Colour me positively surprised, as I had no idea Alpha emulation had progressed this much.
As you might know, I’m involved a bit in the OpenVMS community and the Alpha emulation side via AXPBox. AXPBox (github) is a fork of the es40 alpha emulator by Camiel Vanderhoeven (who is now Chief Architect at VSI, the company that makes OpenVMS, for x86 nowdays). There have been many forks of es40 in the past and recently a new one has popped up with some great new features. Like speedups via a JIT compiler, S3 graphics port from MAME and ARC support, resulting in the ability to run Windows 2000 for the DEC Alpha.
↫ Remy van Elst
Not only can you run the unreleased Alpha version of Windows 2000 on this forked emulator, it’s also capable of running OpenVMS and Tru64 UNIX. In fact, both OpenVMS and Tru64 can run their full X11 CDE desktops on the emulator as well, which is incredibly cool and a huge milestone. As the name of the original emulator implies, it’s emulating an AlphaServer Es40 from the turn of the century, which should be fast enough for enthusiast use.
The last AlphaStation ever made, the ES47, is still very high on my list of computers I desperately want but will never have – they are incredibly rare, and whenever they do come up for sale, incredibly expensive. If you have one, consider yourself lucky, and please, write about it! Tell the world!
Alpha is once of those true missed opportunities.
I still get a reoccuring fever dream of DEC buying Apple and Be.
Macs running a BeOS derived OS on Alpha processors just seems such a weird but somewhat brilliant combination. Of course dual booting your Mac with VMS and Windows 2000 AXP would also be a cool but cursed combination
What stuff are you abusing that you dream about putting together 3 of the worst managed companies of the mid 90s? 😉
FWIW if it makes you feel better ever since x86 cores went 64bits, they are pretty much what Alpha would have ended up being. Since the design teams behind it went to either AMD or Intel.
Xanady Asem,
When AMD engineers released the AMD64 platform, it changed the trajectory of many companies with one strike.
Intel abandoned “Itanic” (and somehow was shamelessly called this EM64T)
hp’s server dreams dies by extension
DEC Alpha, PowerPC, and other alternate chips died out.
SPARC and entire Sun Microsystems had no strategy to counter this.
No more “clean break” nor death of archaic x86.
AMD64 was not perfect, but having good backwards compatibility and cheap availability sealed everyone else’s fate.
(I honestly did not know how to even get my hands on an Itanium workstation back then without paying exuberant enterprise prices to hp)
It is not a given that Alpha would have lost to x86 despite the compatibility advantages you mention. Alpha had quite a lead at the time and, with Windows NT being a different market and ecosystem, there was not as large a legacy of x86 software in the parts of the market where Alpha was strongest.
Alpha was squandered by DEC who did not know how to give up VAX margins to pursue a larger Alpha market and then by Compaq. But when Compaq failed overall and was bought by HP, the Alpha’s fate was really sealed. HP was all-in on Itanium. HP did not want Alpha to succeed because they wanted Itanium to serve the same customers.
So, ironically, it was Itanium that killed Alpha but not because it was better. This led to an exodus of Alpha engineers to AMD where they created AMD64. And it was AMD64 that killed both Itanium and Alpha for good. If HP had backed Alpha instead, there may never have been an AMD64 and Alpha may well have succeeded over x86.
But AMD64 may still have happened anyway. Perhaps it still would have won out in the end. Or maybe we would still have both x86-64 and Alpha today. We will never know.
LeFantome,
Alpha had technological merit, but I agree with sukru the x86 monopoly gave it an enormous advantage over the others.
Yeah it is interesting to ponder what might have happened.
Intel at the time was determined to have itanium become their 64bit platform, but they were leaving behind their massive x86 developer & software ecosystem advantages (itanium x86 emulation was terrible). On top of that itanium was inaccessible for users and developers alike. Whereas Alpha failed to compete against the x86 monopoly head-on, they might have succeeded at competing against itanium. So AMD may have inadvertently helped to save Intel by keeping x86 relevant in a 64bit future.
@LeFantome
Alpha was effectively dead by the mid 90s. The out of order implementations after EV6 became extremely expensive to design, and AXP turned into one of the albatrosses around DEC’s neck, one of the major factors that helped push the company toward collapse. There was simply no realistic way to recoup those design costs through VMS, Tru64, or NT.
By the time HP acquired Compaq, Alpha had already been mothballed for years.
That was the broader story for most high end RISC architectures of the 90s. They started losing momentum once x86 went out of order, and that trend only accelerated after x86 gained 64bit extensions. The markets those RISC vendors targeted were simply too small to justify the escalating cost of high performance CPU design. That dynamic did not really shift again until ARM successfully scaled up into 64bit.
Ironically, a lot of Intel’s 64bit x86 cores during the 2000s and 2010s still carried some Alpha DNA in its development process. The original design and simulation flows used AXP-based models to define many of the fetch and execution structures, and ran Alpha binaries during early performance exploration. The x86 front end was added later on the design process.
Xanady Asem,
A huge issue for them was just getting mainstream software support. Even windows NT itself limited support to a 32bit OS running on 64bit Alpha CPUs.
https://devblogs.microsoft.com/oldnewthing/20170823-00/?p=96875
Microsoft never released a 64bit version even though the hardware was capable. So software devs couldn’t even take advantage of native 64bit addressing without inefficiently thunking application data through windows 32bit API.
For example…
https://learn.microsoft.com/en-us/windows-hardware/drivers/kernel/why-thunking-is-necessary
Obviously this would nullify one of Alpha’s big advantages over x86 at the time and it just goes to show that developing the hardware wasn’t enough to compete. On top of hardware, an architecture really needs a mature OS and buy in from software devs to succeed.
On the ES47, as was typical in the day. “Create me a workstation!” We have a rack mounted server… err.. wait a sec.
Even though I don’t think I’ve seen the rack mount version, it definitely looks like the typical HP approach to a workstation (of that day).