Path: ...!eternal-september.org!feeder3.eternal-september.org!news.eternal-september.org!eternal-september.org!.POSTED!not-for-mail
From: The Natural Philosopher <tnp@invalid.invalid>
Newsgroups: comp.os.linux.advocacy,comp.os.linux.misc
Subject: Re: Rewriting SSA. Is This A Chance For GNU/Linux?
Date: Sat, 5 Apr 2025 12:13:06 +0100
Organization: A little, after lunch
Lines: 139
Message-ID: <vsr383$2421k$1@dont-email.me>
References: <pan$54963$b3f3d4e6$ae35ff46$71fe05c9@linux.rocks>
 <gXCdnTD2YLRBaHX6nZ2dnZfqn_idnZ2d@giganews.com>
 <m4tf1dFmvh3U1@mid.individual.net> <vsd0ui$365s0$1@dont-email.me>
 <vsds7u$2u8h$1@dont-email.me> <wwviknpb1iw.fsf@LkoBDZeT.terraraq.uk>
 <vsksb5$3df6l$1@dont-email.me>
 <ZI2dnQjwJajG9XP6nZ2dnZfqn_WdnZ2d@giganews.com>
 <m581c7Fd22eU2@mid.individual.net>
 <DJOdnXslWrdAbHP6nZ2dnZfqnPudnZ2d@giganews.com>
 <m58mnpFguqjU2@mid.individual.net> <vso5qc$31clb$1@dont-email.me>
 <E2WdnXiNaZ9CTXL6nZ2dnZfqn_idnZ2d@giganews.com>
 <pan$f2307$df5236a$923c4908$a6fb4a1f@linux.rocks>
 <6BidndvG26Vec236nZ2dnZfqnPadnZ2d@giganews.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Injection-Date: Sat, 05 Apr 2025 13:13:10 +0200 (CEST)
Injection-Info: dont-email.me; posting-host="dba3d81f6f0eee3a847ad5e631b58534";
	logging-data="2230324"; mail-complaints-to="abuse@eternal-september.org";	posting-account="U2FsdGVkX19bpnChZBZqgIfOEk/X4AkCVO5kD9+puPs="
User-Agent: Mozilla Thunderbird
Cancel-Lock: sha1:qIcQmgXIPSGMlBBJaqItW11i9g4=
Content-Language: en-GB
In-Reply-To: <6BidndvG26Vec236nZ2dnZfqnPadnZ2d@giganews.com>
Bytes: 6792

On 05/04/2025 09:50, c186282 wrote:
> On 4/4/25 3:15 PM, Farley Flud wrote:
>> On Fri, 4 Apr 2025 08:30:23 -0400, c186282 wrote:
>>
>>>>>
>>>>> I'm not sure there are any little old ladies left to knit magnetic 
>>>>> core.
>>>>
>>>> Last time I looked they were little asian ladies wit teeny nimble 
>>>> fingers.
>>>> Did all the coil winding in that factory.
>>>
>>>     Look up "rope memory"  :-)
>>>
>>>     Hey, it flew us to the moon ...
>>>
>>
>> Who would ever give a flying fuck about this "Neolithic" technical
>> crap?  It's the future that is of concern.
> 
>    No future without a past.
> 
>    And past tricks/thinking/strategies CAN inspire
>    the new.
> 
Indeed.
Many ideas that were infeasible, become feasible with new technology.
Many dont. Windmills being a prime example...

> 
>> My question has always been: when are these memory engineers (or
>> whatever they are called) going to produce cheap RAM memory that
>> can actually keep pace with the CPU?
> 
>    Never ...

The problem is distance between elements times the  size of elements 
divided by the speed of light.

It means that you need to start going 3D on memory to keep the 
speed/capacity withiong bounds.

It has its parallel in human political structures. Huge monolithic 
empires like the USSR simply fail to keep up, because the information 
from the bottom takes a long time to get to the top.

A far better solution is the old British Empire, with governors having 
power over a local nation, and very few decisions being centralised.

> 
>    OK, Moore's Law is getting close to stalling-out CPU
>    performance. Ergo, give it a few years, the memory
>    MAY finally catch up.
> 
Same laws are governing both.  What is happening is more local on chip 
cache.
IIRC the RP2040 has 256K of memory *on the chip itself*.

That's local. Hence as fast as the chip is.


>> For decades we have had to use various levels of high speed, though
>> minuscule, cache memory in order for our software to run, and from
>> a programming point of view cache management is a supreme bitch.
>>
>> The world needs cheap RAM that can operate at CPU speeds.  Then,
>> all programming would be a supreme breeze.
> 
It already has it, juts not in the sizes you want. Because of the 
propagation delay inherently in large arrays.

We can clock the CPUS up to 4Ghz ± mainly because we can make em down to 
10nm element size.

Below that you start to get into quantum effects and low yields.

DDR5 RAM is pushing 3GHz speeds

>    Well, your 'future' isn't providing. MAYbe some odd
>    idea from the past, just on better hardware ???
> 
The better idea is to look at what the actual problems are, and design 
massively parallel solutions to them that do not require a single 
processor running blindingly fast.

>> Cache memory is just another crutch, and its existence is indisputable
>> testimony that modern PC hardware is crippled shit.
> 
>    Well, I'd argue that on-chip cache is always gonna
>    outperform - if for no other reason than the short
>    circuit paths. These days, the speed of electricity
>    over wires is becoming increasingly annoying - it's
>    why they want photonics instead. Of course soon even
>    that will be too slow soon enough ... and you can
>    complain to Einstein .......
> 
Cant yet beat speed of light. Photonics is not much faster than electronics.
Back in the day we measured delay on a reel of 50 ohm coax. It was about 
..95 the speed of light IIRC.

So that isn't the way to go,

Look, you need to study the history of engineering.

Let's take a steam engine. Early engines crude, inefficient and very 
heavy and large. Maybe 1% efficient.
Roll forward to the first locomotives and still heavy but now getting 5% 
efficiency..

Fiddle with that for a hundred years and the final efficiency of a steam 
piston engine approached the theoretical limits of the technology 
without cooling or superheated  steam of around 20%. Now use superheated 
steam in a steam turbine with a condenser strapped on the back - 
suitable for ships or power stations, - and you are getting up to 37%.

But that is fundamentally it.  There is a law governing it
Efficiency is (Steam temperature IN - Steam temperature OUT)/(STEAM 
TEMPERATURE IN) (degrees absolute)

so for 200°C in and say 100°C out TIN = 673°A, TOUT =373°A  gives a max 
thermal efficiency of 45%

You simply will never do better than that with water as the working 
fluid unless you go to horrendous inlet temps of superheated steam


The point is that every technology has a limit beyond which no amount of 
tinkering is going to get you. Engineers come to understand this, the 
lay public do not.  They are always whining 'why cant you power the 
universe form one single bonfire'

Digital computing has a little ways to go, but it is already close to 
the limits

For some problems, precision analogue might be faster...

-- 
Climate Change: Socialism wearing a lab coat.