Bilow's very first Bash on this thread... post #1032.
To: dougjn who wrote (1030)
From: Bilow
Wednesday, Aug 13, 1997 12:52 AM ET
Reply # of 43681
I'm an EE, wondering what all the hype in this stock is about. Here
are some questions regarding RAMBUS and RMBS:
(1) I wasn't able to find a RAMBUS chip spec. Can somebody
save me the trouble of searching the net? By the way, my current
project uses SGRAM and I was able to find spec sheets from 10
more or less pin-compatible suppliers on the net in just a few hours.
(2) How does RMBS claim that they are providing 10x performance
improvement on current DRAM? SDRAM (and SGRAM) already
burst with 7.5ns periods. Seems to me thats only a 5x performance
improvement.
(3) I can expand multiple SDRAM/SGRAM chips on the same data
bus. I suspect that with a 600MB/sec bus, this is impossible. If so,
this means that I lose the pin-out advantage on the memory
controller if I have to design for expansion.
(4) What about the fact that historically, RAM makers have been
incredibly conservative with new interfaces. The current interface
is only a mild change from the original 1Kx1 DRAMS that required
-5V, +12V, -12V and ground (the power supplies were a minor
nightmare.) For instance, VRAM never gained much of a market
share, despite the much faster access rates.
(5) As far as pin-out advantages of RAMBUS goes, note that
VRAM sells for a very small price over DRAM, despite having twice
as many data lines. (and extra control lines too. The industry
standard part is now in a TSOP-100) I suspect that RAMBUS DRAM
will not trade cheaper than SDRAM for many years. Given this, I doubt that many cheap computer makers will
convert to RAMBUS
until long in the future.
(6) If I have the RAMBUS interface specification, I can design a
controller for the chips without paying a dime to RMBS. I simply
design without their technology. If I am a small guy, they will never
find out whether I violated a patent. If I am a big guy, then my
corporate lawyers can keep theirs busy. In any case, it is impossible
to patent all the ways of doing something.
(7) If I am a DRAM maker that has not signed an agreement with
RMBS, what is to stop me from making chips that are compatible
with the RAMBUS interface? All I have to do is design them in a
way not covered by the RMBS patents. Then I can sell my chips
to computer makers (and anybody else) without having to pay the
royalty.
(8) In light of points 6 and 7, how is that RMBS can possibly charge
a very large royalty? It seem that if they do, their customers will lose
out in a competitive environment.
(9) New chip packages all have the same similarity: Lots of pins.
Given this, what is the big advantage of a smaller number of pins?
Recent projects that I have been on (Virtual i/O) ended up with
unused I/O because of large die sizes. I have the suspicion that
RMBS is providing a solution to a problem that only occurs to a
small minority of the DRAM users.
(10) The one place where a high speed memory interface seems to
me to be useful is in video display cards. The classic problem for
memory designers (and I designed memory interfaces for APPN,
RENX, and ZMOS) is too tall memory. That is, you end up too high
a ratio of memory capacity to memory bandwidth. RAMBUS will
certainly help there, but SGRAM has already solved that problem.
The bandwidth out of a single -10 SGRAM chip peaks at 400MB
per second (32-bit bus), and -7.5 parts are available. Furthermore,
the parts are double banked, and actual performance in the area
of 300MB/sec is easy to achieve. Furthermore, they are fully
synchronous, so there is no need for the asynchronous interface
that I suspect RAMBUS requires.
This stock doesn't look like a good bet to me. Perhaps some of the
more technically knowledgable of you can help me see the value
I am missing here. I took a look through www.edgar.gov, and it
seems that the stock price is incredibly, unbelievably, horribly,
over-valued.
-- Carl |
