Path: ...!eternal-september.org!feeder3.eternal-september.org!news.eternal-september.org!.POSTED!not-for-mail
From: BGB <cr88192@gmail.com>
Newsgroups: comp.arch
Subject: Re: Historical evolution of CPU perf
Date: Fri, 11 Oct 2024 03:03:14 -0500
Organization: A noiseless patient Spider
Lines: 101
Message-ID: <veam42$3i733$1@dont-email.me>
References: <jwvh69l6zms.fsf-monnier+comp.arch@gnu.org>
 <ve6hmr$2nbq0$1@dont-email.me>
 <20241009201840.31c1e818f9d5e789691c718b@127.0.0.1>
 <1fd00af7b706a22d8e588cef3be10a6b@www.novabbs.org>
 <ve9rdc$3bqg4$1@dont-email.me>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Injection-Date: Fri, 11 Oct 2024 10:03:15 +0200 (CEST)
Injection-Info: dont-email.me; posting-host="fc39995b3786a0a07ce43dd90b274fee";
	logging-data="3742819"; mail-complaints-to="abuse@eternal-september.org";	posting-account="U2FsdGVkX1/TVHzfX9hnRVvCECNoFBukNztj0eVfMp4="
User-Agent: Mozilla Thunderbird
Cancel-Lock: sha1:SPKWBSHwGhO1On+Evs3T0VFC+eY=
Content-Language: en-US
In-Reply-To: <ve9rdc$3bqg4$1@dont-email.me>
Bytes: 5769

On 10/10/2024 7:27 PM, Sarr Blumson wrote:
> MitchAlsup1 <mitchalsup@aol.com> wrote:
>> On Wed, 9 Oct 2024 19:18:40 +0000, Kerr-Mudd, John wrote:
>>   
>> I remember the PDP-11/20 in the computer lab at NCR.
>> Last person out at night would flick the power switch OFF, and
>> the computer was OFF in 1/60 of a second.
>> First person in would flick the switch ON and the computer was
>> back where it was turned off in 1/60 of a second.
>>
>> We used the 11/20 as a remote debug device for the 8085 cash
>> register machine(s) we were building.
> 
> Core memory: slow to access but also slow to forget.

Core memory was before my time, but I remember reading somewhere that it 
needed to be preheated to a certain operating temperature in order to 
write to it (because the hysteresis energy of the ferrite rings was 
temperature dependent, and it needed too much power to flip the bits at 
lower temperatures).



I am left to wonder if one could have made a non-volatile 
electrochemical memory with "ye olde" tech;
   Say, for each memory cell:
     Two lead plates and a thin separator
       With a water + sulfuric-acid electrolyte;
       3 states:
         A: Plate A is lead dioxide, Plate B is lead
           Electrolyte is live.
         B: Like A, but with the plates reversed;
         C: Plates are lead sulfate, preferably avoided;
           Too strong of sulfate, memory cell dies.
     Or, two iron plates, and a separator:
       Similar, but using a potassium hydroxide based electrolyte (1).


*1: In the C state, the plates would try to convert to iron-potassium 
ferrate, but this process would cause the electrolyte to become slightly 
acidic (depleting oxygen and leaving an excess of H+ ions), which would 
decompose the ferrate. In the A and B states, one plate will be metallic 
iron and the other iron oxide, with a basic electrolyte.

Each memory cell would be constructed more like a capacitor, but would 
be "charged" more like a battery, but with (comparably) very little 
capacity. Bits would be stored in the charge-polarity of each cell.

Likewise, for RAM like operation, would likely use a fairly dilute 
hydroxide solution (likely much weaker than if it were being used for 
energy storage).


Memory cells could be separated either by distance or by an insulating 
barrier, though potentially the electrolyte reservoir could be shared 
between all of the cells (assuming primarily 2 state operation).

For the iron-hydroxide case, the total amount of free oxygen in the 
system would likely need to be controlled, so it could not be open to 
the atmosphere. If more oxygen could leak in, likely the plates would 
turn into ferrate and the electrolyte would turn into water. With any 
further oxygen turning the remaining iron into iron-oxide.

Though, charging the cells may drive out the extra oxygen (but, this is 
likely to require considerably more power, and if the process goes on 
too long, it may structurally damage the cells). Similarly, if the 
electrolyte becomes over-saturated with oxygen, then the oxidation 
process would resume as soon as power is removed (unless the cells are 
charged for long enough for any excess free oxygen to diffuse back out 
of the electrolyte).

For a lead-acid chemistry, free oxygen would not matter.



Here iron-iron-hydroxide might be preferable for memory cells, as it 
would likely be more stable and have a longer cycle life than a 
lead-based design. Cells would likely have a very low energy density, 
but this doesn't matter for memory cells (and would be preferable for 
RAM like use). One would need a chemistry where cells can be charged in 
either direction and which can tolerate cells being rapidly driven from 
one polarity to another. Here, less capacity also means less energy 
needed to flip a bit (but likely also less stability).

One would also need the wiring to not interact with the electrolyte.
Should probably be able to wire up the cells with a crossbar configuration.

Similarly, the separator would need to be non-reactive with the 
electrolyte. More modern materials, like porous plastic or fiberglass 
would work. Within the limits of older tech, dunno. cellulose was a 
common (such as paper or cotton) but would likely slowly dissolve in the 
electrolyte solution (I guess, old time solution would probably be to 
use asbestos or similar, or maybe a porous ceramic).

For similar reasons, could not use bone or leather/vellum (also weak 
against both acids and hydroxides), ...


....