| Tuesday,
April 13, 1999 I shall do my taxes today, since tomorrow I have to go down
to Internet World and see what's new there. Meanwhile, a new way to waste my time:
I have very little experience with the Bank of New York, whose letterhead proclaims it
"New york's first bank -- founded 1784 by Alexander Hamilton," but what
experience I have had makes me anxious to avoid more.
It began with a letter dated April 1, 1999 and arriving here on April 12 after close of
business in New York. It was addressed to myself and my wife as joint tenants, the form of
the address making it probable that the names and addresses had been obtained from the
Prudential, which acted as my broker for some years. The letter says that "Our
records indicate that we are holding funds belonging to you… In accordance with
abandoned property regulations, failure to contact the Bank will result in your monies
being escheated to your state of residence.
"To claim these funds you may contact out Customer Service Area [gives an 800
number] or return this notice … no later than April 12, 1999."
Well, since their notice didn't get to me until afternoon of April 12 and they closed
before I got it, I couldn't do that. So this morning I telephoned the 800 number. After an
unconscionable number of electronic messages presumably designed to get you to hang up and
stop bothering them, I got a human being on the phone. I presume human, and male; I don't
think artificial intelligences are up the conversation I had.
He wanted to know what this was about. I pointed out that I didn't have the foggiest
notion because other than in my history books I had never heard of the Bank of New York
and had no reason to understand why they had any of my money. The letter didn't say. I
gave my social security number (which wasn't supposed to be a national ID number, but
which seems to have become one). No joy. Way down at the bottom of the letter, off in one
corner, is a number that appears to be a form number, but it was #0222xxxx. I read them
that number.
Enlightenment. The bank can't find me by name or social security number, but this is an
account number. Why it is in small print down in one corner of the page, and isn't labeled
"account number," isn't known to me.
I had, until March 27, the grand total of $7.53 (Seven Dollars and Fifty-three cents)
in their bank. On March 27 the money was turned over to the State of California. On April
1 the Bank of New York sent me a notice that they would turn the money over to the State
of California if I didn't get to them by April 12; but they mailed the letter in such a
way that I didn't get it until April 12. That letter was the first indication that I had
any money in their silly bank.
Now, they would be willing to give me a telephone number (not toll free) that I could
call to try to get California to give me my money back. Wasn't any responsibility of the
Bank of New York, New York's oldest bank founded 1784 by Alexander Hamilton, no sir, not
no way.
Since I had used up more than $7.53 worth of my time, I declined to go try to persuade
California Treasury bureaucrats to give me back my money.
Someone should tell the Bank of New York about computers. The letter that told me of
this account could as easily have given me the amount, saving us all some time.
The letter is unsigned: where there might be a signature it says "The Bank of New
York, Trustee." I am not sure what Trustee means, but apparently it does not mean
that the Bank of New York will inform the account holders about money paid into the Bank
of New York, New York's oldest bank founded 1784 by Alexander Hamilton, as part of a class
action court settlement until it is too late to collect it. If I ever need a trustee in
the State of New York I will remember that. The sum was trivial and California is welcome
to it. The waste of my time wasn't trivial, and I resent it.
The one clear message from WinHec
was that the limiting factor in small computer speeds is no longer CPU speed, but memory
speeds. This is because raising the CPU speed without raising the bus speed is getting to
the point of diminishing returns. One needs to make the whole system run faster: not just
internal calculations, but data transfer from Level 2 cache to chip internal level 1
cache; faster snooping to see what to put into that cache; faster transfer of calculated
data to the display; and so forth. Most of this is accomplished by raising the bus speed.
For some time now PC motherboards have operated at 66 MHz, but 100 is becoming
increasingly popular. Bus speeds at 100 require PC-100 certified memory, which costs more
(but not a LOT more) than standard memory. Now the tendency is to even faster bus speeds:
133 MHz seems to be the next plateau where we'll pause for a while.
Intel is backing a system known as RAMBUS, which I won't explain here: the important
fact is that works at 133 bus speeds and is capable of operating at high transfer speed
rates, but does little about the latency. Loosely defined, latency is the time it takes to
get the FIRST byte of data, while transfer rates are concerned with how fast the rest of
the data flows.
Traditional memory doesn't work well at 133 MHz; most PC 100 chips will begin to show
soft errors at 120 or so. PC-133 chips exist but they're expensive. Rumor has it that one
reason they are expensive is that some PC box makers are buying up PC-133 memory in order
to hedge their bets on RAMBUS. They are doing this because RAMBUS may cost more than was
thought, while so far performance improvements have been disappointing, in the 15 to 40%
improvement range, not the 2x to 5x that we keep hearing about.
This may be because a great deal of memory transfer is from the program in main memory
to the Level 2 cache, and the latency of memory calls can be more important than the
sustained transfer rate. At this point we get into rocket science, and I don't much trust
the theorists: at this point I want to see hard benchmark numbers, and I have not seen
them.
All this may be rumor only, but I have heard it from more than one authoritative
source. The problem is that sources tend to have attachments. Those attached to Intel
don't show me test results, but they do point out that Intel isn't going to support other
kinds of memory: they are committed to RAMBUS, RAMBUS is good now and will be better soon,
and competition will force PC-133 RAMBUS memory down and down until it costs no more than
ordinary memory does now. How long it will take for Moore's Law and competition to work
this magic is not so clear.
Those attached to non-Intel chip makers point out that they can make fast PC-133
machines with traditional but upgraded memory cheaper than Intel can make RAMBUS machines,
and for AMD at least their chip speeds are competitive. Non-Intel systems made great
inroads into Intel's market share down at the lower price end of things. Intel doesn't
like that and wants that market back. RAMBUS may be a big factor in price.
There is, in a word, a lot of FUD. (Fear, uncertainty, and doubt.) It is not, I think,
deliberate rumor mongering. It's quite real uncertainty. In particular, the performance
enhancements of the new systems with RAMBUS may not be enough to justify the increased
prices, which can be substantial. What people will pay for 3x they won't pay for 1.3x.
And for the moment, here endeth the epistle. If anyone knows a lot more -- particularly
if anyone has hard data on performance and price -- they're not talking to me.
Me, I want things to go well. I'm rooting for RAMBUS, which I wrote about in the old
BYTE several years ago as a promising technology; we may even have given it a technology
award. We certainly considered it. But the proof of the chipset is in benchmarks…
We have mail on this.
I have opened a new discussion page on GULF
WAR SYNDROME sparked by some revelations coming out of Tulane and published in Insight
Magazine this week. It will be an ongoing story. There is plausible evidence that it is
real, and that we know the cause. How much of that is true I don't know. The waters are
very muddy on this. My own interest is as a novelist, I suppose: there's a fascinating
story in there no matter how it turns out.
TOP |
CLICK HERE for
a Special Request. 
