Re: Revival of Alpha?

From: Chip Coldwell (coldwell_at_gmail.nospam)
Date: 06/22/05 

Date: Wed, 22 Jun 2005 17:41:50 -0400

On Wed, 22 Jun 2005, David Turner, Island Computers US Corp wrote:

> If HP were to go x86 - would HP put hardware restrictions on the systems
> to force VMS users to purchase specific models? If HP didn't, I am sure
> they would have a huge "black market" of VMS installed on DELL PCs etc.
> They still do it with certain NICS, Gigabit cards, scsi controllers etc.
> I don't think it would be in their interests to develop the X86 platform
> if not.

Most of the existing hardware restrictions for OpenVMS/Alpha are a
reflection of the availability of device drivers or sometimes the
arcana of PC hardware that expects to interact with a BIOS instead of
the SRM console. If you're willing to write your own drivers and
forgo vendor support (or perhaps get vendor support from the people
that wrote your driver) then you can use quite a bit more hardware
than what HP certifies. I realize that neither of these options makes
sense for people in a production datacenter environment, but I'm
speaking here only of technology limitations.

I think the SRM console has to be one of the biggest technical hurdles
to an x86 port of VMS. x86 platforms generally use a BIOS. In fact,
many of the Alpha platforms that don't support VMS are disqualified
because they use AlphaBIOS (designed for Windows NT/Alpha) instead of
the SRM console. I don't know what Itanium uses, but if it's similar
to the x86 BIOS then I would imagine that this constituted a big part
of the Itanium port of VMS.

Even bigger than the console is a MACRO32 compiler that targets x86.
I suppose if they can write one for Alpha and another one for Itanium,
then they can write one for x86. But it's a big effort nonetheless.

Many people have erroneously cited the lack of four CPU privilege
levels on x86 as a reason why a VMS port is impossible. In fact, the
x86 architecture has supported four CPU privilege levels for some
time; at least since the Pentium.

All of this effort pales in comparison to what would be required to
restart Alpha production, though. The cost of developing CPUs these
days is just enormous. It might cost you only ~$200K to hire a good
engineer, but he will need upwards of $1 million per year in software
licensing for the tools he uses. Some firms even work their engineers
in shifts in order to maximize the value they get from their software
tools. That's how HP and Intel managed to get the cost of developing
the Itanium up so high.

Of course, the Alpha team was unique in this respect: they developed
their own tools in house. (A friend of mine who did a summer
intership with the Alpha team at DEC in 1997, told me that the tools
ran on VMS -- not surprising, really.) I have no idea what the fate
of those tools is today, but I think it's safe to assume that they are
no longer being maintained. Software doesn't wear out, but the
expertise to use, maintain and extend it does. Since the Alpha team
has been scattered to the four winds now, reassembling a quorum is
probably impossible. Most likely, before resuming work on the Alpha,
the existing designs would have to be ported over to use the standard
tools that fresh-out-of-college engineers know how to use.

Apple's recent decision to switch to x86 is a reflection of the harsh
reality that you can't justify the development and fabrication costs
of a CPU unless you have huge volume. IBM certainly knows how to make
a low-power high-performance PowerPC (witness the Cell processor in
the PlayStation 3), but Apple doesn't generate enough volume for IBM
to justify putting resources into the G5 speedups that Steve Jobs
needed. And switching to IBM from Motorola (or "Freescale" as the
spun-off semiconductor division calls itself) was already an act of
desperation because Motorola couldn't provide the speedups. Jobs
really had no choice but to switch horses.

I really, really wish they hadn't pulled the plug on Alpha. It was
the best RISC architecture on the market. And x86 is a really, really
ugly ISA in comparison. But as the Preacher said, "The race is not to
the swift, nor the battle to the strong, neither yet bread to the
wise, nor yet riches to men of understanding, nor yet favour to men of
skill; but time and chance happeneth to them all."

Chip 

-- 
Charles M. "Chip" Coldwell
Turn on, log in, tune out

Relevant Pages

  • Re: New itaniums out at 2.5x perform gain
    ... stuff installed and running on VMS. ... somehow easier to port it to x86-64. ... Given that active investigation of porting VMS to x86 was actually under way in the early '90s, suggesting that there's any insurmountable obstacle involved seems a bit unreasonable - especially now that x86 has become 64-bit. ... Alpha sales at this point in time is a very wrong mistake). ...
    (comp.os.vms)
  • Re: New and Refreshed Advertising Colateral for OpenVMS
    ... For many the cost may not be such a big thing. ... But enough about the Alpha ... Time will tell what Intel will do with the itanic. ... >> control of whether CPUs that run VMS would be available. ...
    (comp.os.vms)
  • Re: AMD blew it big time!
    ... The only two OS's that run on the Alpha are VMS and Tru64 Unix. ... x86 boat anchor performance wall ... ...
    (comp.os.vms)
  • Re: Revival of Alpha?
    ... If HP were to go x86 - would HP put hardware restrictions on the systems to ... I am sure they would have a huge "black market" of VMS ... >> What would it take for HP/Compaq and Intel to rekindle the Alpha ... > I have only seen one Itanium running VMS. ...
    (comp.os.vms)
  • RE: Cost of used alphas vs 8086+charron
    ... Cost of used alphas vs 8086+charron ... If one were to pitch VMS to a new customer, ... charron-alpha, to match a ds15/25 machine's perfomance? ... i.e. will industry standard servers with the charron alpha ...
    (comp.os.vms)