Re: Why are TI Calcs so inferior?

[Prev][Next][Index][Thread]

Re: Why are TI Calcs so inferior?

On Sun, 17 Nov 96 04:13:55 -0400, you wrote:
>> Totally different technologies.  It's nonvolatile static ram.  Regular
>
>Static RAM is NOT "none-volatile" and it is NOT the memory of choice for
>calculators!!


I didn't say that it was regular static ram.  I said it was
NON-VOLATILE static ram.  And yes, it is the ONLY form of memory that
most calculator's use.  They don't use flash, dram, cache, bubble,
masked, fprom, or any other kind of solid-state technologies.  If it's
not the memory of choice for calculators, what is?  I doubt you can
prove me wrong.


The HP100LX and other pcmcia palmtops do use flash-file storage for
some things, but it's not used as the main battery-backed storage.


>> computer ram (simms) is not the same, and it's not even close to a
>> hard disk.  Computer ram is a form of DRAM (dynamic ram), which uses a
>> capacitor to store each bit.  Static ram, on the other hand, uses
>> (about) 6 transistors to store a bit.  The latest chip manufacturing
>> processes only allow for about 1 to 2 million transistors on a single
>> chip, so that's why high-capacity static ram is so expensive.  If
>> someone were to come up with a way of making static ram as cheap as
>> hard disks, a new age in computing would result because NO ONE would
>> even think about buying the "slow" hard drive.  Static ram can be kept
>> intact on very low power, whereas dram requires refresh circuitry and
>> takes up too much power.
>
>You are totally Wrong about the power consumption of power by the two
>types of RAM.  Static RAM consumes much more power then DRAM!  That is =
the
>reason that DRAM can be packed to high densitys on a single IC and =
Static
>RAM can not even realize a fraction of this density because it would FRY
>itself with the heat build up from all the power it consumes!  And who
>told you that static RAM is none-volatile?  It is not!
>
>Don't rave on about a subject till you are qualified!


Ummm, I'm an electrical engineering student at Michigan State
University, and I've designed and built a number of microprocessor
based computers, and I'm currently developing the 4 megabyte memory
Expander for the TI-83/85/92 (you can visit my webpage), which now
works BTW.  You're obviously not an electrical engineer or anything
close to one, so I think I'm more qualified than you :).  You're the
one who's raving!


And no, you're wrong about power consumption.  Go to www.dalsemi.com
and look at dallas semiconductor's nonvolatile static ram chips.  I
didn't say that all static ram is non-volatile, I'm saying that most
static ram these days is made non-volatile with a battery-backup (with
the exception of cache ram).  On standby, low-power static ram chips
draw less than a few MICROAMPERES.  That's why it can be battery
backed with a small lithium battery.  The latest non-volatiles sram
chips hold their data for 15 *YEARS*.  Dram, on the other hand, easily
draws 10,000 times MORE power than that during standby, and probably
more than that during use.  A lithium battery could never source that
much current for more than a few minutes.


In terms of heat, I don't think you've got your facts straight.
You're thinking about microrpocessor chips, that have such huge
transistor densities that they'll overheat.  Static ram, on the other
hand, never has it's transistors all in use at once.  Therefore, it
won't produce much heat at all, and it allows itself to draw extremely
low power.  Cache ram is slightly different.  Due to the extremely
high speed, it ends up drawing as much (if not more) than dram.


The reason dram requires so much more power has to do with it's
operating theory.  Each capacitor must be kept charged to the correct
on/off state.  Since the capacitors are so small, each one must be
refreshed almost 1000 times per second.  Special dram refresh
circuitry (that works in the background) has the daunting task of
keeping each capacitor up-to-date, otherwise information would be
lost.  That refresh circuitry must ALWAYS be on, no matter what.  This
is why dram is NEVER used as solid-state storage, like flash eeprom,
regular eprom, low-power static ram, mask prom, etc.  If static ram
could be made with very cheap processes, dram would become extinct in
a second.  This is also why *NO* small battery powered devices (like
calculators or PDAs) use dram.  You can't name one that does.


>Do you also know that static RAM is also used in computers?  Main memory
>is usually Dynamic RAM and cache memory is usually high speed static =
RAM!


Umm... yep, I knew that.


>> This is also why a modern pcmcia 2mb ram disk costs almost $200.
>> Static ram is very low power, no moving parts, extremely fast (easily
>> 10 times faster than the fastest hard drive) and very small.  You get
>
>Let's see....DRAM at 60-70 nanoseconds and a Hard-Drive at 10
>milliseconds......where's the TI-85?  That would be 142,857 times as
>fast!


I didn't say that static ram is faster than dram.  Some of the newer
static ram is, but all in all static ram is about the same speed
(well, due to the refresh latency times, static ram is usually faster
given an access time).  What I said was that static ram is much faster
than a hard drive.  In the case of the TI-85, I'd say that the access
time of the ram module is 100 ns. Ten times 100ns equals 1000
nanoseconds, which equals 1 microsecond.  A 100ns access time gives
you transfer rates of over 10 megabytes per second!  That's faster
than the vast majority of hard drives.  Go up to cache-ram, and you've
got 100 megabyte per second transfer rates.  And yes, given a
mass-storage memory request, the TI-85 IS FASTER THAN A PENTIUM, that
is, a pentium accessing the hard drive takes 1000 times as long as the
TI-85 takes to access it's memory.


>The reason that a calc is a rather slow device is that it's routines are
>coded in masked ROMS, slow compared to DRAM, and it is an interpreted
>language vs the faster compiled code run in DRAM on computers.  The
>computer requires a ROM BIOS for it's IO routines, but that is why they
>use Shadow RAM to transfer the ROM code into and speed up execution.


FYI, masked ROMs are usually as fast (if not faster) than DRAM.
Masked ROMs are used in virtually everything, including video game
cartridges (this is why the new nintendo ultra-64 kicks so much ass,
because the processor accesses the cartridge at 50-60ns speeds).  I'd
say the latest masked rom's are much faster than DRAM.


The only reason why rom bios is shadowed is because of the speed of
the processor.  That is, even though the BIOS probably has an access
time of 100ns, a pentium is literally twiddling it's thumbs waiting
for the bios to respond.  The same thing happens when it accesses the
main ram, even though it's a "fast" 60ns.  The bios routines are so
important for performance that it's shadowed into a form of cache ram,
which has access times of < 12ns, which significantly improves
performance.  This is why a L1 cache is so expensive to manufacture,
because it runs at nearly NATIVE speed, with transfer rates almost
exceeding 1000 megabytes per second.  As soon as the technology gets
better, L1 caches of 512k will be possible (regular caches are L2,
which is slow compared to L1 caches which cannot be upgraded).


-Mel


<pre>
--
The TI Memory Expansion Homepage
http://pilot.msu.edu/user/tsaimelv/expander.htm
</pre>
References: