NetNews Usenet Archive 1992 #27/NN_1992_27.iso/spool/comp/sys/dec/6052

Internet Message Format | 1992-11-18 | 3.7 KB

Xref: sparky comp.sys.dec:6052 comp.os.vms:18187
Path: sparky!uunet!sun-barr!cs.utexas.edu!zaphod.mps.ohio-state.edu!darwin.sura.net!paladin.american.edu!news.univie.ac.at!hp4at!mcsun!uknet!mucs!lilleyc
From: [e-mail] (Chris Lilley)
Newsgroups: comp.sys.dec,comp.os.vms
Subject: Re: alpha memory requirements (vs vax) ?
Message-ID: <[e-mail]>
Date: 18 Nov 92 01:39:28 GMT
References: <[e-mail]> <[e-mail]> <[e-mail]>
Sender: [e-mail]
Followup-To: comp.sys.dec
Organization: Dept Computer Science, University of Manchester, U.K.
Lines: 67

>>> In article <[e-mail]>
[e-mail] () writes:

>>>>The memory for the new systems is priced at $125/MB

>>So in article <[e-mail]>, [e-mail] (Jurgen
Botz) writes:

>>> Nonetheless it's too expensive... 70ns 1Mb chips are selling for less
>>> than $2.50 per in volume... that works out to less than $25 for a
>>> 9-chip SIMM.  If your manufacturing cost is as high as $10/SIMM,
>>> that's still nearly a 300% markup.

But in article <[e-mail]> [e-mail] (Bruce C.
Wright) points out:

>>A minor nit, but ECC memory won't be 9 bits wide.  9 bits only gets
>>you parity, you need more than that for ECC (depending on the method
>>used).  I don't know offhand the memory architecture used by the
>>Alpha so I don't know exactly how many chips it would need per SIMM.

>>I suspect that the DEC memory is still on the expensive side though,
>>just not as badly so as your computation would suggest.

Which is followed up in article <[e-mail]>
[e-mail] (M Darrin Chaney):

>There's another factor that our genious didn't take into account, that is,
>the speed of the memory.  70ns memory won't hack it on a machine that is
>running at 150-200MHz.  In case you didn't notice, 200MHz is a 5ns cycle,
>a bit smaller than, oh, say *70*.

This is presumably why all the systems have sizeable secondary caches
in addition to the on chip primary cache. Does anyone know what speed
the main RAM is clocked at? More to the point is this static (I guess
not) so what is the actual access time including latency and refresh?

Also what speed is the secondary cache going at (still slower than 5ns
I bet, unless the SRAM comes with a metal top to fry eggs on)

>I know one 3rd party vendor, whose prices are significantly lower than DEC's
>on most products, just quoted me $6495 for 64MB of memory for the Sandpiper.
>DEC isn't too much higher than that, especially when you consider the price
>disparity in the DECstation-line memory cards.

If you feel that is not significantly higher, you could buy from your
third party and donate the insignificant US$ 1505 to a worthy cause
such as myself ;-)

But yes, DEC has to get this one right, and it is in their own long
term interests to have all the Alpha AXP systems stuffed full of
memory. I suspect they are not trying to make a fast buck out of
memory. That would be stupid...

Actually, the main problem I have is not with memory costing US$125 a
meg in the states, but that it is likely to cost {compose L minus -
damn this emacs editor} UK pounds 125 a meg over here. Similarly for the
prices of the main systems. Does anyone have UK pricing yet?

--
Chris Lilley
---------------------------------------------------------------------------
Technical Author, ITTI Computer Graphics and Visualisation Training Project
Computer Graphics Unit, Manchester Computing Centre, Manchester, UK
Internet:   [e-mail]        Janet:   [e-mail]
Voice:      [phone]          Fax:     [phone]
---------------------------------------------------------------------------