13 Jun
2012
13 Jun
'12
10:58 p.m.
On Tue, Jun 12, 2012 at 07:48:18PM -0700, Chuck Guzis wrote:
On 12 Jun 2012 at 19:55, David Riley wrote:
Well, from what I'm currently reading in "Dealers of Lightning", very
early semiconductor memory was quite a bit less reliable than core,
coupled with miserable yields. Presumably both improved reasonably
quickly. It had the advantage of being smaller, cheaper and faster,
though.
Kind regards,
Alex.
--
"Opportunity is missed by most people because it is dressed in overalls and
looks like work." -- Thomas A. Edison
Of course, "working reliably" does
involve a lot more than just
eliminating bit errors from cosmic rays, I'll give you that one.
I thought it interesting that the CDC 6600/CYBER 70 series did not
have any memory error checking whatsoever (the old "Parity is for
farmers" quote), but that the CYBER 170 series had semiconductor
memory and SECDED.
From my recollection, if the old systems passed
deadstart
diagnostics, the chance of seeing a memory error that day was very
small.
Was early semiconductor memory less reliable than core? I don't
recall that part.