Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 27 Apr 2001 13:41:43 +0200 (CEST)
From:      Remy Nonnenmacher <remy@boostworks.com>
To:        ajh3@chmod.ath.cx
Cc:        keichii@iteration.net, jgowdy@home.com, freebsd-hackers@FreeBSD.ORG
Subject:   Re: x86-64 Hammer and IA64 Itainium
Message-ID:  <200104271139.f3RBdNC63674@luxren2.boostworks.com>
In-Reply-To: <20010426192352.A2341@cec.wustl.edu>

next in thread | previous in thread | raw e-mail | index | archive | help
On 26 Apr, Andrew Hesford wrote:
> On Thu, Apr 26, 2001 at 06:08:36PM -0500, Michael C . Wu wrote:
>> | and a bunch of ARMs for low-level I/O tasks. Back to imagination. (Take
>> | a look at 0.15um copper process FPGAs with embeded ARM at Altera, for
>> | example, and you will see why no one, in the futur, will never ever need
>> | a proprietary and undocumented 'server class' SCSI or network card).
> 
> Do you mean to say that, in the future, we will all blow our own FPGAs
> at home? That to get the next processor upgrade, you download the
> schematics and use Xilinx Foundation Series to build a chip? FPGAs are
> not the answer... they can never be as fast as custom-fab chips. It may
> be, in the future, we get circuits to run fast enough that we wouldn't
> notice given today's speeds, but remember, if FPGAs get that fast,
> custom-fab chips will be even faster. 
>

In fact, the move is to hardwire some core functions (CPU, PCI,
RAM/SDRAM controllers). Here you get a 'fab' speed. Then you have a
programmable area to specialize the whole stuff. All this in one chip.
Fastest chips are needed by very few peoples for very specific needs.
I'm talking of daily needs: Audio, LAN, WAN, SCSI.. Areas where existing
FPGA's speed can match for 50-100$.

> The future isn't FPGAs; it's GaAs BJT circuitry, designed and built by
> the guys who have money to set up a fab and roll out millions of chips.
> It will be a long, long time before anything changes about how we get
> our chips; the only thing that will change is how they are made.
> 

Futur can be GaAs, copper or whatever you want. The fact remains that
these fab guys (and anyway _who_ setup a fab, by now ?) will build a
chip for their own needs and we (if we can access the product) will
have to deal with stupid behaviors and write more code to adapt
things... just to see,  sooner or later, the chip disapear from world's
surface for any good reason they will have.

This is more a philosophical issue than a practical, technical, one:
Chip makers do not care at all of code writers needs. FPGAs give us a
chance to escape from dependancy in a lot of domains (not CPU, anyway)
and push a little further in their influence sphere. From my experience,
the only two words that make your fab guys become very attentive are:
"no thanks".

RN.
IeM



To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200104271139.f3RBdNC63674>