Slashdot Mirror
Compaq sees Linux as selling Alpha chips
phlebas writes "Cnet news has an article about how "Compaq Computer sees Linux as a way to increase sales of its Alpha microprocessor, and the company is trying to encourage other Linux distributors besides Red Hat to support the chip." They ar also plannig to "soon introduce new pricing geared to encourage Linux users to "step up" from Intel to Alpha chips." Maybe after all i'll be able to get one those in my lifetime " I think it's pretty clear that Linux is the way for Compaq's to really push the Alpha-let's hope they actually do it.
I don't think the Alpha can really take off ;).
unless Compaq allows vendors to OEM the chip
(i.e. Dell etc)...then of course you have to get
vendors interested in OEMing the chip
Open Genera costs quite a bit, plus you have
to get Digital Unix.
Now _that_'s a reason not to buy any other
Linux system.
I don't think it would cannibalize their system,
after all, I would not change IRIX for Linux,
for example (but I don't care about Digital
Unix, Solaris, HP/UX).
Don't forget Debian! Slink just recently shipped for Alpha too. Here's the news page,
http://www.debian.org/ports/alpha/news
Would Compaq be better off opening up the src for DEC *NIX in such a way that it's features could be more easily copied into Linux?
Intel got a rocket ride as THE chip for Windows. DEC tried to replicate that by trying to position Alpha as THE chip for NT and they've failed.
Linux offeres a new opportunity. Why not make a play at making Alpha synonymous with Linux? IBM acknowledged that supporting Linux will cut into AIX sales, but the new opportunities more than make up for it.
Compaq might be better off committing it's DEC *NIX engineers into getting DEC *NIX features into Linux.
It's right there in spot #113 - it's just called "Avalon", not Beowulf.
We won't get a 64-bit chip from Intel anytime
soon, because Intel has to wait for Microsoft.
There is just no way Intel will ship Merced
without a native NT port.
Intel is pissed at Microsoft. Apperently Microsoft
has been lazy about the compiler: they add two
NOPs for every useful instruction. (Merced packs
three instructions in a 128-bit chunk)
Microsoft fucked up portability too. Rather than
correct their own erroneous assumption that a
long is 32 bits, they made pointers larger than
longs. This is a first, and it breaks standards.
Avalon made noise, but another Linux machine
quietly broke the top 100. Few people even knew
that the machine at rank 98 ran Linux.
Actually, the 21264 was just released, the 21364 has been underway for a while, and if I understand how things work, the 21264 team has started on the 21464.
For info on the 21364, look at this presentation.
Compaq appears to intend for the Alpha to be a competitor to Merced, and there's no reason to think that it won't be considerably faster. They're also going to migrate their Tandem customers over to Alpha (from MIPS).
It's perceived as trailing others such as HP-UX, Solaris and AIX because it's now owned by a "Wintel" company.
Just look at Apple computer. If you have reasonable margins, you can make a killing off hardware!
Apple's margins are wholly _un_reasonable..
I want to buy equipment from a company with supermarket margins...
huh huh...
dual 21264... huh huh...
Grimlok want one NOW...
There's been many vendors building and selling
Alpha systems based on Digital's OEM mobos for _years_! You just probably never heard of 'em..
http://www.microway.com
http://www.aspsys.com
http://www.pandaproject.com
among others... Shit, Tatung (among others) resells Ultra2i/333+ systems based on Sun's OEM Ultra mobos.. You just need to look a mite harder..
... check out this bad boy...
dunno, the site says they rank based on LINPACK, but IIRC that's a DB benchmark.. Are there any Beowulf-enabled DBs?
Look at entry no. 113 using Alphas (Avalon) and also entry no. 97. Also clustered Alphas running Linux.
Is there a GPL Fortran 95 project??
VAST/f90 is free for noncommercial use,
and generally works very well (IMHO), but it ain't Free Software...
a g90 or g95 command to go with g77 would be cool...
The Avalon cluster currently ranks about 113 on the top 500 at http://www.top500.org Beowulf isn't dead. It's alive and kickin a@#......
hmm... recently, amiga coders have been running into 64-bit stuff when coding for their PPC dual-processor boards.
They tend to call a 32-bit value an "int" or a "long int", a 16-bit value a "short int", and a 64-bit value a "long long int". Ah well...
Amiga C programs are supposed to use the APTR and BPTR wrapper types for OS pointers, so, in theory, it's pointer constructs could move to 64-bit anyway
If you check out the autodocs for PPCAdd64 and PPCMult64 in the ppc.library, you can see this.
It's also pervasive in WarpOS code. (and WarpOS is the standard for PPC classic amigas)
N.B. WarpOS=="Classic" Amiga PPC microkernel that runs on the PPC side of a dual-processor system running AmigaOS 3.1 (and soon 3.5) on the 68k side
WarpOS!=OS/2Warp
Digital Unix is based primarily on the Mach 2.5 kernel and BSD code. Digital uses and pays royalties on some OSF code as well, hence its previous name OSF/1.
Will these new "Cheap" machines run OpenVMS? Or are they going to simply be WinNT machines minus the OS. I'm looking for a couple cheap powerful systems for my home VMS Cluster, which currently has two Alpha's and will soon have some VAXen also.
I've just got a catalog from Tigerdirect.com.
Look at these prices:
Slot 1 mb LX chipset 35.99
sillyceleryA 300Mhz 74.99
64 Mb SDRAM 79.99
For home use, no Alpha will EVER compete with these prices. Heck, garage sales can't compete.
It's just too little, too late.
Man there is HUGE profit in digital Unix. There is no way in hell Compaq would walk away from that, costs be damned. The lost profit and customer base from a wholesale switch to Linux, not to mention the re-engineering costs, would be an absolute killer. Linux is great, Linux on Alpha is great, but Digital Unix still far more capable at the high end on Alpha systems and in any case the installed base would view this as a betrayal and run to Sun and HP.
Not too long ago, Digital had Maddog and other insiders working extensively on getting Linux to run on all the Alpha products. That lead to the development of Milo.
I personally don't prefer to run Linux and instead prefer a BSD variant. I own a 300MHz XLT system. Unfortunately, it runs the Windows console stuff and won't run SRM which the BSD variants all require to run.
It would be wonderful (and I'd have to say a good move on Compaq's part) if Compaq were to work with the BSD community to get the various free BSD's to work well on machines that won't/can't run the standard consoles.
The early Alpha workstations did not have a fan on the CPU, they had the hugest heatsink you ever saw and a 90mm case mounted fan blowing through it from the side. The fan was mounted in the front of the case right next to the CPU but not on it.
the celeron cant compete with em and neither can
the p2 clock for clock the Alpha's have better
Integer and Floating point performance than any
intel chip. if I remember correctly the only
chip that beet alpha was the sgi and those chips
were friggin amazing. they run slower than most
modern chips but do more Fips than any of em.
saw it on cpu maddness unfortuanatly I think the
article is down now.
jojobe
Ever heard of Netscape? I'm typing this on an Alpha Ruffian w/ 633MHz and 256MB of RAM running the DEC OSF 4.0 version of Netscape 4.51.
You don't really need to have a Digital license, knowing someone with access to the libraries is enough.
uh.. you're wrong.
>They don't appear to claim on their web page that >their port is `abandoned,' though it is >admittedly well behind the NetBSD port these days >in terms of functionality.
thats a bunch of bs to start with, but check this out....
"The alpha port was derived from the NetBSD/alpha port written by Chris Demetriou . Unfortunately we cannot track it closely anymore as some copyright issues prevent us, check our policy page for details."
> No, there are no copyright issues at this time. > Please don't spread FUD.
please dont spread lies, *netbsd troll*.
or you can simply download mozilla and compile it yourself.. i'm sure they would be happy to have input from an alpha box..
remember, its not a battle of the cpu's that makes linux good or bad, its the software, the os is only as good as its apps, windows has made intel lucky.. there processor blows comaparitably speaking, BUT it has the apps and the software to make it run, and its been running..
linux on a sparc is in the same boat, linux on a mac is in the same boat, its been an effort to build around intel and people who ultimatly have better hardware are trying to pull linux in..
but intel has an easier chance of compatibility since they and most PC vendors don't have a problem providing specs and information to developers to port for the hardware.. only recently has sun opened up and from the looks of it alpha will be opening up a bit more..
and intel has better licensing for most of its software, you don't have to pay up the wazoo to license this or that so you can build an application on that specific platform..
well, have a good one.
Where can you find the complete ISO Fortran 95 specification? - maybe a f952c++ program could be produced like f2c - although the parallel processing stuff would be "fun"...
/is/ a pretty complete implementation-more than enough for most projects. - and vf90->g77 within pentium-egcs (www.goof.com) produces zippy enough code for my (distinctly undegraduate) purposes.)
Preferably without paying for the spec, of course... Has anyone donated the spec to the FSF/egcs people ?
VAST/f90 works by changing the f90 code into code suitable for g77, and drops some of the more advanced features of fortran 95/HPF/90 because of this - mapping f95 to another OOP language, such as C++ would possibly be a better approach.
(I do use VAST/f90 myself - I'm an engineering student, and it
Then again, the egcs framework is intended to allow separate optimising language implementations without such mappings - i.e. the g++ complier is not just a preprocessor for gcc, and nor is g77. The full f95 spec would still be required for myself and/or anyone else interested ( seeing as I'm a mecheng student, not compsci, anyone else is a more likely option) to have a stab at producing a Free g95 though
Many thanks to the Debian-Alpha team and AlphaLinux developers everywhere :)
RR
eBay just had a dutch auction that ended a day ago and these sold for $250 each. Check with the seller to see if anyone canceled their order after the auction ended. These are the 21164a (not the slower 21164PC) processors.
I bet if you bought the top of the line alpha...
you would still have it in two years and it would
still be competitive with the three or four
intel upgrades you would have to buy in the mean
time....
Alpha=700mhz
Celeron=300mhz
Alpha also does two or three times is much per
each of those clock cycles. Nothing (affordable)
runs faster than Linux on an Alpha.
Well, basicaly, gong from a celeron 400 to a Alpha 500, expect a 5x to 6x improvement in floating point and a doubling in integer. I can only guess that a 21264 @ 700 Mhz would rate somewhere around :
21264 @ 700 Mhz SPECint95: 37.6 SPECfp95: 75.4
PIII @ 500 Mhz SPECint95: 20.6 SPECfp95: 14.7
key:
+ : Found these off of intel's site http://www.intel.com/procs/perf/index.htm
* : Found on a Alpha sales web site dcginc.com
! : Found on www.specbench.org
Who is is talking about Alpha on consumer market? Since the beginning it was intended for high performance servers and workstations no matter what OS they run. And there is a wide variety of OSes for Alphas including Windows NT, DU, Linux, OpenVMS ..
Just bought one. It has SPECint95 of 15.3 and SPECfp95 of 20.1
Has a 256-bit memory bus (when using 8 SIMMS). I tested it by running Wintune 95 (the Intel version) under fx32! The scores that came back, EVEN UNDER 32-BIT EMULATION, beat out the 200mhz Pentium Pro in CPU and ALL Intel boxes in everything else (memory, disk I/O, etc). Remember, this is 32-bit software (not 64) and running under emulation. You could expect the performance to more than double when running 64-bit native OS and apps.
This was a new board bought from a distributer. It is just like the newer LX and SX boards but with less 3rd level cache (doesn't need as much cache because it runs the faster 21164a processor rather than the cacheless 21164PC). Just saw a bunch of these on ebay go dutch for $250 each.
Now what was you saying about price/performance..?
try this
http://www.digital.com/alphaserver/
And the R&D
money they save can become a capital investment in chip fabricators. Intel has the big chip fab investments today, which is why they
have such a price/performance ratio. If Compaq sinks money into chip fabs, and does it well, they can again drop the price of an Alpha
Didn't Digital sell their fabs to Intel? I suppose that gives them a nice clean slate to build fabs for the next generation of 2xx64 chips, but I thought they wanted to be out of the production business.
Besides, Samsung makes Alphas, and though they might not be on the NASDAQ, Samsung is _huuuuuge_..
DEC sold its fabs to intel (using the lawsuit as a bargaining tool) because it was not cost effective to run wafer fabs for low volume high margin production and it got their bottom line in shape for sale to Compaq. Both Sun and HP outsource their chip production very effectively.
Conventional wisdom says you have to be IBM sized to run your own fabs. There are rumors that Compaq wants its own fab - they want to be IBM's size - but the best way to do that would be to pick up a state of the art chip manufacturer such as AMD, rather than build their own from sctratch.
There's lots of info there and if you have any questions just ask me or hop on the axp-redhat list and fire away.
AlphaPowered,
Peter Petrakis
--
www.alphalinux.org
ppetrakis@alphalinux.org
The best place for Linux based alphas right now is as low end single function servers where the market has historically been NT/Netware for real and perceived cost reasons. For example many dedicated Unix customers use NT or Novell for print services because the hardware was cheap even if the OS license price was ridiculously high for the functionality received. An Alpha with Linux can replace two to four x86 NT print servers for the price of one two two servers and its more reliable as well. In the ISP market Linux/Apache/Alpha makes a great high capacity web server. When the Customer needs a High performance database as well then the ISP rolls out the 14 processor Digital Unix/Oracle solution.
Let's forget about legacy code. We do molecular simulations, and all our code is f90. DEC's f90 is the only reason we haven't switched our Alpha workstations to Linux. We haven't seen a good, native f90 compiler for Alpha Linux as of yet other than DEC's. We know our codes run about 4 times faster on our 21164 Alpha-based workstations than on a Pentium II, III, or whatever. And we are just drooling over the 21264 Alphas...
We keep our fingers crossed that DEC will port f90 to Linux, but I think they will be reluctant to give up their DEC Unix license fees. However, with Compaq pushing Linux on its low-end alphas, it might not be too long before they see some benefits on high end workstations.
Comparing a new Intel to an old Alpha? Not very fair is it? Try comparing the 21264 to a PIII and you'll think different. Just bought a 500mhz alpha + mobo for $450. Even under emulation on NT, it runs faster than a Pro 200mhz. I/O is night and day (150% faster). This isn't even with native apps or 64-bit code. Goodbye Intel...Hello Alpha.
In the Bad Old Days, "big iron" vendors wrote OSs mostly for the purpose of selling their iron. The margin on such an OS is low (perhaps negative), but the tie-in value is high.
The problem with this is that you have to spend a tremendous R&D budget "keeping up with the Joneses". If you don't, you're selling a second-rate OS and thus a second-rate platform. Realize that this is R&D _just to tread water_, not to gain share.
This problem is made worse by the fact that Linux is already ported to Alphas. The "Joneses" have a tremendous R&D budget, measured in terms of engineering hours as opposed to dollars. Compaq is likely unable to keep Digital Unix up with Linux.
The worst part is that, if they keep developing Digital Unix to compete with Linux on their own platform, they split the application base. Some shops won't be able to afford to port to two Alpha platforms. The application base splits, the user base splits, and you again lose market share.
Pushing Linux is the perfect solution for Compaq. Sure, they lose their OS revenue. They also drop a huge R&D budget, merge the Alpha Unix user base into one critical mass, and assure them that they will have a state-of-the-art OS, regardless of Compaq's corporate decisions. They can also direct their attention to selling _hardware_ versus _hardware and OS_. The OS will take care of itself. They also drop the price of an Alpha to the point where people can seriously consider putting it on their desktop _and_ their backend servers.
They can still sell support contracts at a higher rate than any other support organization, because they have the Name and because they sell the hardware. They can produce new hardware drivers cheap, by pre-releasing specs to select groups of users. And the R&D money they save can become a capital investment in chip fabricators. Intel has the big chip fab investments today, which is why they have such a price/performance ratio. If Compaq sinks money into chip fabs, and does it well, they can again drop the price of an Alpha. They can increase sales volume and margin simultaneously.
In short, Compaq loses a direct source of revenue by pushing Linux. They also lose a lot of indirect costs and market share, and stand to make more profit in other ways by pushing Linux. Trust me, the dollar signs say that they're doing the right thing.
"Compaq is realizing how much time and $$ Digital poured into the Alpha. Giving up on that would be pissing money down the toilet."
Actually this is a common economic misconception. An economist would say this is a sunk cost. It is money that has already been spent and cannot be recovered, and therefore it should not be a factor in considering future investments. The MBA's at Compaq surely know this. The point is that Compaq will only continue putting money into Alpha if they think they can make a profit on it (technically, if the project has a positive net present value: the discounted value of future income minus cost.)
sorry, i'm waxing pedantic just to prove to myself that i actually learned something when i tortured myself with all those econ classes.
--ccg
chad at glendenin dot org
Try installing KDE and using KFM as a web browser.
It's almost as good at rendering pages as Netscape, is actually a little faster than Netscape, and the only thing lacking is Java support.
people dont buy alphas blindly (at least lets hope
not). lots of people buy alphas for linux. intel
helped cygnus optimize gcc for intel. those
looking for compiler performance will know this.
it speaks louder than trying to sell the dec
compiler for linux (somehow i suspect thats what you want to do) another compiler is a hard sell on
very GNUish platforms. do the same thing, youll
recover the costs in selling alphas alot faster than in sales of your super optimizing compiler.
compaq has the best catering service i have ever
seen!
"My opinions are my own, and I've got *lots* of them!"
Thats ounds good, right? Why don't they do it? Alpha is an "open" architecture, supposedly. If they don't make BeOS for the G3 because Apple wont release the G3 specs, will Compaq release the Alpha system specs to Be? Is there an interest in Be/Alpha or am I mistaken?
Might it really have something to do with Intels "investment" in Be?
It's far easier to forgive your enemy after you get even with him.
It's far easier to forgive your enemy after you get even with him.
I doubt a high-end free fortran is in the forseeable future (but hopefully I'm wrong).
:) Not that I could, anyway; he's across campus from me, and I ssh into his machine from mine).
:(
In applications for which Fortran is appropriate, performance tends to count. When we bought the dual PII linux box last year, we paid about $10k for the box, and another $1k for absoft fortran (a cray derivative), imsl, and a set of multiprocesor libraries. Had there been a free alternative, we probably still would have bought this over a 10% difference in performance--and it's much more than that. Also, we need F90 features, and could have used F95 (there's places that i have to write several lines of F90 to duplicate F95 features, and I'd kill for true arrays of pointers).
Then again, what we *really* want is Digital Fortran for Linux. It sounds like they're working on it in house, but haven't decided whether to go ahead. Had it been available, we probably wouldn't have even looked at any of the alternatives.
/flamebait{}
Open source meant absolutely nothing to us in this decision. We took linux for it's stability (ok, and I probably wouldn't have agreed to work for him if it meant I had to use NT
We needed the most bang for the bucks he'd been budgetted, and on a stable system. The choices came down to linux x86, alpha linux (too risky a year ago), and alpha DU. Given the prices that month, the x86 with linux was the best choice. Though it's a pity we couldn't have waited another month (the money would have gone *poof*), as we had to take 333's since a dual MB for 100mhz wasn't out yet
And for the record, we're tickled pink with Absoft. The tech support is spectacular; usually by the next day, and frequently the same day if we write in the morning. I even got a message back from they're tech support after I posted a question (thought it was general F90) in comp.languages.fortran. The real kicker: their tech support gets back to us every time faster than NAG sales ever got back [I've been told since then that NAG tech support is also faster than NAG sales. I have no idea how you stay in business that way.
/end flamebait for "GPL uber alles" crowd
For this, Compaq should be congratulated for. I hope they help out free software more in the future.
SPF support for most open source mail servers can be found at libspf2.
We have more than one alpha in house, and they all run Linux now. One is our Samba server, replacing itself running NT. :)
We see Linux as a free UPGRADE, cheaper than buying a new processor!
--
$you = new YOU;
honk() if $you->love(perl)
Uh, Alpha's have been available from many OEM vendors for quite some time.
The LDP lists many vendors which sell Alpha Linux systems.
I doubt it. By embracing Linux, they gain a shot at the home market which would never pay for a DU license. It allows them to break into a whole new market.
For people who need servers, though, they'll still be able to sell DU licenses, because DU does have some features that Linux lacks. For example, Digital's Fortran compiler is a lot better than g77. That's still important for people who need to support lots of legacy code.
So they gain in the home market, without hurting their server market too much. Well worth it.
Also worth noting is that Alphas have been available with NT for years now. Embracing NT probably dropped more DU licenses than Linux would, and NT doesn't gain you much of anything on the home front.
Can you give some more detailed instruction, or put up a How-To or something?
Linux is the ONLY reason I bought an Alpha. Mind you it is a clone box, not Compaq/Digital, but nevertheless, it WAS an Alpha CPU sale.
I would never have comsidered buying this system for the likes of Winbloze NT (Neutered Technology)
I have to say that I'm EXTREMELY happy with the system.
Michael J. Ball
Open Source Who's Who
http://support.lcg.org/Whoswho/ unix_guru@hotmail.com
With Intel running very quickly into the x86 architectural brick wall, and Merced proving to be the CPU equivalent of Winbloze 2.000, I think Alphas have a good chance of regaining some market share, at least server side. Especially since they run Linux faster than anything Intel has put out so far.
Michael J. Ball
Open Source Who's Who
http://support.lcg.org/Whoswho/ unix_guru@hotmail.com
Ok, I'll sit corrected. I remember that now, but my newer Alpha system has a fan large enough to cool a typical four cylinder engine bolted to the CPU.
Michael J. Ball
Open Source Who's Who
http://support.lcg.org/Whoswho/ unix_guru@hotmail.com
This company has some *KILLER* Alpha boxen, they have enough horsepower to blow away *ANY* PII,PIII box on the market. And the kicker is, that they use industr standard peripherals! Standard PCI Video cards, SCSI controllers, standard DIMMS, standard USB stuff... My Alpha board is very similar to my fiance's ASUS P2B Intel based board.
Now I think you should check on that CPU fan thing though. Mine has a honking HUGE fan on it! And, well... All CPU's use IRQ's, it's just up to the BIOS/Operating System, as to how they are implemented. You are correct in that my board assignes an IRQ to a slot though, not a peripheral.
Michael J. Ball
Open Source Who's Who
http://support.lcg.org/Whoswho/ unix_guru@hotmail.com
This is good news! Alpha rules even more! Even the old 6 or so year old Alphas here are running 24x7 without problems! Show me that with a x86!
And let us not forget that Compaq sent Linus a AlphaServer DS20 for his personal Linux development system!
Visit alphalinux.org today!
BTW Rob or Hemos, can we change the d|i|g|i|t|a|l back to the Compaq one?
"rbf"
--
ALPHA POWERED and loving it!
> the company is trying to encourage other Linux distributors besides Red Hat to support the chip.
I was a little surprised when I read this so I checked SuSE, PHT and Caldera but no, it really seems like only Debian and RedHat supports the Alpha. Or is there specialised Alpha distributions hiding somewhere?
From the specs I've seen, if Compaq wants to "help" Linux the best thing they could do would be to donate some time to gcc on Alpha. Everything I've seen shows Digital Unix (oh errr I mean "Tru64 Unix") cc being up to twice as fast as gcc. Getting a fast box is somewhat useless if the compiler throughs away even a quarter of your performance.
If only there were a usable browser, I'd use it a whole lot more.
--
As long as each individual is facing the TV tube alone, formal freedom poses no threat to privilege.
"Reactionaries must be deprived of the right to voice their opinions; only the people have that right." - Mao
Not bad.
Gotta call 'em about buying just parts tho.
--
As long as each individual is facing the TV tube alone, formal freedom poses no threat to privilege.
"Reactionaries must be deprived of the right to voice their opinions; only the people have that right." - Mao
If memory serves, SGI gave these costs as one of the reasons for sponsering Linux/MIPS development.
I believe that they make their money from selling the alpha processor and not the operating system, so in effect, they don't care which operating system you use, if it encourages you to buy an alpha processor.
There's an article in Linux Gazette to help you build that Multia. It's called Personal Linux/Alpha System: 64 bits for Under $500.
One of the reasons why Linux/Alpha is slow compared to Tru64 executables is the math library. That's why Compaq just released the libraries for Linux/Alpha.
> I've replied to the copyright FUD above.
"FUD", excuse me? The future of FreeBSD/axp was in jeopardy because of this "FUD" you refer to.
Should have known better than to try a rational discussion with a NetBSD bigot. Please go on with your life and forget I even brought up these hideous "FUD" facts.
No way! Check out the site above - the Alpha architecture is moving faster than it ever has before. You think it's hot now, you ought to see what the foundry has in store for the end of 1999.
Beg your pardon? I'm using Netscape 4.51 on an Alpha running Linux right now. Not sure what you mean by a "usable browser", because this is what darn near everyone else is using. It helps if you have the required DEC libraries sitting around (like on Digital Unix media), but there are always ways to get around that.
What you are saying would be true if the year were 1995 or 1996. The compiler issue is exactly why I put off buying an Alpha until within the last year. With the progress egcs has made, number crunching code on the Alpha really flies - and it's still not fully optimized!
Compaq donated their math library to the Linux community (check www.alphalinux.org for the link to it), which provides at least twice the floating point performance as the GNU libm. I'm using it on almost every math-intense software on my system right now.
The only directly Linux on the Alpha platform has is straight ahead. Things are only going to get better, especially in light of the speculated failures of Intel products (Merced).
Well, I'm in the opposite boat (prefer Linux over BSD), but was involved in the BSD world (NetBSD/FreeBSD) for several years alongside Linux.
My system is a PWS 500 with both SRM and AlphaBIOS installed, though the hardware is of course only supported by AlphaBIOS. Here's my understanding of the free BSD "situation" when it comes to Alphas:
There are some copyright issues with CMU which have "tied up" the NetBSD alpha port. I only say this because I witnessed a short "flamefest" on the FreeBSD-axp list in which some NetBSD folks cried foul about some code used, and threatened to call their lawyer friends at CMU about it. What I speak is true - look it up on Dejanews if you care to. I also believe this is the reason OpenBSD "abandoned" their alpha port, see their page for more details.
I think that copyright issue is the one major hinderance to free BSD on the alpha platform. I am, however, very pleased to see the FreeBSD port moving along, and planning to add support for ARC/AlphaBIOS firmware this year. Having worked in a predominately SysV environment for a few years now, I doubt I'd be happy going back to BSD - but I still am very happy to see FreeBSD's progress.
Totally agree with you there. Marketing will make or break the Alpha architecture, regardless of technological superiourity. The wise consumer will purchase an Alpha no matter what the marketing department spews, but how will that wise consumer know of the technology available unless a good marketing team presents it to him or her?
Check out www.alphalinux.org - it's a group dedicated to seeing Linux on the Alpha platform continue into the future.
If you're the same "cjs" I know from elsewhere, maybe you can clear this one up about NetBSD. As I posted in a previous comment, I /think/ the
NetBSD alpha port is held up by various copyright "issues" involving CMU - this is why the OpenBSD alpha port is supposibly dead. I do remember a blitzkrieg of flamage on the FreeBSD-axp list a while back when the FBSD crew used source code from NBSD without asking, and two or three people threatened to "call our lawyer friends at CMU" on the issue.
What's the real story on this?
Sure, you can find one at http://www.alphalinux.org/ or follow these instructions.
Compile a kernel with UFS support and mount your Digital Unix 4.x media. You need the following
files:
/etc/sia/[matrix.conf & siainitgood]
/etc/svc.conf
/sbin/loader
/usr/shlib/*
All you do is copy these files into the same locations on your Linux filesystem, and you're ready to run Digital Unix 4.x binaries.
I am very pragmatic about platform support - I believe in using the right tool for the job. It sounds like you're looking for mainly gaming support and performance, which I'm sure a Celeron can accomidate. Me, I do numerical analyses and large matrix operations that would probably gag any x86 (and believe me, I've tried).
Everything is about location, location, location. The market target for Alpha systems is not really the home user per se, but it may become as speeds increase (~1200Mhz by the end of the year) and prices drop.
See http://www.alphalinux.org/.
It's officially funded by Compaq (DEC), and usually has better information than the other places.
Buyers know this and it will limit widespread adoption of the platform (which I would offer has already missed its realistic adoption window anyway).
Any IT manager worth their salt is going to know better than to throw big money at a vendor's stepchild architecture.
This is a very good point. There have been some statements in the comp.lang.fortran group about Compaq porting their Fortran compiler to Linux.
Might it really have something to do with Intels "investment" in Be?
I doubt it. Alpha systems are mainly used for research and as servers; BeOS is intended for personal use for people doing multimedia-related design. There just isn't any overlap. As far as Be is concerned, there is no market for an Alpha port of BeOS.
If Alphas actually do start showing up on desktops in quantity, then this will probably change. I know *I'd* like one.
The use of "code" as a synonym for program seems to be popular with hard core engineers and programmers. When people talk about "codes", they are often referring to programs that do heavy duty number crunching, such as hydrodynamics, finite element analysis, weather prediction. Often in association with fortran programs that run on Crays or other "real computers".
Mea navis aericumbens anguillis abundat
I have an AS200 4/233, 8MB TGA card, 96MB ram and a bunch of drives, running RH 5.2 and kernel 2.2.3. It makes a fine server, though not as fast as a comprably priced Intel box. Apache is repsonsive and samba is tolerable.
As a workstation it sucks; the only stable web browser I know of is the KDE browser. I did run it for a while, but KDE is a dog on alpha.
Installation was also a pain... without comp.os.linux.alpha I'd have never gotten past the initial install issues and X configuration.
I'd say if you have an Alpha sitting around or you can get one for next to nothing, do it. I wouldn't recommend spending a lot of money on a machine as your workstation unless you need the computational power and are prepared to write or modify applications so they will run. Don't expect it to be easy.
Let's face it - Apple can make the margins that it does because of the operating system, not because the box is blue. There's nothing that special about apple hardware.
(Anti-Apple flame retardant -just because schools and new users like MacOS doesn't mean you have to.)
--
Business. Numbers. Money. People. Computer World.
Don't forget the Compaq braintrust really only understands selling hardware and selling services.
DU helps them in the short term because it gets them in the door for the big unix-based datacenter market (along with Sun, HP, IBM), but if a couple years Linux/Alpha can get them in there, I don't see much of a future for Tru64.
Apparently Compaq is treating the VMS community like a redheaded stepchild. Unlike DEC, Compaq will probably much quicker to drop their legacy base when it suits them.
--
Business. Numbers. Money. People. Computer World.
I've never seen an alpha in my life. I've heard various (all good) things about them... from what I've heard they are THE BOMB. Anyone who has one, can they point me to a good link for info on the hardware?
The things I heard that really struck me as cool is the lack of CPU fan and the lack of IRQs (devices are referred to by slot?).
--
Paranoid
Paranoid
Bwaahahahahaa.
Hmmm... I thought OEMing was what Compaq mainly DID...
--
Paranoid
Paranoid
Bwaahahahahaa.
From what I've seen, Compaq is providing a fair amount of support for the Alpha, and has no intention of phasing it out or slowing development. They've shown in the past they they're not married to Intel; a fair number of their home systems use AMD processors. Given the delays with Merced, they may (and should!) see the Alpha as their chance to make a pretty big dent in the performance computing market.
Digital Unix is quite a ways ahead of any of the free Unix systems out there. As has been mentioned here already, the compiler generates significantly better code than the free compilers, but also the multi-CPU and large system support is better. I certainly don't see them dropping support, or even development, for DU any time in the near future.
As far as the `lack of IRQs,' well, it depends on what sort of Alpha you get. Some of the low-end models have very PC-like hardware, and even an ISA bus bridged off a PCI bus; this obviously has the same IRQ problems as the ISA bus on any machine. But these days many people stick to PCI peripherals on both Intel and Alpha machines, and it's not a big deal.
WRT the Red Hat install pain: I've heard this from quite a number of people. I've seen word from several sites that they switched from Red Hat to NetBSD due to this pain, in fact.
Which brings up the question: why is NetBSD never mentioned in any of this? It's certainly a viable alternative to Linux on the Alpha platform. I find it rather sad that the `Open Source' hype mechanism tends to such a great degree to promote Linux and ignore other systems. (I should mention, too, that FreeBSD has a port to the Alpha in progress, though I'm not sure of the current state of it.)
Anyway, here's my quick comparison of the Unix-type operating systems available for the Alpha. I'm open to comments on any of this.
Digital Unix: Advantages: most featureful; best support for all Alpha systems. Disadvantages: not open source; costs a fair amount of money; requires SRM console.
Linux: Advantages: has massive Linux hype machine behind it (I suppose to some it's debatable whether or not this is an advantage); open source; some support for SMP?; can be run on some (many?) systems for which SRM console is not available; has DU emulation. Disadvantages: really painful install. Lack of support for many Alpha machines, especially high-end ones.
NetBSD: Advantages: better peripheral than any other system, at least for PC-type peripherals; support for more machines than Linux (particularly older ones and the larger servers); fairly easy install; open source. Disadvantages: requires SRM console.
FreeBSD: I don't know the status of this, so I won't comment. I gather it's to the stage where it's running reasonably well, though.
cjs
The world's most portable OS: http://www.netbsd.org.
I've replied to the copyright FUD above. If you're thinking of a different incident, you're going to have to dig up the messages for me, because I can't find them.
I wouldn't be surprised if the OpenBSD Alpha port is fairly dead, but that would be due to the fact that NetBSD's seen a lot of internal changes on all architectures since OpenBSD split, and OpenBSD hasn't followed these. Therefore they can't do straight code imports any more, and I don't believe they have the manpower to do the re-coding necessary to keep pulling in NetBSD/alpha stuff.
cjs
The world's most portable OS: http://www.netbsd.org.
You're better off with an AS200 or something like that, if you're looking in this price range. If you want something with modern performance, however, you'll need to get a 21164 running at a reasonably high speed (433 MHz or above).
cjs
The world's most portable OS: http://www.netbsd.org.
cjs
The world's most portable OS: http://www.netbsd.org.
You, on the other hand, not only cannot come up with any record of messages that back up your statements, but can't even get the name of the FreeBSD mailing list or port right! (It's the "alpha" port and mailing list, not "axp".)
So unless you can come up with something more than argument by unsubstantiated assertion, I think we can lay this to rest.
cjs
The world's most portable OS: http://www.netbsd.org.
Most schools use win95 as clients and Novell Netware as servers.
Look at elementary schools. They are flooded w/ macs. It was only recently that any educational institution started using wintel boxes.
-matt
This is a good move. Our research group buys alphas with Linux (without DU) because it saves us so much $$. That gives us more $$ to blow on hardware.
Compaq is realizing how much time and $$ Digital poured into the Alpha. Giving up on that would be pissing money down the toilet. But Compaq also knows how to move computers: Make them fast, reliable and affordable. Linux does these things.
How do you turn on Haifa on egcs?
They laughed at Einstein. They laughed at the Wright Brothers. But they also laughed at Bozo the Clown. -- C. Sagan
I believe you are wrong. Where can they make the profit? Digital, er, Compaq has to pay *huge* license fees to sell Digital Unix. They even have to pay license fees on the PROMs they have to ship for booting DU (the SRM console).. when they sell NT (or Linux) they can ship the ARC PROMs instead which is cheaper.
Look at the price difference between the two alternatives: DU with SRM console, Linux with ARC console. Now tell me where they make their huge profit?
Better to rip it away and sell them cheaper. Hell, if DEC had aggresively pushed Alpha workstations with Linux (in my country) a couple of years back I guarantee we had been an Alpha workstation company now, instead of Sun/SGI..
TA
Huh?
Samsug makes most Alpha chips and Alpha motherboards are available to anyone who wants to pay for them.
I think loyal slashdotters everywhere should remind the keepers of that web site of the existence of the Beowulf class supercomputer, don't you?
---
"Eternal vigilance is the price of Freedom."
You're really right about the compiler performance issues. Code compiled with gcc on Alpha machines is ridiculously slow--there is a need for some REAL work on optimization. Since most of what's in a Linux installation is compiled using gcc, an Alpha Linux installation tends to crawl as well. The horrible consequence of all this is that many people wanting to use their alpha boxes for intense floating-point calculations end up installing Windows NT--a criminal thing to have to do with such nice machines.
Alpha is the fastest, and will remain so until
at least midway thru 2003, in Q4 99' they will have a 1.2 Gz chip coming out. Gartner Group
just put out a piece detailing all the UNIX/PC
harware plans thru 2003, and Alpha is ahead at
every checkpoint. If YOU want 64 bit that flies NOW, why fart around? Sure DUNIX makes em' more
money, but it costs more too, if they can sell MORE chips because of Linux, believe me they will
Compaq + Dec are HARDWARE outfits with services
NOT software outfits. If I was running an IT shop
or a medium+ internet shop I'd ONLY spend money on
Alpha + Linux + Oracle, everything else be damned.
CPU Micromart (www.cpumicromart.com) is selling 166MHz Alphas for $100. 32MB of (required) parity RAM is about another $90. The machine comes with a built-in network card, floppy, External SCSI controller, PCI riser and 256-color VGA video. It takes a standard PC monitor and a PS-2 keyboard and mouse, none of which are included.
I don't work for these people, but I did just order one of these boxes. I know a few people who use them for Linux and speak highly of them. And you can't beat the price.
Yeah, they're not real speedy. I plan to use it to use it as a development platform for web and db-server stuff, and to dig around inside of Linux. I hear that they're fine for this sort of thing. There's another company on the east coast selling Multias for a lot more money, actually, and when I called to ask about this they said that they're configuring them and selling them as fault-tolerant web servers. They claim people are pretty happy with them.
What's an AS200? Where would I find one?
cdj
I tested one of the 1st of 4 alpha xp1000 ev6 pre-production systems last year under Linux.
I think I was the only person who did, because
everyone else used digital unix or NT. It was,
by far, the fastest Linux system I've ever used.
I have another ev6 system coming in this week.
Just one stat: BogoMips=890
That would be nice, as DU has a few nice features that could blend very well into a Linux setup. The most important thing for Compaq to do, though, and the thing that will help them to compete with Merced, is to work with Cygnus to get their compiler technology for Alphas into the hands of developers. DEC's compilers kick much in the way of ass on the Alpha platform, and the GNU compilers can't reasonably compete (except on cost, where I will always favor a GNU compiler over a $3K/yr. DEC C++ license, thank you very much).
IMNSHO, Compaq/Digital will _HAVE_ to make their compilers available reasonably if they ever want to see Alpha take off. I'd spend $200 or so for a C/C++ compiler suite from Digital for an alpha, but I'd also like that compiler suite to be able to target, say, OpenBSD/Alpha.
--Corey
Not only will they not deserve liberty or safety, Mr. Franklin, they will be DENIED both!
Doesn't the Alpha have zero development budget?
I expect Compaq just want to sell off their old stock of Alpha chips...
I have to admit that I am biased (in large part due to my employer). I would not consider a dual PII generally a good multiprocessor for numerically intensive codes. This has nothing to do with Intel or Linux, or g77/pgf77/absoft, etc.
:-) ) in favor of Linux. Market size and all that.
What I have found on my codes is that small (actually tiny) problems run well on pentia. But reasonable research sized problems cause it to huff and puff. Machines like the alpha or the R10k (and R12k) kick serious butt on the larger problem sizes. What is just insanely cool is to watch your code (efficiently) use all 32 processors, and get something like a 28-30x speedup.
But, as I said, I am biased.
Back to fortran. Jeff Templon has an excellent page on Linux and Fortran. Better is the big fortran link page. This is really a nice resource and is a nice intro to the general Fortran Market setup by Walt Brainerd. I strongly advise visiting this site if you need to think Fortran.
Ok, now some thoughts. Craig Burley and crew have done a positively bang up job on g77. It is IMO a useful productive research tool... with a caveat or two.
First, it really is just a front end to the gcc back end, so there are many... gcc-isms... floating about.
Second, while optimization is OK, it is generally tied to the gcc optimization, which has traditionally not been very good. The egcs project has had a much better track record on getting real optimization into the compiler. Folks, if your runs can take years, 5% DOES matter. Optimization on pentia is not just -O, you need things like
-O3 -malign-double -malign-functions=2 -funroll-loops -ffast-math
among others for decent performance.
Third, and most important for me, it (nor egcs) knows nothing whatsoever about multiprocessing.
In short, g77 and egcs in general are awesome tools. But unless you work on small problems, they are not suitable. You will need some better tools, and that involves passing over some money in this case.
I like the Portland Group tools, though the KAI tools are effectively identical to what you use on big supers like Origins. Unfortunately, I do not think KAI supports Linux any longer. Maybe we can all write them a nice letter on how they could drop support for some underused platform for computation (some come to mind here
As the author of the referant article wrote, most fortran users want all the speed they can get, so you need to look at what your code spends the most time doing, and figure out if it is doing it the right and most efficient way, or if your system is correctly designed for speed, or if you are hitting one area of your system really hard, and thus causing a bottleneck. In short, if you need to design for speed, start out with a workstation design, and not a PC design. You likely will need massive memory and IO bandwidth to complement an insanely fast CPU. Putting an Alpha into a PC architecture should be considered a capital crime. It makes much more sense to put it into something like a DS20, a T3E or some other design (I can fantasize about an Alpha in an Octane or an Origin, that would be a complete screamer... a memory and IO bus capable of feeding the processor at its full speed... shudder).
The language and its implementation are important, but so is the fundamental system design. You need to avoid bottlenecks everywhere.
Joe
AMD and Compaq FOREVER!
--
Beef
"Raging Moderate" of the
Alpha's don't need development money. Digital/Compaq gets them from those little pink aliens who gave us PrimeCo cellular phones.
--
Beef
"Raging Moderate" of the
They'd make a lot more money selling Alpha chips and tech support than they would simply from selling Digital ... er, I mean, Tru64 Unix licenses.
--
Beef
"Raging Moderate" of the
Research, ISPs and Education - all superb markets for Alpha - they're clued, they generally run Linux already and price/performance isn't as much as an issue as performance alone (Alpha price/performance *ratio* isn't as good as Intel, IMHO - think Irish Guiness vs. Special Brew Lager and you're someway there...)
I suggested that strategy to my former employer, but they didn't follow it, continued to (try to) sell on price/performance and went bust about three months after I left, having sold five machines.
Shame, 'cos I enjoyed playing with 500MHz Alphas.
If Alpha has zero development budget, then who's paying my salary?
In grad school I did some analysis on this subject and there is probably a 30% difference on average between -O5 output from the dec compiler and the best egcs can make.
Without a compiler, the alpha is a fairly uninteresting system. Intel chips, PowerPC chips and Sparcs can out perform it but with a good compiler it can hang with anybody.
This is my signature. There are many signatures like it but this one is mine..
Any IT manager worth their salt is going to get
the system that best solves their problem.
this "vendor's stepchild architecture" happens to be the best platform for doing hardcore computational biology and high-throughput bioinformatics. Sun? too slow. HP? no software.
SGI? good but expensive and I don't need floating point performance. Linux? cool and has potential but I want a system that can handle 12GB RAM.
I'm a huge fan of Digital Unix on alpha. Fast, easy to maintain/admin and not super expensive. I've bought $1M+ worth of their stuff for my group every year for the past several years.
It's all relative though. I know my systems are going to be replaced frequently as faster systems come on the market. Who knows what kind of systems I'll be using a year or two from now...I'll go with the best stuff available at the time. You raise a valid point though-- If I was planning on larger scale purchases with longer expected service times I'd really want to be sure of Compaq's support for the platform. That said though, the more Merced gets delayed, the better Alpha looks.
-chris
That 25% is based on better compilers - the gap disappears when running DU compiled code on alpha linux.
There's nothing to stop Compaq releasing their compilers for alpha linux though, which is a win for them because they can still make money selling good compilers, without all of the R&D involved in writing the entire OS.
When I can get the kind of price/performance out of an Alpha that I can get out of a Celeron 300A at 500MHz, I'll sign up.
So far, the only cheap Alphas I see going around are those old clunky "It's almost as fast as my near-dead P100" things... possibly a neat way to explore the architecture for cheap, but not quite in the same league in terms of bang:buck ratio as what's out there in the x86 world now.
I'm no fan of the PC platform -- hell, I'm getting mighty tired of it, having used PCs since 1981 -- but you can't beat the kind of volume that the x86 moves for creating good honest market pressure. Commodity hardware is a good thing.
We know our codes run about 4 times faster ...
I've always known code as a collective noun: "Our code is poorly commented, but it sure is fast." But I have seen the term codes used occasionally in a way that appears awkward to me, as in this quote.
I would have said "We know our code runs about..." and thought it more natural. Where does this usage of codes stem from? One generally associates the term "codes" with, for example, a series of cryptographic codes; and "code" as program code.
Mostly what I do (image manipulation, some rudementary data analysis, and development) uses flat-out integer and small matrix performance 80 or 90% of the time, and is more or less I/O bound over CPU for the majority of tasks.
Yes, the tool-fit is vastly more important that overall performance, and at least in my (fairly common) case, x86 is a win. Don't get me wrong, I'll be excited to see the Alpha descend into the mass market, but it's certainly not there yet.
Sounds highly desirable. I'm glad to see newer Alphas getting into the market. If what you're describing is a trend, I imagine my next box will be an Alpha.
Thanks, that's usable information! I've read the term often enough to get curious as to where it comes from.
There's a lot more money to be made licensing Digital Unix than Linux (be it RedHat or whatever)...
I don't think it's clearcut...
Open Source. Closed Minds. We are Slashdot.
The enthusiast market, myself included, would... and we would want Linux... I can't see myself with any compelling need for DU...
Point duly noted
Open Source. Closed Minds. We are Slashdot.
- nr
- nr
- nr
Compaq has committed to continued development of the Alpha line and has just released the next level system. Their new 21264 based systems are the fastest thing running today.
Visit Compaq at http://www.compaq.com or at http://www.digital.com for more info on the new Alpha's.
D. Keith Higgs
CWRU. Kelvin Smith Library
My office has been taken over by iPod people.
Promoting GNU/Linux to the point that new installations go there instead of to other 'nix's helps Compaq get back to what the really want to do anyway. Ultimately, they downsize the software part of the business and keep a smaller coding force on hand to continue 'nix development for GNU/Linux solutions.
As for finding Alpha systems cheap,look at places like WebAuction (run by Micro/PC/MAC Warehouse) for deals. I picked up an Alpha box from them for $350, added 4.6GB HD, 24X CD-ROM, 128MB additional RAM, 17" Monitor, 56K/V90 external Dava/Voice/Fax modem, NT4 Workstation, MS Office Pro, and the Alpha firmware for a total cost under $2500 by shopping the web and local computer shows.
D. Keith Higgs
CWRU. Kelvin Smith Library
My office has been taken over by iPod people.
linpack is a linear algebra suite, it is used for supercomputer testing because many real world problems involve large matrix manipulations and because optimized libraries are widely available.
There are two specialized versions out there: DLD and Stampede, plus SuSE did, on March 20th at CeBIT in Europe, hand out version 6.1 Beta for Alpha, with an intended commercial release in May. Additionally, all it takes for kernel support is a simple kernel recompilation, as described in the appropriate HOWTOs. This has been in effect since 2.1.xxx. Versions 2.2.0 and 2.2.1 do not require patches for it to work, although 2.2.2 does. Additionally, several of the later 2.0.xx kernels will work on an Alpha with the appropriate patch.
Goto http://www.alphalinux.org for more information
The main advantage of the Alpha over the x86 is speed. One reason why DUX is better than Linux on the Alpha is because of the reasonably powerful compilers available for DUX. Compaq can really show their support of Linux by making a complete development environment available...
"Just bought a 500mhz alpha + mobo for $450"
Where can I get this deal?
Compaq does OEM their products out! Compaq's focus is however not OEM however Digital's was. They have an OEM website, and that is actually where I got my first Alpha system.
OEM products are done through Samsung by way of Alpha Processor, INC. (API). They are bulding new motherbaords and OEM solutions.
I would take a look at our site if you want to know more. www.alphalinux.org
-Barrett
In the next week I am sure everyone will see exactly what Compaq is doing with the Alpha.
Alpha is not just a Compaq/Digital thing, it is a superscaler open industry standard 64-Bit RISC based architecture that is engineered by Compaq and manufactured in volume by Samsung Electronics, Mitsubishi, and their *subsidiaries*.
Some of the best advancements might happen much further awawy from Compaq itself.
A lot of the Alpha push is not only from Compaq but the subsidiaries of Samsung and Mitsubishi.
-Barrett
Samsung makes CPU's, they sell them to API (Alpha Processor, Inc.) that is owned by Samsung. API makes OEM motherboards for the Alpha, and Samsung sells OEM processors. Not only does Samsung do this but Mitsubishi does too, and a few other vendors as well.
Digital made and sold OEM products also, so when Digital merged with Compaq they contnued these products but they not seem to have the focus to sell a lot of OEM. Thus the Alpha OEM market will be done by others.
What I find even more annoying is that I have a Alpha 233 and A PII 233 running exactly the same
software (Redhat 5.2, EGCS 1.1.1 etc.) and with essentially the same memory and disk space and the PII can usually compile something in half the time of the Alpha.
If Intel is funding Cygnus to optimize for KNI and Merced why can't Compaq do the same for Alpha?
IMHO compaq makes the same mistake as digital to
promote the alpha chip. Alpha means Unix, and in the consumer market there is no demand for Unix. There is a demand for an system that uses smp, dsp, video, connectivity and real expansion options (not just another slot/bus system like pci,agp but like VME)with an new desktop(real time, object technology, and themes)for $1000 to $2000. The pc cannot deliver this.
And compaq is not in the position to create yet another niche.
It would be a better choice to go the way as Elbrus (http://www.elbrus.ru) and transmeta and
let the Alpha run all intel like software,
In the mean while the new AMD K7 is an opportunity as it uses the same bus protocol as the Alpha. No need to develop motherboards for the consumer market Give Be $20 mln and make an native alpha port available. Amiga inc. might be interested in the hardware combination.
As demand rises compaq can introduce more options
for the alpha like dsp, dvd hardware en/decoders on chip, which can be used by consumer software.
This way there is an future for alpha
Unix and alpha will not succeed on the long term.
Also, take a look at the IBM Power3. It's used in one of their new RS/6000 workstations (rather than a PowerPC), which is reviewed in UNIX World/Performance Computing. I think it had something like 4x the FLOPS of a PII while running at half the clock rate.