHz to the final regulated DC with no intervening stages, then you've just created a very difficult problem for yourself. How to regulate it? Integrated circuits need very precise voltages. It's at the chopping stage that precision control/regulation portion of power conversion takes place, typically by pulse-width modulation in response to a feedback loop.
There _are_ ways of regulating without having control of the chopper,
e.g. the use of mag-amps, but they are costly and complicated, compared to good old fashioned pulse-width modulation.
So, really, the first thing you'd have to do as the 30kHz aproaches a point of consumption is to rectify it to DC. That being the case, why not just distribute it as DC in the first place, and save youself the rectifying and filtering components?
Not usually! Every telco setup I've ever seen connects the distribution buses straight to the batteries.
I stand corrected.
You almost seem to be arguing for power distribution to be higher
frequency AC,
No, I'm not. I've heard others argue for it, but I don't agree.
I beleive there are very good fundamental reasons why distribution
should be either DC, or low frequency AC. I'd say DC is best for
short range (within the data center), and low frequency is best for
long distance/large scale, i.e. national power grids. (That is, of
course, opinion. Any statement about what's "best" always is.)
Why not use high frequency (RF) distribution? Well, it radiates.
Horribly. No matter how well you shield it, and how well you design
it, even a tiny percentage of leakage is a huge RFI problem. Audio
frequencies are right out, too, because they tend to interfere with
audio equipment.
Another objection to RF power distribution is that the little tiny (if
expensive) ferrite tranformers and inductors that work so well at high
frequencies, don't scale as well to very large distribution systems,
whereas the clunky, heavy low frequency iron core tranformers we all
know and hate, _do_ scale well to very large sizes. I'm guessing this
problem would be quickly solved if we ever to adopt high frequency
distribuiton, but there are plenty of other reasons why high frequency
distribution is a bad idea.
One unshakeable advantage of low frequency AC, at least at the
national grid level, is that it can be transformed with only
electromagnetic components. DC needs to be temporarily chopped into
AC whenever you convert it to a different voltage level. Chopping
very high voltages is very hard (you can't use "normal" transistors),
so it's prohibitively complicated/expensive to convert. You can't use
lower voltages at the national grid level, because the I^2R losses
would be too high. So, DC isn't practical because of conversion, high
frequency AC isn't practical because of radiation, that leaves low
frequency AC. QED.
The data center is a different matter. There, DC is practical -- more
so than any other option in my opinion.
The slashdot story intro implies that the advantage of DC is that you
save a conversion step. Well, maybe you do, maybe you don't, but
counting the number of AC-to-DC and DC-to-AC conversions is very
misleading.
Converting 50 or 60 Hertz to DC is much more costly and less efficient
than converting in either direction at a higher frequency. Low
frequency rectification requires large filter capacitors, complex and
expensive inrush current limiting, and active power-factor correction.
By doing that front-end work in one place only, preferably from a
3-phase source, you save power and increase reliability. You probably
still want multiple 50/60Hz to DC rectifier stages, of course, but now
they can be in parallel (for redundancy), rather than each one
downstream of the other where a failure of either one will bring down
the system.
Just because you're distributing DC to the racks, doesn't mean you
don't have to convert it again. It typically gets converted to AC and
back to DC at least once, usually twice before it reaches CPU and
memory chips. That's equally true in data centers that distribute AC
or DC. The fact is, memory and CPU devices want very low DC voltages
and very high currents. To make matters worse, not all parts of the
system want exactly the same DC voltage, you almost always have to
have multiple supply rails. You can't distribute very low voltages,
because it would require wires as thick as your arm and they'd still
be too resistive and inductive, so instead you distribute the DC at,
typically, 48 volts. The subsequent conversion to low DC voltages has
to happen via an intermediate AC, but it's a high frequency AC, so it
can be done much more efficiently using ferrite magnetic components,
active rectification, and often resonant mode filters. This high
frequency AC is confined to the internals of a power supply unit, it
never travels over wires or between boxes, thus reducing typical
high-frequency problems such as RFI.
I haven't mentioned battery-backup (i.e. UPSs). They make the system
more complex, but don't change any of the fundamental concerns. Even
on a DC distribution system, the UPS system requires it's own
additional stages of DC->AC->DC conversion, both while charging
(standby) and while discharging (during AC power failure). This is
because battery charging has to have a precisely controlled current
envelope. And batteries don't discharge at the uniform and
well-regulatted voltage that your DC distribution wants. They need
regulators, and switchmode regulators (typically DC->AC->DC) are the
most efficient choice.
A scary thought indeed, considering the current volume of email spam!
IANAP (I am not a physicist), but here's how I understand it. Nature loves middle-weight neuclei. Extremely light neuclei (e.g. Hydrogen) and extremely heavy ones (e.g. Plutonium) are less stable.
In both cases, you release energy by moving towards middle-weight elements. If I recall correctly, Iron has the most stable neucleus of all. The raw materials for fission, such as Uranium and Plutonium, are much heavier than Iron. By breaking up the neuclei into lighter elements, you move closer to the ideal middle-weight stable elements, thus releasing energy. Likewise, the raw materials for fission, such as Hydrogen, are much lighter than Iron. By fusing their nuclei, into heavier elements, you move closer to the ideal middle-weight elements, so you release energy.
There's no perpetual motion involved. You can't get energy back by reversing either type of reaction. For example, you'd have to put energy IN, if you wanted to fission Helium back into Hydrogen, because you'd be moving further away from the ideal middle-weigh neuclei.
So, if someone asks you to invest in their iron-fuelled nuclear power plant, your money is probably best invested elsewhere!
I'll leave it to others to advise you on whether to do it yourself, or how to learn, or how hard or easy it is. I'll just add one thing:
If you _do_ decide to learn to solder, use some form of eye protection, every time. I never took eye protection while soldering seriously, until the day I met a one-eyed technician who would have been a two-eyed technician had he worn safety glasses.
Seriously. Safety glasses are cheap. Wearing them is no hassle. Just do it.
Get a Samsung ML-2151N if you still can. I'm not 100% sure if they are still available.
Very sturdy. Very reliable. Duplex printing. Large paper tray. Talks both PCL or Postscript. Network-enabled, just plug in the 10-baseT and go. Also has USB, (not sure how well that works, never tried it). Also available with wireless network, but at extra cost.
It's an especially good choice for networks with a mixture of Windows/Linux machines.
Doesn't need a special driver for Unix, since it's native postscript. The windows drivers work great too. This is the only printer I've owned that works trouble free on both Linux and Windows, and doesn't need a PhD in driver psychology to get working.
I learned Russian at the age of 40+, having many years earlier wrongly concluded that I had no talent for language learning. What changed? Nothing much, I just developed an interest after visiting the country, which resulted in a higher level of motivation. Motivation is everything.
Music is another example. I plateaued my piano learning at the age of about 12, then gave up altogether (bad teacher too, but my own lack of motivation was a huge part). Then, at the age of 42, I discovered the violin. I've been learning at a great pace ever since. What changed? I heard a performance of the Beethoven violin concerto by Corey Cerovsek, and it got me interested in the instrument. Once again, interest and motivation were the deciding factor, not age.
As for technical knowledge, I learned relatively little in college (EE degree), but have learned a huge amount since leaving school. I learned more when I had a concrete reason to want to learn, rather than the abstract motivations that I had at a younger age.
So for me, the equation for ability to learn is simple... age is irellevant, motivation is everything. (Don't know if I'm typical, though. YMMV.)
In either case, I hope you're right about efficiency gains in a dual core chip. I want a computer I can turn on in the summer.
Just to clarify... I was talking about Montecito. So, (1) You won't get one this summer, (2) you can't afford one on your desktop anyway. Just because Montecito will be low power does not imply that the chip in your PC will be. Not yet anyway.
And, I didn't mean to imply that the efficiency gains have anything to do with the dual-core architecture. Not so. It took heroic effort and some amazing innovation to make Montecito such a low power chip. Eventually, other CPUs will _have_ to follow suit, because we are at or beyond the reasonable limit for per-socket supply delivery and cooling.
Note that cache sizes have fluctuated around 256-512 kb since the P2 days. My P2 and P4 both have 512 kb. I'd be shocked if the reason was something other than that being a sweet spot.
:)
Sorry, I live in a 64-bit world, to the point that I'm quite ignorant of X86 state of the art. I've been blindly (and wrongly) assuming a 64-bit context for this whole conversation.
Your posting reminded me that caches of only 512M still exist! Montecito has 24M between 2 cores. Also, re-reading your posts in the context of 32-bit systems, they now make much more sense to me. X86 die aren't the same huge monsters that I'm used to. No wonder you and I have different views about yield cost tradeoffs -- we live on different parts of the curve.
Unfortunately, most of what I know about Itanium I really can't talk about, other than in very general terms -- assuming I wish to remain employed, that is.
The "sweet spot" for cache size is really determined by a race between core performance, and memory latency/bandwidth. Doubling cores doubles the data production/consumption rate. Doubling the frequency also does. The former is less demanding on memory latency and mostly requires more bandwidth. The latter is equally demanding on both. If you double the data production/consumption, and keep the same memory/bus bandwidth, ideally, you'd like to halve the cache miss rate -- but that's pretty unlikely in practice. That's why caches keep growing (at least in the 64-bit world). There is a point of diminuishing returns, because there's an upper limit to both temporal and spacial locality, but we're not quite there yet for Itanium.
When more CPUs start to have integrated memory controllers and point-to-point links instead of multi-drop busses, I predict that cache sizes per core will actually decline for a while, since the memory performance side of the balance will lower the "sweet spot". After that, caches will probably creep slowly upwards again, because no memory or interconnect technology ever scales fast enough to keep up with CPU core performance scaling.
So the $64 question is... When cache sizes per core start to decline or level off, will we see smaller die, or will we see even more cores per die? The way Intel seems to always position Itanium for high-end heavy metal, I expect huge die with more cores, although for X86 they went the other way. Or so I infer from your posting.
Of course, I'm no more thrilled with a 200 watt CPU than anyone else, but that's what you get with a two CPU system anyway.
Not sure which CPU you're referring to here. Is it something in the X86 line? If you're taking a single core power and multiplying by two, then you may be very pleasantly surprised. Montecito, despite having two cores, will have significantly lower total power than its single-core predecessors. (That much I can safely reveal, because it seems to be common knowledge already.)
Having designed systems, I can tell you that difficulties arising from high power-per-socket are very non linear: 200W isn't merely twice as hard to deal with versus 100W. It's easily an order of magnitude more difficult to cool with the same MTBF reliability. Luckily, Intel have realised this, at least in their 64-bit line. Once again, I am too ignorant of the 32-bit world to know the state of the art there.
It's almost as if they've reached some kind of break even point, where the probability of a defect on a dual core die falls to the point where the gains outweigh the losses.
The gains definitely outweigh the losses, or they wouldn't do it. But the gains don't only come from CPU cost-per-core. There are lots of other factors, such as density, power efficiency, potential for core-to-core lockstep, etc.
I have no first-hand knowledge of AMD, but for Itanium, smaller process geometries do not increase yield through smaller die size. As they've shrunk to smaller geometries, they have not shrunk die size at all. All the extra real estate goes into larger caches, and the die size, and thus the (raw) yield, remains about the same. They have improved yield, but it's not through shrinking the die.
They have dramatically improved yield in other ways. As your execution units shrink and cache dominates an ever-increasing percentage of the die area, it becomes easier to use redundancy to make the chip tolerant of defects.
Intel calls it Pellston Technology (I hate marketing speak). And it it this, more than anything, that makes such massive die as Montecito even possible. In the old days, one defect trashed the die. With this sort of technology, most defects are worked around through redundancy. And, if you have too many defects to allow that, you may still salvage the die by selling it at a lower price with a reduced-capacity caches. Most chips shipped to customers have several completely corrected defects.
You're wrong.
Probably. Wouldn't be the first time.
If you take into account the overall _system_ cost, dual-core is definitely far cheaper than dual-socket. System cost also includes cost of board area, power delivery, cooling, etc. OEMs will happily pay well over double for a 2-core vs 1-core because of the savings they will make elsewhere in the system. Dual-core also gives system vendors much more flexibility by allowing the same board design to support twice as wide a range of CPU counts.
I was comparing CPU+package cost only. Having been both a CPU designer and a system designer at different times in my career, I know how to look at it both ways.
Basically, they print two processors onto the same hunk of silicon and it's cheaper to manufacture than two seperate processors because all the other costs like packaging it and testing it stay the same.
Packaging and testing, sure. But overall cost of 1-dual vs 2-single isn't as clear. Big die are expensive -- they require more costly fab techniques, and result in low yield. Beyond a certain size, the loss in yield is just huge.
You can partially recover the lost yield by salvaging some of the failing dual-core die to sell as single-core parts. There are limits to this, though. They make lousy single-core parts for many reasons, including very high leakage power and a larger die that you have to package. To be viable, you need a high dual-core yield.
Bottom line: for equivalent complexity and cache size, I seriously doubt that it's any cheaper to produce one dual-core chip compared to two single-core ones, knowing how sensitive IC economics are to yield.
If only one core is defect free, is it possible to disable the dud and sell it as a single core CPU?
Yes, it is possible, in most cases. (Although there are a few types of defects that would prohibit this, such as power shorts).
For example, hypothetically, Intel could sell a single core version of Montecito called the Half Monte and a dual core version called the Full Monte.
Proliphix's web site is an example of VERY bad marketing.
At the right price, I'd probably buy one. Even if they don't sell them directly, surely Proliphix's web site ought to give some clue how or where to buy one. What retailers carry them? Who sells them on the internet? How much they cost? Something!
There's a link labelled "DEALERS", but it only describes how to become a dealer, not how to find an existing dealer.
I invested 5 minutes searching for this info, and found nothing. Even a Google search turned up nothing. During those 5 minutes, I stumbled over many competing products (not identical, rather more X-10ish, but still, other people who will gladly take the customers money before the customer ever tracks down how to buy a Proliphix.
Titanic was a great movie! Really!
Forget all the silly Leonardo and Kate crap, though. That story line was bogus embellishment for a great, otherwise-true story that needed no embelishment. Ignore all of it (except perhaps Kate's tits, which _did_ add something to the movie).
Consider the rest of the movie: The historical facts, and they way they were presented, the special effects, the drama behind the real events (ignoring the silly fictional ones). It was a great movie.
I saw it twice. The first time, I was just blown away by the awesome spectacle I had seen. As I left the theater, overhearing conversations, I was astonished to realize that everyone else leaving the movie were intently discussing the Kate & Leo story, which I'd more or less blocked out as annoying distraction.
Sure, it could have been better. And shorter. All it would have taken to achieve both goals is to leave Leo and Kate on the cutting-room floor. Then it would have been one of the best movies I've ever seen.
How did Cliffhanger somehow not make the bottom 100 list?
Sure, you can always rely on a truly appaulingly bad movie from Stalone. But Cliffhanger is in a class of its own, awful even by Stalone movie standards.
I don't know if there's some quantum lower limit on movie quality, but if there is, this movie defines it.
My LASIK turned out wonderful! My vision is now perfect, absolutely no complications, no dry eye after the first week, no pain, no discomfort, just absolutely magnificent vision. I've had glasses and contact lenses in the past, but they just don't compare to this, for both convenience and quality of vision. My night vision does now include a tiny bit of haloing, but barely enough to be a minor annoyance.
Best $4000 I ever spent!
But I still wouldn't recommend it to my friends. It's all a matter or an individual's risk tolerence. I took the risk knowingly, and am VERY glad I did. But if a friend of mine took the same risk based upon my recommendation, and it turned out badly, I'd feel responsible. So, all I can say is it was the right choice for me, and I'm delighted with the outcome. Not every story I've heard from others has been as positive, though.
HOW MANY shares?
It's right there in the FAQ. 10 shares.
I'd like to use that. I use MBNA, but can't find it on their web site. Any clue to help me find it?
Never mind, just found it.
The trick is, you have to be logged into your account to find it. There seems to be no info about it on pages accessible while not logged in.
Thanks for telling me about this. Very useful.
I'd like to use that. I use MBNA, but can't find it on their web site. Any clue to help me find it?
Thx.
I thought Beatty Beatty was a little too off the wall.
:-)
No worse than Boutros Boutros-Ghali, though.
(Yup, that was an unfortunate typo.)
Pat isn't too unusual a name for a man. It's short for Patrick. (Go to Ireland, some time).
Oops, my typo! Her name was Betty, not Beatty.
Thanks god she didn't take his surname!
She did take his surname. I wrote her name as "Betty Rowell" in the posting, to keep it amazon.com-friendly, in case anyone was interested in her book. We knew her as "Betty Beatty", but she wrote under the name "Betty Rowell Beatty", and amazon.com only lists her under "Betty Rowell", for some reason.
Back in the 60s and 70s, Pat Beatty and Fritz Johl did similar work with glider wings. Obviously, with much different technology. They flew their prototypes, and raced them competitively. In addition to variable-geometry, they also expeimented with variable-span!
The technology of the day was far less sophisticated than today, but it's an interesting bit of aeronautical history nonetheless.
Although I met Pat Beatty once or twice during the early 80s, I was too young to have seen his variable-geometry and variable-span creations fly, first hand. Most of what I know about them I heard from the old-timers in my flying club, who had been active in gliding competition during the 60s.
Sadly, there seems to be very little surviving literature available on the Beatty and Beatty-Johl designs. Google turns up a few grainy photographs, and articles in ancient editions of Soaring Magazine and Krautkorant (Cape Gliding Club Newsletter), but that's about it.
Pat's wife Beatty Rowell also made significant contributions to aviation, both as a pilot and meteorologist, and wrote the book "Just for the Love of Flying". Time for a re-read, I think.