Re: F21d - Video, Mup21 instructions.
Hi Wayne,
>>>>> "Wayne" == Wayne Morellini <waynemm@hotmail.com> writes:
> Dr Ting had a manual for Mup21 out, it might be on Jeffs
> shopping site: www.dnai.com/~jfox.
> Jeff, good news, that 0.35 micron process (or 0.18 Micron
> process) might be good for a second edition of the chip.
Seconded..
> More graphic colours and sound would be good (but this is
> I21 type matarial). All we really need is preferably: video
> bus, memory bus, USB (or Firewire bus for I/O) buss, and
> legacy parallel/serial/Analoge and microwire (borad level
> electronics buss) ports. Almost any thing could then be
> plugged in.
Sounds like a good wish-list to me...
> I think we however are getting beat on the memory side of
> things, a lot of memories are being designed for sequential
> access as backups to cache systems, solutions:
> small on chip cache
> - A snow flakes chance in ..
> in package memory die
> - A reduction in headache for Chuck designing memory
> modules and less
> charging time on the pads.
I like the idea of an MCM job for a number of reasons, not least the
throughput. I'd guess with the size of a 0.35u F21 the biggest package
problem would be the cache yield ? (assuming home-grown cache dice)
*dream* maybe 2 F21's, a two bank cache with one bank per chip and a
shared FPU all in a little bitty BGA...
> On chip memory (or should I say on memory F21)
> - I listed some alternatives of people with the facilities
> in last
> message. Benifits, very tight interface to memory.
Think I must have missed your last message Wayne, any chance you could
mail me a copy if you have one ?
> One of the things I forgot to put in the last message:
> Optical Ram, over 10 years ago Australian National
> Unitversity made a discovery that could increase the
> capacity of a CD by Several Million times. Latter this
> croped up as the Blue disk (I think it might have been
> called rainbow disk). Basicaly variouse rare earth
> materials could store millions of frequencies of light. I
> believe that the disks exposed to light turned blue.
> Recently the technology has cropped up again in the USA:
> Optical Ram is a flat card working off of simular principle
> as above, using an ultrasonically guide laser it can read
> and write data at upto 120Mb/s. Realisticaly in the future
> (sciencefiction future) this could fit on the same chip. I
> don't have the details but quick seek times are ussured.
With the way microelectromechanical devices are going, especially wee
bits that move predictably and fast, and the range of chip-scale
semiconductor lasers this day may not be to far away (famous last
words aside) although they sure ain't in the standard cell libraries
yet...
> Now realistically (non-sci-fi) this is a serial data
> stream, if random access is fast enough and multiple lasers
> could be synchronised together (say in parralel lines at 1
> per bus bit) memory throughput could be vasely increased.
> 32*120MB/s (with I think at least 30Gb per square
> centimeter), a lot of instructions.
Had a look at their web page and it looks interesting. The sustained
throughput is certainly (potentially) very good, as you'd sort of
expect from most types of holographic memory, but the bug-bear is the
scanning latency - 10 usec per slot is just too huge, although you do
at least get a sort pre-fetch for free with the bit density. They also
look very appealing for neural-net based computing esp. pattern
recognition and so-forth. I think CalTech spawned a commercial company
based on their holographic memory which, IIRC was a page at a time,
alignment system - zap a bit map of the object (face) into the crystal
for all the people, say, who are authorised to enter a building using
the reference beam at a different angle for each; when the person
looks into a image capture system on a door or what-have-you put the
image through the crystal and hey presto, you get a reference beam
back out. You can then use the beam spread to work out if the body
belonging to the face belongs in the building - but there's a million
similar uses for this. I been trying to think of ways to avoid the
switching or scanning, or at least integrate it with the memory
read/write request directly for a while now, but it's really back of a
fag packet sort of stuff and usually end up with more gotchas than
got-its; if anyone has heard of anything interesting or new on this
front i'd love to hear about it.
> Does anybody know of any fast semicoductor memories with
> fast random access though? I suggested multibank ram before
> due to the fact that it yeilds a minium of 66 mhz (or was
> that 166Mhz) of random access in a worst case scenario (but
> now is old technology and is only used on some graphic
> cards). What about sram, still used in low powered (f21
> equipement)?
On the architecture side the Direct RAMBus stuff looks quite promising
<http://www.rambus.com>, with around 1.6G-Bytes/sec per channel into
low cost memory modules on a standard DIMM package. The infrastructure
is there to have multiple channels so with the right controller you
could multiply that figure by the number of channels which could be
rather nice for a multi-processor box (or a uni-processor with really
parallelised read/writes). It's probably bound to have a struggle as a
competing technology even with support from the likes of Compaq, Dell
and IBM on the system side and LG Semicon, NEC, Fujitsu, Siemens etc
etc on the component side, but who knows..
As far as semiconductor technology goes the last general tit-bit I saw
was an article in New Scientist (4th July Vol.159 No.2141 ) dealing
with memories based on a ferrite mineral, (perovskovite or something)
which has been known about for 30-40 years but got overlooked in the
rush to silicon. The write-up hints that it could be significantly
faster than Si-based chips and it's starting to get some decent
research funding, but for the time being it's still Star Trek time
alas.
> Have fun, Wayne.
Sure hope so. Oh yeah, really liked the Micro-Display stuff. Haven't
got the cost/availability yet, but seriously interested in making up a
near focus display a la Private Eye using these displays - thanks for
the pointer.
> ______________________________________________________ Get
> Your Private, Free Email at http://www.hotmail.com
Just me tuppence worth,
Cheers Tim
--
------------------------------------------------------------------------
Tim Parker, Leading Software Engineer
Redac Systems Ltd., Tel : +44 (1454) 207800
Zuken-Redac R&D Centre, Fax : +44 (1454) 207803
1500 Aztec West, Almondsbury, Email : timp@redac.co.uk
Bristol, England, BS32 4RF. Web : http://www.redac.co.uk
"Any opinions are mine, and mine alone. They do not reflect
reality let alone this company"
------------------------------------------------------------------------
"In a word -- im-possible!"
"That's two words," said Dibbler.
-- (Terry Pratchett, Moving Pictures)
