Re: Different Implementations of VLIW .
- From: rpw3@xxxxxxxx (Rob Warnock)
- Date: Tue, 20 May 2008 05:15:03 -0500
Erik Trulsson <ertr1013@xxxxxxxxxxxxx> wrote:
+---------------
| Eric P. <eric_pattison@xxxxxxxxxxxxxxxxxx> wrote:
| > - Main memory was expensive. IIRC dram was around $25,000/MB
| > circa 1985. ...
|
| Main memory was certainly more expensive back in 1985 than it
| is today, but I don't think it was quite *that* expensive.
+---------------
No *way*!! [See below...]
+---------------
| Some datapoints:
| In 1984 the IBM AT was introduced. It shipped with either 256KB or 512KB
| RAM depending on configuration.
| In 1984 the original Mac was introduced by Apple. It shipped with 128KB
| main meory.
| In 1985 Commodore introduced the Amiga 1000, which came with 256KB RAM as
| standard.
|
| None of those computers were exactly cheap at introduction, but nor
| did they cost anywhere close to $25,000 (not even after adding memory
| upgrades to a full MB of RAM.)
+---------------
Those systems were all both late & small (but cheap). The Fortune Systems
"32:16" was demo'd at Comdex in Fall 1981, was pre-sold to distribution
over that Winter & Spring 1982, and was (finally!) shipped to paying
customers in August 1982 with 1 MiB of RAM -- and a 5.5 Mhz Mc68000
CPU and a 5 MB hard disk, running Unix with bits & pieces from v.7,
BSD 4.1a, and SVr3 -- for a system *total* unit sales price of $5000.
+---------------
| Based on vague memories and (more importantly) a bit of research with
| Google, I would say that memory prices in 1985 was around $1,000/MB.
+---------------
Hmmm... We were using 64 Kibit DRAMs in the FS 32:16. That would make
the DRAM chips be just under $8.00 each ($7.8125). Yeah, that's in
the ballpark. My memory's not so great on those details after all
this time, but ISTR that we had cut some deal to get them for ~$6/chip,
which would make it more like $770/MiB [our cost], which is close enough
to your $1000/MiB.
Note that at the $5000/system price we were getting almost *no* margin
on the DRAMs, since the original plan had been for the base system to
come out with only 128 KiB(!!), except that software bloat and schedule
pressure [and poor swapping performance with the first version of the
hard disk controller!] forced us to jack the RAM up to 1 MiB for the
early shipments. [IIRC, we were later able to squeeze things down a bit
and trim the base system -- still priced at $5000 -- down to 512 KiB,
and then charge another $1000 for the "upgrade" to 1 MiB. Or something
like that.]
Anyway, I don't know whether you're talking "OEM cost" or "retail
sales price". If you ignore our artificially low initial DRAM sales
price [needed to get the early systems to run reasonably!], at a
$770/MiB OEM cost an normal 3-4:1 markup for anyone else would have
made the customer retail price be $2300-3000/MiB, which is also in
your general ballpark.
-Rob
p.s. FWIW, as a data point from a decade later, in Summer of 1992 I
bought a PC clone at one of those *cheap* outlets that had sprung up
around the SF Bay by then, and I paid $3000 for 32 MiB of add-on DRAM
for it. [Yeah, really. And $2400 for a 1.5 GB SCSI disk!]
-----
Rob Warnock <rpw3@xxxxxxxx>
627 26th Avenue <URL:http://rpw3.org/>
San Mateo, CA 94403 (650)572-2607
.
- Follow-Ups:
- Re: Different Implementations of VLIW .
- From: Eric P.
- Re: Different Implementations of VLIW .
- References:
- Re: Different Implementations of VLIW .
- From: Eugene Miya
- Re: Different Implementations of VLIW .
- From: Eric P.
- Re: Different Implementations of VLIW .
- Prev by Date: Re: Register-less CPU
- Next by Date: Re: Register-less CPU
- Previous by thread: Re: Different Implementations of VLIW .
- Next by thread: Re: Different Implementations of VLIW .
- Index(es):
Relevant Pages
|
Loading