Actually, many of them have a good basis in logic, but are used beyond their original purpose.

Part of the reason there are individual companies that separated items like brokerages and commercial banking is historical structure created in the Great Depression, known as the Glass Steagall Act

Other responders to your post have pointed out various specific details, e.g. reason for commercial paper, but let me cover a more general point of view on why so many different products exist: Risk.

The issue, however, is that risk doesn't come in only one form. There are different types of risk:
- Default Risk (if a company goes bankrupt, you don't get back your principal)
- Inflation
- Interest rate risk (if interest rates change, then the value of underlying loans change)
- Tax Rate Risk (different tax rates due to different income or time)
- Counterparty Risk (risk of entering into a contract, but the other party failing to fulfill the contract)
- Secured nature of the debt (recovery in case of default)
- Opportunity cost (cost of not doing an alternative with the money)
- etc.

Looking at various products we can see how they are different. An IRA vs. a Roth IRA actually transfers the Tax Rate Risk onto the government (you pay a known tax rate, and the unknown benefit or penalty due to the future difference is absorbed by the government)

A TIPS (Treasury Inflation Protected Security) vs. a normal Bond issued by the government transfers inflation risk onto the government (presumably the normal bond is accounting for perceived inflation in the offering price, but the TIPS accounts for real inflation, thus allowing one to eliminate the risk of the perception of future inflation being incorrect.

We can see today that today's 4 week Treasury Bill Auction resulted in zero yield (give money to the government for 4 weeks, no interest). This presumably would mean the return one could get in a non-FDIC insured bank account (over the current $250K limit) is entirely bankruptcy risk premium.

Also there are organizations that do market clearing of bonds and stocks that absorb counterparty risk. Part of the problem with credit default swaps was that the holders of those products actually bear the counterparty risk, as they are not regulated like other products. (When combined with a lack of market data on quantity and concentration of the risks around default of bonds, this led to one of the underlying issues in the problems we have right now)

Lastly, there are also "positive" values that are priced into different financial products, such as recovery in case of default. That's why secured loans of [statistically] appreciating assets (e.g. home mortgage) are lower rates than loans on depreciating assets (e.g. automobile) and those are lower than unsecured personal loans. Same reason bonds will maintain value longer than preferred stock.

As a specific example of what is good (and bad), look for a moment at interest rate swaps. They actually serve a valuable purpose, in allowing an investor (or the loaning company) to convert a variable-rate instrument into a fixed-rate one (or vice versa). This is valuable to companies to be able to "lock-in" lower interest rates when rates fall, for example. What is risky is when someone speculates on interest rate swaps without having an underlying asset. This results in significant leverage that can be wiped out very quickly if interest rates (in this case) move unexpectedly in the "wrong" direction.

If you lay out the different forms of risk (to both parties in a transaction) and level of security in a loan, I'd suspect you'd find very few financial instruments that significantly overlap (There will be a few, but in many cases, those can be differentiated by size, e.g. commercial paper vs. personal loans). The problems come in speculating on items to make money.

You *can* integrate memory and CPU on the same silicon die, but the overhead to do it in terms of additional error rates and processing tasks makes it economically inefficient to do so. (It is more economically efficient to build a larger cluster). It's possible (I'd even hedge likely) that we will see 3D packaging technology try to get around some of the latency. (More in my post a few years ago). However, the overall difference in speed between memory and processor is unlikely to change in the near future, so we need to continue to architect around that limitation at a system level, rather than a chip level.

The other thing I'd point out is that your analogy to "balanced" general purpose computing systems can (and should) fail for supercomputers... there is no rational reason to continue scaling in a linear fashion.

Then again, this is seriously old news. Trying to optimize a supercomputer to get anywhere close to 100% CPU utilization is known to be a problem. Others have already pointed out IBM's Blue Gene, and there's a reason it's a good example.

From Marc Snir, et al. at IBM, September, 2001 in a file called BlueGenePublic.pdf, which discussed the design philosophy for the Blue Gene supercomputer:

"Standard microprocessors are optimized for running as fast as possible one instruction stream...
Standard nodes suffer from 'von Neumann' bottleneck: computation speed increases much faster than memory access speed"

"Let's think from scratch.... in order to build general purpose systems that overcome constraints of conventional architectures.
Let's accept that significant improvements in cost/performance can be achieved by building an 'unbalanced' system" (emphasis added)

"CPU is a vanishingly small fraction of total system, in silicon area, or power
It is rational to build systems with a surfeit of compute power, so as to reduce memory requirements and reduce the need to move data around
It is cost effective to have a low CPU utilization"