Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > it.comp.retrocomputing > #53042 > unrolled thread

Rise and Fall of RISC :D

Back to article view | Back to it.comp.retrocomputing

Contents

  Rise and Fall of RISC :D "zfkirone@gmail.com" <zfkirone@gmail.com> - 2020-11-04 00:14 +0100
    Re: Rise and Fall of RISC :D Ammammata <ammammata@tiscalinet.it> - 2020-11-06 07:31 +0000

#53042 — Rise and Fall of RISC :D

Tra i commenti di un articolo, ho trovato questa testimonianza:

«Yeah, I think he's misremembering the 90s. I worked at HP at the time.
Not only did the killer micros do a "good enough" job, they knocked the
socks off anything we could produce ourselves. The creamed us.

Here's a little history. HP had 68k workstations (the HP 9000/300 line)
and we had PA-RISC timeshares (HP 9000/800 line). Sun, SGI, and IBM came
out with RISC based Unix workstations and we bought Apollo. The Apollo
guys helped us bang out the Snakes systems, HP 900/700 systems with
crazy fast PA-RISC CPUs, running an a mind blowing 50 and 66 MHz. Wow,
don't use it all in one spot. For comparison, a fast x86 at the time was
around 10-12 MHz. Clearly RISC was the faster architecture.

Intel didn't sit still. They kept cranking the clock rates of x86,
adopting a RISC core below a hardware (!) x86 instruction translator.
Wow. Just wow. The RISC vendors did the same and managed to keep a
shrinking performance lead.

HP launched a moon shot, the Tahoe program which turned into Itanium.
This was a superscalar processor and initial projections were it was
going to ship around 1997 at a blazing 500 MHz! Other processors at the
time of the announcement were in the 100-150 MHz range so that was a
pretty huge performance jump.

Then Intel and AMD got into a clock rate war. They poured enormous
amounts of cash into building new fabs and improving their designs.
Turns out a few billion dollars beats a better architecture. We all know
how this ended. The killer micros with enormous investment got better
much faster than RISC systems did. By the early 2000's it was clear x86
beat every RISC processor both in raw performance and price performance.

And that's when all the RISC architectures petered out. It's not that
people had anything against them, it's not that the market was too
small, the market entirely went away. There was nothing you could do
with a Sun workstation you couldn't do better and cheaper with a stock
PC running Linux or Windows. Literally nothing. To use examples from
TFA, want a JTAG interface? Plug it into a PCI slot or USB port. Want a
FPGA? PCI slot. Visualization? Latest graphics card from Nvidia or ATI.

TFA is right, we don't buy a system and go to the App store to see what
it can do. We still start with an app and pick a system to run it. It's
just that today, the market has consolidated to basically Windows,
Linux, OSX, and The Cloud. There's a lot of stuff available on all three
desktops (e.g. all the cloud apps). Some stuff is still specific to one
OS or the other. People are savvy enough to pick the one they want and
that will blow the socks off what the author nostalgically remembers
from 2000.»

fonte:

https://it.slashdot.org/story/20/11/03/1512208/the-untimely-demise-of-workstations#comments

CYA

[toc] | [next] | [standalone]

#53051

Il giorno Wed 04 Nov 2020 12:14:02a, *zfkirone@gmail.com* ha inviato su 
it.comp.retrocomputing il messaggio news:rnso7q$1usg$1@gioia.aioe.org. 
Vediamo cosa ha scritto:

> Clearly RISC was the faster architecture.
> 

ricordo la recensione dell'Archimedes su MC: toni entusiastici per le 
prestazioni

-- 
/-\ /\/\ /\/\ /-\ /\/\ /\/\ /-\ T /-\
-=- -=- -=- -=- -=- -=- -=- -=- - -=-
........... [ al lavoro ] ...........

[toc] | [prev] | [standalone]

Back to top | Article view | it.comp.retrocomputing

csiph-web