It's only the SPEs in the PS3 that are specialized: the main CPU, the PPE, is similar to a PowerPC 970.

For the sake of argument, I think we can agree that the SPEs (aka SPUs aka APUs aka SPCs) are the specialized part of Cell and barring the PowerPC 620, that Cell and Xenon derivatives (featuring the more or less common PPE) are the only other 64 bit processors in the same timeframe as the PowerPC 970. The 970 could retire up to 5 instructions per cycle wheras the PPE could complete 8 single precision (or 4 double precision) instructions per cycle. FAR more importantly, the PPE was limited to in-order execution, wheras the PPC970 could handle out-of-order. The PPE's pipeline depth was horrendously long, allowing for 3.5GHz execution at the 90nm technology node, while the PPC970 was restricted to 1-2GHz operation. By any meaningful benchmark the 970 was a far faster core.

The PPE's sole purpose for existing was to feed SPEs data to crunch. Microsoft had a barely managable deadline for which the PPE was repurposed and Xenon was created.

If you're interested in the creation of Cell and Xenon, read The Race for a New Game Machine. A lot of people I know are in there, some by different names.

With the billions of people who have by now inhabited Earth and died here, we'd by up to our armpits in protoplasm if they really did exist.

I've read that more people exist today than have ever died. If we accept a modern population explosion of that magnitude, even if not quite true, you can simply imagine doubling the population of the earth. While there are local areas where people would notice the population doubling, most of the earth is sparsely enough populated that we wouldn't even need to cross paths. Part of the premise lies in "ghosts" not needing to eat or needing shelter, although I'm sure there are enough attics and basements that quite a number of "ghosts" could get shelter. Additionally, most of the people I know who believe in "ghosts" believe they're a small fraction of the dead, largely those with unfinished business. Assuming there's a positive rate of "ghosts" who are able to "finish their business" (perhaps along the lines of, "everybody I know of has now died, there's nobody of meaning to torment), I don't see how your statement is necessarily a barrier to accepting the theory.

Obviously, I'm skeptical, but I have other reasons. Firstly, a lack of evidence.

I could say my home is haunted. Only on some occasions, there are some strange clanging noises coming from the walls and ceilings. At some times, it feels like it's coming from all around me. My home is only 15 years old, but we're not the first inhabitants. Now, breaking that down, it's only from November to March, and it's when I'm using the hot water and for the next 5 minutes after that, and the rooms affected are between the hot water use and the water heater... I conclude, with zero direct evidence, that it's hot water expanding the pipes and later cooling. Really, it's an assumption on my part based on my beliefs. For all I know, the Flying Spaghetti Monster Himself could be reaching out with His Noodly Appendage to knock on my walls and ceilings to encourage me to drink more beer.

Don't you worry, the GHz war is not done! There's talk of exotic materials (SiC, diamond, etc...) going to 10 GHz. If someone figures out how to make the Rapid Single Flux Quantum [wikipedia.org] digital chips with high temperature superconductors, then we may seriously start to see 1 THz clock speeds in practical computers, using extreme Peltier cooling to get the CPU core down to cryogenic temps.

The GHz war is over. The speed of light won. A long time ago, it stopped being "all about the transistor" and started being "all about the wires". IBM won the race to copper in 180nm (back when it was 0.18um), and that helped make those technologies even better, but about the time we hit 90nm, semiconductors were "fast enough", or even by some measurements stopped being able to speed up. Since then, almost all speed increases have been largely (but not exclusively) due to the transistors getting smaller, reducing the distance wires need to go.

The RC delay of wires is the major problem. R isn't going to be getting much better than copper. Silver has a lower resistance by a little bit, but it's too reactive to be used anywhere real. In these geometries, any alloy would be insufficiently mixable to be reliable, to say nothing about more exotic materials (like ceramics). There's some room for improvement in the dielectric (the "C"), but by the time you make a box with corners covering water permeability, thermal coefficient of expansion close to the wires, mechanical properties friendly to sub micron manufacturing, you have to concede you're not going to be able to get more than 20% faster there (and that we could dispute separately).

Take a cache. The slowest path is having a memory cell read. That tiny little device needs to have a measurable change in voltage on the bitlines, and be sensed by a sensing structure. That sensing structure has nothing to do with storage, so it's pure overhead and thusly you want as few of them as possible. Can you have it 16 bits away? 32? The days are gone that it was 64 bits away for any meaningful performance. There's nothing you can do to the characteristics of that little device (which needs to be minimum feature size to maximize the density of the cache) to dominate over the characteristics of the bitline he's trying to affect.

Take a data path. Even if 95% of your data is highly predictable, easily pipelined stuff with local signals, your critical path is going to involve signals from other areas of the chip, and they're going to have to be rebuffered and trucked from hundreds of microns away. No giant buffer in the history of man can dominate over a long distance wire. The signal will show up "eventually".

3GHz is a good place to stop. We make it to 4GHz with compromises in power, but beyond that and you're dedicating so much of your chip to rebuffering that you're blowing a lot of power on that. At that point, your pipeline is so many stages that branch mispredicts are very painful. You're devoting so much of your cycle time to setup and holds for your latches that you're going to be embarassed at how little work you can do in each cycle.

1 THz clock speeds are on their way, and maybe even higher. But they're not useful to CPUs or GPUs. They're useful for more exotic applications, primarily technology demonstrations.

I'd like to submit a vote for hooking this system up to a lawn mower. Automated lawn mowing is why I became an engineer, but I wasn't successful in finding a job doing it. If I could now purchase a lawn mower that would automatically mow between midnight and 3am and a rate slow enough to be nearly silent (a reel mower can do this), sign me up. In fact, this appears to be possible or even easier with the Kinect requirement of little ambient light and IR interference being a problem. I don't like current automated lawn mowers for various reasons, including the requisite subsurface wires and the noisy high RPM small blades.

My site at home has been under a distributed hack attempt (a long list of IPs all trying to ssh in as root*) for days now. ... Yes I know I could just change my ssh port and much of this would go away. But I find it amusing and I have bandwidth to burn.

I have a home server exposed to the wild internet by only port 22. It's an old machine, and it only allowed a single authorized user to log in, only with key authentication, not password. Nonetheless, the attacks would sometimes come in at such a rate that the CPU was pegged too high for the system to be usable for any of its primary functions (firstly being an Apache proxy through an ssh tunnel from work). I looked into a number of options to mitigate this CPU use, but none of them were as useful as using /etc/hosts.deny (the whole internet) and /etc/hosts.allow (my employer plus 192.168.1.*). I still get a few dozen logged messages every day to feel good when attacks are denied, but my CPU no longer gets pegged from authentication failures -- face it, denying an authentication doesn't cost much bandwidth, but it can take a few cycles to fail to authenticate a key.

When I buy a razor, the blade is what I'm really interested in and really paying for. It's where the real functionality is and what makes all the difference from one maker or model to another. The handle is just a cheap piece of plastic or metal. They don't even have to sell it at a loss, it costs a few cents anyway.

On its face, your statement is very valid and may be true. If packaging and transportation are the dominant factors, however, the far larger handle and "necessary" package for logos and eye catching (most aren't cheap monochrome boxes, there's almost always metal foil on them) my indeed be more expensive.

Now with printers, it's different. The printer itself is what has the real funcionality and what defines the quality, speed and other aspects of the output. The ink on the other hand is relatively unimportant. So long it's up to spec and doesn't clog the heads or smudge, any ink is as good as any other.

Sadly, printer manufacturers have no such warranty with regard to clogging or quality control. A single individual jet is almost guaranteed to fail within a few weeks. People shop based on reviews of a printer in the first few hours of operation and based on the sticker price. Honestly, neither of those are dominant factors in a printer that we want to last for 2 years and 3-4 cartridges that should last 6 months. The reality is that printers are sold at a loss, or darn near it, with the understanding that you'll buy at least one refill pack when you buy the printer, and probably another in 6 months before you get fed up and get a new one or decide to print it at the laser printer at work.

I didn't take a lot of Economy at university, but I thought corps are supposed to maximize their profits

Any economy course in high school teaches about the razor / blade model. Companies will sell you the latest Mach 7 Closest Shave Plus handle at a loss in order to sell you their replacable blades at a good markup. Perhaps it's more vogue on Slashdot to discuss ink jet printer manufacturers selling their printers at a loss (doesn't that start to explain the quality?) in order to sell you refill cartridges at a healthy profit later. Drawing the parallel between these "loss leader" models and the game system model is an exercise to the reader.

From Wikipedia: "As of November 6, 2008, the EyeToy has sold 10.5 million units worldwide.". 10.5 million sales is most definitely a success by any definition.

Interesting, though, that 10.6 million is a failure by every single measure.

You're describing what happens to fandom when expectations are too high. Do you recall hearing anyone criticize George Lucas before Phantom Menace ("Yoda was a bad muppet", etc)? Anyone who dared such a thing was immediately ridiculed. We were so enamored with the fictional universe the man had created he could do no wrong. Until, of course, he did.

Or perhaps the genius that was Joss Whedon. Firefly was pure art. The actors cast were perfect for their roles (and I say this while not caring much for Summer Glau). You don't create a long running show like Buffy without something going for you -- and then a spinoff like Angel that lasted a few years to boot. Then, with a gigantic "thud", Dollhouse. For whatever reasons that everyone had, Dollhouse had loyal viewers, but I'd never have considered them fans.

Fans will zealously overlook imperfections in order to better enjoy and advocate for the object of their affections. And then, when they're shunned, they have a habit of either just walking away or even turning on the remaining fans. You shouldn't be surprised that Warhammer (a franchise with 20 years of fans) had some people who were less than receptive to legitimate criticism. Nor should you be surprised when the fans who now feel scorned by their idols acknowledge the problems were there all along.

The physical difference between your uber cpu and a z80 is half a teaspoon of sand and some subtlety in the arrangement. You don't think you actually paying that much for the physical material in your processor are you?

No, I don't think I'm paying for the physical material in a processor. I think I'm paying for the time on the lithography machine to expose the 40-60 mask layers -- the most expensive part of wafer processing. Assuming Intel C4 bound instead of area bound, they were going to spend the same time here, so their costs are fixed. The second biggest component is wafer and module test -- the "locked" cores must have been fully qualified in order to advertise on the box that they are unlockable -- so Intel has already spent that money too.

Every cost that goes along with locked cores is already paid. This is a pure profit move on a piece of silicon I hypothetically bought and fully own. That doesn't set right with me.

I want to see machines following two parallel paths. Same internals in each, of course. I want to see an XBox 360, PS3, Boxee (or whatever) that looks ugly, can't stack for shit and has all these pointy lines or curves, and another device that is just as wide as my amplifier and is 10" or whatever deep with holes all over the place for good ventilation.

If you want to sell to children who have a table to put the thing on, fine. I have an entertainment center. I won't make room for some stupid form factor machine, unless it can be put out of the way (Wii). My DVD player, CD player, amplifier, etc. all sit in a nice stack. The way the XBox slim and PS3 slim were artfully redesigned on the inside makes it more than apparent that I should be able to buy one of those that's mounted properly in a stackable device with the fans pointed back. Not some kludge metal box that has the same game chassis inside, but a real front with the indicators and buttons on the front and real jacks on the back.

You misunderstand the point of the first amendment, and the founders' conception of rights. The first amendment does not GRANT rights; it merely acknowledges that the right to free speech exists, and constrains the federal government

Isn't that one of the big debates these days? Whether the Constitution needs to be read as a constrictive document, without any room for interpretation, or whether it's more of a societal framework intended to outline more than just a handful of federal policies? Sure, it says that congress can't infringe on those rights, and that's a specific thing. Implying that those rights are unalienable is a different concept -- while the Declaration of Independence holds a certain legal place in the formation of our country, I'm not sure of its place in the unalienable right for Liberty. Maybe the first 10 amendments were intended as a broad place to start from, with hundreds of amendments to follow every decade. Maybe they were intended to be interpreted liberally largely as Congress wished (within some checks and balances from the executive and judicial branches). Maybe they were intended to represent a tiny federal government that only barely held together a federation of states.

To some people, the Quran is just a book.

To some people, the American flag is just a piece of cloth. No problem burning it, cutting holes in it, using it as clothing, etc. even if all those things are disrespectful.

To some people, ethnic slurs have no malicious meaning. To others, they hurt more than violence.

You have no place in judging the value others place in any action -- including burning a specific book. Personally, I'm of two minds. I think we need to find ways to reach out to those cultures who appear offended by ours and help them realize we're simply people. We need to find common ground and work forward from there finding ways to eliminate hate through mutual understanding. There's quite a bit of self reflection we can do to realize why people in that region think we're in a holy war. I also think we've tiptoed around for far too long. Our culture is increasingly based on non-offense and tolerance, and not based on learning how to accept others' actions with a tough skin. We're in a time when even the phrase "those people" can have a negative meaning. Maybe it's time we all grew up a little.

Network Solutions used to be the place to go for domains. Now they are completely redundant.

I'd argue "irrelevant", not "redundant". If their prices were sane as they provided the same commoditized service, then they'd be redundant. In this day and age, the default parking provider should probably be someone like GoDaddy. If you have any content, stick with DreamHost or some dedicated colocation.

I've been a happy DreamHost customer since 2006 (when I relinquished control of a dedicated service on a commercial DSL). Yeah, I don't have root anymore, but for $120/year, I don't have to pay extra for the dedicated IP and I don't have to worry about applying those patches and being a "competent admin" either. I've noticed a sum total of an hour of e-mail downtime and I haven't observed any web downtime. Unlimited e-mail forwards is really nice -- I create a new address for every site that thinks they need one so I can track spammers to their source and delete that alias.

Latency was also reduced still further simply due to the masses of bandwidth FiOS offers compared to bog standard ADSL: in our case, 25mbps.

Damn it, kids, Latency and bandwidth are not the same thing and anybody who makes that mistake should be forced to use a "1Gb/s" connection via fedex.

If that's the stance you take, one might suggest you be forced to use a 150bps connection with a 1us latency. No, latency and bandwidth aren't directly related, but one who believes that the other is completely irrelevant in all situations has been misled. In this case, we're dealing with full frames of video being compressed and sent back to the customer. I wouldn't necessarily believe my 5MBps RoadRunner would be a neutral test for that situation. I don't usually get more than 400KBps for whatever reason, and when I'm pushing that much my other connections (ssh) experience additional latency. If the video is 300KBps, I'd think FIOS would indeed be a better test because of the additional bandwidth.

It's been a while since I worked in a data center, but time was we tried to keep ethernet saturation below 20%. Higher than that and performance went down noticeably.