NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/sys/amiga/hardware-part1/1306

Text File | 1996-08-05 | 3.7 KB | 76 lines

Newsgroups: comp.sys.amiga.hardware,comp.sys.amiga.misc
Path: magnus.acs.ohio-state.edu!csn!ub!dsinc!scala!news
From: [e-mail] (Dave Haynie)
Subject: Re: EDO RAM on March Amiga?
Sender: [e-mail] (Usenet administrator)
Message-ID: <[e-mail]>
Date: Mon, 15 Jan 1996 22:05:42 GMT
Reply-To: [e-mail] (Dave Haynie)
References: <[e-mail]> <4d33po$[e-mail]>
Nntp-Posting-Host: gator
Organization: Scala Computer Television, US Research Center

In <4d33po$[e-mail]>, [e-mail] (Martin Caspersson) writes:
>[e-mail] (Wells Fargo Bank) wrote:

>>This would be great as the upwards of 10% improvement in performace with
>>the RAM, and at only about $10 per SIMM more than standard DRAM SIMMS,
>>it would be a great way to boost the performace and a low cost.

>>They are going to be adding RAM slots, thats a definite.  I can't imagine
>>that supporting the EDO RAM would cost much if any more than for the
>>standard RAM.

>I don't think EDO RAM will do anything for such a slow processor as
>the 030.

Sure it would. The '030 may not process as fast as an '040 or '060,
but that's an architectural difference. It can still hit memory as
fast as an '040 at the same clock speed. And, due to its internal
cache size, the '030 is a whole lot more dependent on memory
speed for good performance than more recent chips.

If there had been EDO RAM around back then, I would have used it in
the A2630 (and, by extension, the A3000). In fact, early versions of
the A2630 used a bursting scheme which could deliver single clock
burst cycles at up to 33MHz. Unfortunately, though, you couldn't get
standard page-mode memory with output buffers fast enough for my
design (basically a 128-bit bus, 32-bits enabled at a time) to work
reliably. Some worked great, but others would self destruct from too
much buffer contention. Page-mode itself is too slow for single-cycle
bursts from an '030 at these speeds (the A3000, of course, does get
two cycle burst). EDO, which really just tweaks the behavior of the
output buffers a bit, would have been a perfect solution. Today's no
different.

>I read in an article in BYTE that EDO RAMs were developed
>because the older types of RAM just couldn't keep up with the new
>processors like the Pentium 166 and the forthcomming Pentium 200, as
>well as the new Pentium PROs.

Older types of RAM couldn't keep up with 25MHz 680x0 systems, but you
could fake it acceptably by using with page-mode burst, interleaving,
etc. The older 80x86 chips didn't hit memory quite as hard. EDO is
hardly a miracle, it's just a very small tweak that can often drop a
wait state on a fast system. Because it's a small tweak, it was
trivial for DRAM vendors and board vendors alike to support --
standard page-mode memory will work just dandy in a system that
supports EDO, just a little slower.

The high speed CPUs in workstations, like 300+MHz Alphas, certainly
push the envelope on DRAM, but they're not in enough volume to make
any radical changes overnight. The PC helps evolve large scale
changes, but most PClone CPUs run their local bus at a fraction of the
internal clock speed; rarely above 66MHz, often at 33MHz. That's
changing some with faster CPUs, but mainly just because they can hit
synchronous cache SRAM up to around 100MHz (though not necessarily
every synchronous cache system springs the big bucks for 100MHz
SRAM).

Dave Haynie          | ex-Commodore Engineering |   for DiskSalv 3 &
Sr. Systems Engineer |  Hardwired Media Company | "The Deathbed Vigil"
Scala Inc., US R&D   |    Ki No Kawa Aikido     |     [e-mail]

         "Feeling ... Pretty ... Psyched" -R.E.M.