[rescue] Let's ask Fhourstones how much faster - Re: The good old days really were

[Toby Thain]

On 10/06/13 12:23 PM, Lionel Peterson wrote:
> The good CP/M machines ran 4 MHz CPUs in the late '70s, and current multi-core
> CPUs knocking on 4 GHz, clocks are easily 1,000 faster. Of course, factor in
> multi-core and multi-byte word length, and compensate for the vast increase in
> work done per clock cycle and I think a claim of 'million times faster' is
> about right.

It's unlikely a multiplier anywhere close to that could be supportable 
when memory latencies are considered.

Let's check out a decent integer benchmark like Fhourstones for a more 
realistic ratio.

The highest Fhourstones rating on this page 
http://homepages.cwi.nl/~tromp/c4/fhour.html is a

   3.66GHz Xeon at 12,032 Kpos/sec.

An older listing at 
http://performance.netlib.org/performance/html/fhourstone.data.col0.html 
records:

   33.3MHz '486 at 26.7 Kpos/sec

On a very hand-wavey basis the clock speed ratio is something like 110x, 
and the benchmark is about 450x. I don't think you'll find a ratio near 
1,000,000:1 for any benchmark that accesses memory.

--T


>
> Of course, most mainstream 'PC' CPUs prior to the debut of the IBM PC were
> slower than 4 MHz...
>
> Lionel
>
> On Jun 10, 2013, at 7:28 AM, microcode at zoho.com wrote:
>
>> And then they have a way to
>> understand what it must have been like writing code with almost no memory
>> and and CPUs thousands (millions?) times slower.
> _______________________________________________
> rescue list - http://www.sunhelp.org/mailman/listinfo/rescue