Faster than Moore's law

Major Variola (ret) mv at cdc.gov
Thu Jul 8 19:08:52 PDT 2004


At 03:05 PM 7/8/04 -0700, Steve Schear wrote:
>At 09:31 PM 7/7/2004, Major Variola (ret) wrote:
>
>
>>At 02:55 PM 7/7/04 -0500, J.A. Terranson wrote:
>> >"A few years ago".  Lets call it two years ago.  That would make the

>> >average hi-cap drive around 30gb.
>>
>>Just want to remind y'all that drive capacity has increased *faster*
>>than semiconductor throughput, which has an 18 month doubling time.
>
>But access time has not nearly kept pace.  Which is why all manner of
>database architectures have been created to make up for this
shortcoming.

True 'nuff.  But DRAM is so cheap you put a few tens o' megs in the
drive
and cache the hell out of it.

And the busses on the x86 have gotten much faster too.

And the CPU itself is caching things up the kazoo.  Remember the
640KB days?    Its easier to add cache than to add a Montgomery
unit, though instruction-sets are evolving towards symmetric crypto
too.   The FPGA players are noticing all this too.  Get an Arm (tm)
by default, or synthesize a soft CPU,
and buy a third-party verified Montgomery unit.  Maybe even
get to add your own instructions to the basic CPU unit.

And yeah, you keep enough pointers around, you can have some
pretty fast DBs.  Dereferencing spaghetti.  Or dodder in the forest
of trees.

------
PS: if the TSA goons detect a tumor with their highres Thz scanners,
do they have to tell you?





More information about the cypherpunks-legacy mailing list