Great explanation.

The reason is simple : most of the large PC manufacturers write contracts with the parts manufacturers so that they can buy up to a given number of stuff at a predetermined price. Not only disks, but also DRAM (price known to fluctuate wildly), flash memory, capacitors, etc etc etc. These are the guys who are dating those 9 out of the 10 girls.

I'm actually quite surprised that ASUS didn't secure a stable supply chain - now they have to shut their business down simply because they can't get enough disks.

So, essentially, what he is saying is that 'under first-to-file, the inventor has to patent every invention instead of just the useful ones, because it takes time to figure out which ones are useful'.

Well, here in South Korea, the patent law is first-to-file, but the inventor may publish it, and STILL can file the patent within six month since being published. Have a nice idea? publish it, and for the next six month, you have the exclusive right to file the patent of the idea on the publication (of course, as long as the publication itself doesn't have any prior art). If you found out that it is a bad idea, just don't bother filing a patent and still, nobody else will be able to file one since the publication itself will count as prior art.

Plus, here in Korea, we have 50% discounts for small businesses, independent inventors, universities, students, etc. (read: anybody without a deep pocket). The last patent I got here (which was filed three months after being published on an international conference paper) cost something like $2.5K including the cost of hiring a patent attorney.

My idea was:

CPUs/GPUs :
Yeah, they used a off-the-shelf design for the GPU, but I'm not sure if Cell can be called an "off-the-shelf" processor - the problem was the SPEs were a completely new design, which not only requires a whole set of new compilers & software infrastructure, but which actually needed a completely new research field on compiler design. Because the SPE could only address the limited amount of scratchpad memory (external memory could only be accessed via the DMA) that added so much headache on software design, which could only be done by a crazy amount of hand-tuning. Add other typical VLIW scheduling problems, and you have a ton of headache optimizing software.
As a result, they couldn't rely on off-the-shelf software & tools, and had to design their own software stack & tools, which had zero guarantee to have better quality than proven-to-work tools.

DRAM :
The problem of DRAMs is that it's a ultra-cheap commodity product, where one design gets manufactured in billions. Once you turn to something nobody else uses, it turns extremely expensive - not only the manufacturing cost, but it becomes difficult to deal when there are supply issues (e.g., nobody else can supply the DRAM if your primary manufacturer gets hit by an earthquake/tsunami/strike/thermonuclear war/whatever) which adds a lot of risk.
Using these low-volume products on limited places will make the problem even worse because it will further reduce the volume. I'd rather have less performance than use a low-volume part which can potentially add a lot of risk on mass-manufacturing. Additionally, if you want bandwidth, you always have the possibility to add more parallel DRAM channels - although it may be inefficient & costly, I'd rather choose that.

Blu-ray:
What I meant was that the PS4 shouldn't use something beyond Blu-ray. One of the biggest problem of PS3 was that they tried to use a next-generation optical medium (instead of the proven solution) and thus, had a lot of manufacturing problems. Now that Blu-ray is quite stable and affordable, I guess PS4 should use Blu-ray (yeah, probably a 4-layer design) instead of developing something else.

Or, put it the other way : for the last ten years, every PC manufacturer moved to just a normal gaming console, albeit capable of running any generic operating system, and being capable of running generic software quite well. Actually, most modern PCs got all those "custom circuits" (GPUs, sound processing DSPs, vector instructions, etc.) which previously only existed inside those custom chips inside the consoles.

What I thought was not just something of 'generic'ness, but that Sony (and IBM) took a far too ambitious goal, which failed miserably. Their initial though on Cell was that it should become a generic processor which can be used for various home appliances, supercomputing, and possibly other embedded applications. Their intention was to have PS3 to be the initial Cell customer, and find many other customers later to cover the development cost.

However, the problem of the Cell Broadband Processor was that it was too generic to be used for games (since it must also be capable of running HPC or home entertainment applications), while being too difficult to use properly. Thus, they failed to find a customer other than the PS3, and as a result, the sales of PS3 had to cover the development cost of the whole Cell project. The final nail in the coffin was that IBM killed all future Cell projects, probably because they couldn't find any future customers.

In short, Sony and IBM's goal was to create a new "general-purpose" CPU, which failed miserably. The issue wasn't about "generic PC" vs. "custom circuits for gaming" : it was about "generic PC" vs. "a different generic PC".

Sony didn't say that they are going to produce a less powerful design, but a design which costs less, in terms of investment.
Although the outcome may be a not-so-powerful console, the other possibility is something with less "custom" solutions.

Such as:
  - Off-the-shelf CPUs/GPUs, or custom ASICs using 3rd-party licensed CPU/GPU designs (instead of designing one from scratch)
  - Off-the-shelf DDR(1/2/3/4/5/whatever) SDRAM (instead of using something from Rambus)
  - Blu-ray, instead of a new kind of optical disk design (or, even eliminate the physical medium altogether in favor of online purchases)

...and therefore, most of the people in the world won't need anything beyond an Intel Atom because their tasks aren't compute-bound.

Seriously, I don't know what kind of code you write for a living, but the code I write is almost always has some portion of compute-bound submodules, even if what I do has nothing to do with video codecs or 3d or whatever field that there are convenient libraries.

Additionally, an average server has 2x cpus, 8x memory, while having 0x graphics compared to an average desktop. Another problem is that we are running out of tricks for reducing dram power, which means that the portion of dram power may increase steadily in the near future.

Even graphic cards have a sizable, high-bandwidth ram on-board.

Trust me, DRAM power consumption is becoming a serious probpem.

A common misconception is that people think that the Android market from Google is essential and irreplacable. However, I find that there are plenty of ways to replace Android market with something that may work.

The Android market may be fine and essential for people who live in U.S, or at least, many English-speaking regions. However, my experience is that they are a somewhat half-baked solution to non-English speakers. Don't speak English? all you have is a bunch of apps written in some indecipherable language, or some wierd application with machine-translated, cryptic text.

For example, in Korea, there are zero games avaiable on the Android market, and zero paid apps published by the Koreans. Until recently, paid apps weren't even available to the Korean market. Although half of it isn't Google's fault (e.g., in Korea, games are required to have a rating to be sold.), but that doesn't matter. On the other hand, the T-store from SK Telecom (the largest telco in Korea) does have a functioning Android app store which actually provides localized games, localized contents, and applications that are written by real Korean-speaking developers. Since the Korean developers cannot register paid apps in the Android market, they upload the free version on both the T-store and the Android market, and leave a note on the Android market version as something like 'if you want the paid version, search on T-store'.

Now, this is the situation in Korea. Imagine what can happen in China, which may require even more localized content, a potentially huge installment base, and many people who doesn't speak English. If Google fails to provide a localized market, and Baidu does, suddenly Android without Google starts to make sense.

Okay, I think I understand what's going on (at least, what Microsoft is claiming).

1) The sunspider cordic benchmark is actually a huge dead code : runs something, throws the result away, and just measure how long it takes.
2) IE9's javascript engine has a dead code elimination engine. However, since analyzing large amount of code for searching dead code is expensive, they have a parameter specifying how far they will search for dead code.
3) The engineer who was looking for the optimal parameter number simply looked for whatever benchmark he/she could find, and tuned the parameter which yields the highest score,
4) ...which happened to be precisely the size of the cordic benchmark.

The claims on the blog has a point : sunspider is a bad benchmark. Not just because it is stupid to write some code that computes sin() in javascript when you can get super-fast native implementations, but because THE WHOLE BENCHMARK IS A FRIGGING PIECE OF DEAD CODE.

Choose my build order. Rock, paper, or scissor?

However, books with good stories also have a 20-40 hour experience, and somehow still can be compressed into an hour thirty..??

Even if the patents that Google need to defend against are really crappy, it may worth having some ammo because bogus patents still can be used for suing, while zero patents can be used for nothing.

Moreover, invalidating bogus patents is quite expensive, risky, and time-consuming.

OTOH, how many patents does Palm have, and how many of them are valid ones?

Wikipedia to the rescue:

http://en.wikipedia.org/wiki/Micron_Technology

They split their PC manufacturing business into a spearate company, which declared bankruptcy in 2008. Now, they focus on manufacturing memory.

To most of the people, Micron is known as their consumer brand Crucial Tehnology and Lexar Media.

Well, I think this would be the article Zed needs to read:

http://www.joelonsoftware.com/articles/fog0000000332.html

Basically, many programmers feel that everybody else around him(or her) is a stupid asshole. However, if you want succeed, (e.g. have everybody around you learn statistics) you should never, ever, ever make enemies.

Be productive, work hard, listen to others, and try to do the work in the *right way*. Gain respect from yor collegues, and then they will get interested.