Conference 7.286::digital

>   For 45 seconds: 30MB/second x 45 seconds = 1,350,000,000 Bytes which 
>                        32
>    is A LOT less than 2   or 4,294,967,296. 

Since you're doing back-of-the-envelope calculations, only order of magnitude
counts, and you're well within that with this estimate.  Make it 24-bit 
color, and it's right on.

	Steve

Or consider that many systems only get to use at most half of their
address space for "user data", and frequently even less than that...
(Consider a VAX, where you'd be hard pressed to map even 1GB file,
since that would be all of P1 space...  Likewise, on MIPS, you don't
get the "negative" addresses for user data.)

    re .1
    OOPS!  Had my PC mentality hat on.  Lucky I wasn't using the
    information from my PC which is only capable of 16 colors.
    
    Thanks for setting me straight.  I now feel confident is quoting that
    information.
    
    	Regards,
    
    	 kam

    
     Then again if you compress it using MPEG 1 for CD-ROM, you can get an
    hours worth of video in to 650 MB that has a reasonable, but not great
    quality. Double the bit rate and it gets quite good. Using the CD data
    rate, you can get 6 hours of video in to 2^32 bytes.

And once you're talking about more than X seconds
worth of video, why do you have to be able to "address"
all of it at once?    You can read it just slightly
ahead of playing it (classic ring-buffer setup).

(Btw, addressability of >32 bits on disk is not tied to
 having a VA size of >32 bits or a native integer size 
 of >32 bits)

Of course, you need sufficient processing and I/O
bandwidth to read the data at the same time as you're
playing it.

    Absolutely.   Width of CPU-to-memory address has nothing to do with
    the maximum size of a disk file.  Don't use that as a reason to
    buy a 64-bit address computer.

    	Yea but now serve 50 to 100  movies from a single box memory
    address range becomes critical. I assume we are talking about video
    servers.

re .-1

Is one "video server" process serving all the users?  If so,
virtual address space is significant.  However, a design
which uses one address space (=one process) per user
could avoid that limitation, and is probably easier
to program anyway.

--

I remain to be convinced that 64 bits is much of an advantage
for this application.  However, sheer brute force processing power
*is* important, so buy an Alpha.

    I think the point of .0 was an analogy of the potential difference
    between 32 bit and 64 bit addressing.
    
    Early machines had a 16 bit address space, but it didn't stop them
    physically having or addressing more than 64k.  It just means the
    operating system and/or application had to jump through hoops to work
    properly.  Examples - PDP8, PDP11, Intel 8086 and so on.
    
    Larger address spaces are easier!
    
    I have also seen the difference between 32 and 64 bits described as the
    difference between being able to enumerate the people on Earth, or
    every subatomic particle in the universe.  An analogy, right?
    
    Martin

    
    	Isn't the point with 64 bit addressing that 32 bit was the "norm"
    and users are currently approaching the need for 33 bit addressing? 
    The next technological step after 32 bit is 64 bit which WE can supply
    6 years ahead of our nearset competition!!!!
    
    				Malcolm.

    Re .8
    
    If several users want to see the same data at the same time (or nearly
    so) then shared buffers would be a benefit, and the virtual address
    space to map the buffer pool would be significant.
    
    Peter

One lesson from the PDP11 is that having physical address space greater
than virtual address space is a major hassle.  The VAX reached that stage
a few years ago.  (Earlier, physical addressing was limited to 32 bits or
less, but that is no longer the case.)

That said, 64 bits is not a universal good thing, as Jeff Mogul has pointed
out.  It would be good for compilers to offer the option to use 32-bit types
rather than 64 bit types where appropriate for the application.

	paul

>    I have also seen the difference between 32 and 64 bits described as the
>    difference between being able to enumerate the people on Earth, or
>    every subatomic particle in the universe.  An analogy, right?

FWIW - 2^64 isn't even close to the number of subatomic particles
in the universe (maybe 2^128? :-) ).  (A 2^64 byte memory "chip"
could be fabricated from a single gram of matter with >100 atoms/bit -
or on the order of 2^80 "traditional" sub-atomic particles just
for that "chip".)  Of course, even just reading the data out of such
a "chip" in a timely fashion would be "a challenge" for today's
technology (and I sure wouldn't want to sit around waiting for
a sequential memory test... :-) )

		6.02252x10^23 atoms/mole * 1/28.086 mole
1 gram Silicon: ---------------------------------------- = ~145 atoms
		    2^64 bytes * 8 bits/byte

as long as we are going on with this little amusement, what would it
cost (at current prices) to buy that much alpha memory (even if it
won't fit in the backplane)? How many backplanes would it take?

	Joe

1.25Gb/in3 isn't going to make a 2^64 memory feasilble any time soon (that
would still be something like a 60x60x60 kilometer cube of memory... ;-) ).
However, keep in mind in the last 40 years we've gone from 4K drum memories
to 1Gb in 8 Type I PCMCIA cards...  (If the same progress is made in the next
40 years, that would be very close to 2^64 bytes/meter3.)

VNS TECHNOLOGY WATCH:                     [W. Stuart Crippen, VNS Correspondent]
=====================                     [Acton, MA, USA                      ]

                Technology gives memory densities huge boost
                --------------------------------------------

        From EDN, August 18, 1994, Vol. 39, No. 17, Pg. 18
        Author - Gary Legg

    A new assembly technology for memory wafers and dice promises
    dramatically higher densities for packaged memory devices - ultimately
    as high as 1.25 Gbytes/in3 with 16-Mbit DRAMS, compared with only 40 to
    80 Mbytes in the same volume with conventional SOJ or two-sided TSOP
    packages.  The first products using the technology will be available
    this year.

    The technology achieves the higher densities through its method of
    stacking and connecting wafers and dice.  A key feature is a patented,
    pyramid-shaped via through the silicon.  The small opening on the top of
    the pyramid penetrates the silicon on the circuit side, and the large
    opening comes through on the back, where the interconnect makes contact
    with a number of circuit elements on the silicon directly below.  This
    interconnect method takes no space because all pins fan out under the
    silicon stack instead of residing around the perimeter of the active
    silicon.

    The first products to result from the technology, due in the fourth
    quarter, are 88-pin, Type I PCMCIA cards with 16, 32, 64 and 128 Mbytes
    of DRAM.  Expected prices are $1995, $3995, and 7995, respectively.


<><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><><>
  For information on how to subscribe to VNS, ordering backissues, contacting
  VNS staff members, etc, send a mail to EXPAT::EXPAT with a subject of HELP.

    Permission to copy material from this VNS is granted (per DIGITAL PP&P)
    provided that the message header for the issue and credit lines for the
    VNS correspondent and original source are retained in the copy.

<><><><><><><><>   VNS Edition : 3159     Tuesday 20-Sep-1994   <><><><><><><><>

    Did anyone else see the new ads from Atari for the new Jaguar game
    system.  It's 64 bits - the ad has a way of making it seem important,
    and ends with the slogan: "DO THE MATH".  Don't know where they get
    their chips, or even what 64 bits means to them - but can we borrow the
    slogan and ride the tide?
    
    bill

They get their chips from IBM (not a general purpose microprocessor).
The tagline is a good one for their market.

				Steve

> The tagline is a good one for their market.

What, a bunch of pimply youths who probably *can't*
do the math because they've been mentally mutilated
by what passes for culture these days?

RE: .19

You're a generation behind.  Gates is no longer a youth and hasn't 
been for 15 years.

--PSW