The Death of Alpha on NT (1999) | Hacker News Reader

The Death of Alpha on NT (1999)

itprotoday.com

62 pointsxattt43 comments

I remember Alpha in the 90s, it seemed like it was going to be the RISC architecture to finally take on Intel. I was so sad when it fizzled out and I was never able to get my Alpha-based computer(since its heyday was mostly while I was a teenager). Especially since it was the RISC architecture that seemed like Linux was going to work the best on. Eventually I had to settle for Mac OS X and PowerPC as a my accessible RISC Unix work station.

I can relate. I was learning 3d modeling and animation software at the time, on a lowly P120, and it'd take all night to render even the shortest low resolution clips. Meanwhile I'd read trade magazine stories about what the Alphas could do with the same software, dreaming of maybe one day getting to work on a machine like that.

That's crazy, I was also on a P120 machine looking longingly and Alphas and how much better they were. They just seemed so fast. If I'm not mistaken, they were also one of the early ones to 1GHz.

That they were, though as a sibling comment points out it was apparently a mixed bag, as you really needed native apps to tap into the performance. They were slower than Intel when running software that required emulation.

So for the stuff I was interested in, namely LightWave3D, which was native, they were a big jump. If you were using Visual Studio or Office, not so much.

If I recall they were marketed as “screamers” for Lightwave3D distributed rendering. The main customers seemed to be the 90s TV visual effects houses (Babylon 5 etc).

I ran a Softimage house in the mid-to-late 90s and we had SGI Indigo workstations and a Dec Alpha mental ray render farm of about 30 machines or so. There were certain renders, scenes, particularly anything with ray marching or volumetric light, where we couldn't have the SGIs participate in the render because their contributions were slightly different and resulted in a distinct checkerboard look.

I loved Softimage — sad when the big 3D apps all consolidated and we lost the competitive development of Wavefront / Softimage / Alias. I’ve been getting into Houdini this year — the app is still going strong, and actually affordable (vs back then when these apps, let alone the machine, cost as much as a car).

We had a seat of Prisms, the pre-Houdini offering from SideFX, purchased solely for one job. I found it inscrutable and the documentation was… well… it seemed as though the writer(s) just gave up even trying to explain anything.

You weren't missing much. The Alpha workstations most of us got to play with were dog slow. It was a funner architecture to read about than to use.

They had a whole room of Alpha workstations when I was at University in the late 90's, and I always preferred them to the Sparcstation machines in the main computing room as they were quite a bit more responsive.

The Sparcstations ran pretty much okay on Linux though, even though in benchmarks Solaris seemed faster.

yup. I had my choice between my cheap linux PC, Sun workstations, SGIs, and Alpha. The alpha (in 1994) was by far the fastest machine of the bunch.

If you want a deep dive into the Alpha AXP architecture from a Microsoft perspective, Raymond Chen has a wonderful series in his blog:

https://devblogs.microsoft.com/oldnewthing/20170807-00/?p=96...

As much as I love Cutler... I'm also impressed a news website that existed in 1999 still exists today in 2021!

CSB: He called me an asshole once. I got a prototype x64 delivered to my office (at Microsoft, circa about 2000), and I politely asked the Win64 distribution list where to get a build to install on it. After a week of silence I wrote back and said something like "without the OS it's a doorstop to me". Apparently he took umbrage at that.

My manager once presented to Cutler. They said it went remarkably well as Cutler "only interrupted the presentation twice to cuss"...

Worked in Microsoft back in the day, as a QA, and we had an extra Alpha NT machine which was used as a test machine for our software. I didn't know much about Alpha architecture at the time, just that it was this non-Intel CPU with some strange assembly language, and I remember being amazed that NT looked and seemingly behaved the same on that CPU as on my own Intel workstation.

Then one day I was told that Alpha machine is to be taken, there was no need anymore to test on it. And that was that.

I never learned the full story until much later.

The referenced "Intel Merced" took me back -- that was Itanium. I remember it being super hyped, before the advent of amd64.

The merced / itanic hype contributed greatly to killing off most of the vibrant bazaar of 90s RISC architectures . Eg, Alpha, PA-RISC, SGI mips64. Sun's Sparc and IBM's power were the few high-end RISC chips to survive the itanic hype.

As a huge alpha fan (user since 1992, helped/maintained the FreeBSD/alpha port, etc), I was so happy when AMD's amd64 beat itanic in the market, and when Intel finally killed the itanic I was ecstatic.

The recent Ryzen/ARM offensive makes me happy for similar reasons (and I even bought Bulldozer-era AMD processors because I didn’t like the idea of buying from Intel)

I remember ordering a free copy of the IA64 architecture book from Intel in the early 2000's(actual paper even!) thinking I was getting some head start on the future of computing(when it was clear that Alpha wasn't going to happen), but that never panned out. Clearly I've got a poor track record in guessing the path of CPU architecture(I even went 32-bit Core Duo over AMD64, once I got disillusioned over the idea of "better tech" winning out), so I'm still hoping that my buying an M1 Macbook Air doesn't doom Apple's move to ARM.

RISC-V may win over ARM. Would be nice.

If I recall correctly, Alpha was important in the history of the Linux kernel. Originally the Linux kernel only worked on Intel chips, but the effort to modify it so would also work on Alpha set the stage for it to run on many architectures.

The Alpha folks talk about being the first port away from x86, but Linux/m68k (1993) predates the Alpha port (1995). I guess the difference is that m68k was happening independently and lived out-of-tree for a long time, while Alpha was happening with the direct involvement of Linus and lived in-tree from the start.

I think the Alpha port may have been more significant because it required making Linux work on SMP systems with a very weak memory ordering model, in contrast to x86's strong ordering. I'm not aware of any SMP 68k systems, but that was definitely an important part of the Alpha ecosystem.

I remember going at the Compaq workstation launch of what would be the last Alpha workstation. They would tout how great it was and so on, but never bench it to any Intel counterparts (hint: it was destroying Intel for 3d rendering). For people using lightwave initally on Amigas, running on an Alpha was the equivalent of going from a raspberry pi to a ryzen 5950. I was barely recovering from commodore killing my beloved Amigas, I had to witness this crap all over again. The only good thing that came out from all of this is that I am not longer emotionally invested in any platforms anymore. Well yeah… Amiga is still in my heart ;).

NT wasn’t the only Lucy-and-a-football of that era. I worked on processor independent Netware (Alpha, MIPS, PPC, PA-RISC, perhaps another I’ve forgotten). Or course it all was for naught, though I got some good stories out of it.

Was this for Cygnus? I know Apple was talking a lot of crap about PIN for the Workgroup Servers (at least what became the 9150, anyway). Did the PowerPC version ever boot on any Power Macs?

Yes it was. I’m not sure anyone had anything booting before it all unraveled (perhaps that was why).

Netware and its history is something I'd love to hear more about. I got into IT support in 1995-96 and saw a fair amount of Netware 3.12 and later 4.11 but chose a Microsoft-centric path for my "day job" and a Linux-centric path for my personal computing. Netware seemed interesting but much harder to find low-level details about. NDS was interesting, but the promoted advantages over NT 3.51/4.0 domains weren't an issue in the size of NT deployments I worked on.

I'd certainly never heard of this processor-independent Netware effort. Its history sounds intriguing.

I worked at Novell in 2006-7 after they bought (what became) Domain Services for Windows from my company. This enabled eDirectory to become an Active Directory compatible domain controller. Integrating the two worlds (whilst imperfect) was a lot of fun and there were some great people there. And, as someone who loves computing history, it was great to delve into the back story of NetWare and NDS.

YouTube Computer History Museum "Oral History of Kanwal Rekhi" (Excelan, Novell CTO) https://www.youtube.com/watch?v=ox0e7yVgsXM has some interesting stories about early Ethernet adapters made by Excelan, Novell, their strategies, Lite products and bonkers ideas (Mormon run, trying to compete with Microsoft, hate of Unix despite owning it).

About a few years after this article came out, I was working for a local hardware reseller. One of the engineers on staff for support told me about NT on Alpha. He also told me there was NT for PowerPC somewhere in a development lab at Microsoft. I didn’t get that far into digging, but he made it sound like Microsoft was doing this as a backup solution should Apple want to offer it.

There was. NT3.51 and NT4 were available for PowerPC (and MIPS) and was commercially sold, but only for IBM's workstations and the PowerPC ThinkPads, not for Power Macs. It was dropped by W2K.

And it's a single product - the NT3.51 and NT4 CDs you'd use to install an x86 PC also support PowerPC/MIPS/Alpha. Googling suggests you can run the MIPS version inside QEMU, but PowerPC is as much a bust as it was on PowerMacs.

IBM RS/4000 Unix computer PowerPC workstations? They came with AIX.

Post-Copland, Apple apparently looked at Solaris and the NT kernel (source: Wikipedia), before BeOS and NEXTSTEP.

The operating system runs on the chipset. To me, the title is nonsensical but perhaps that's the point?

I apologize for nitpicking, but shouldn't that say "NT on Alpha"?

The Alpha died, not NT.

Maybe: The Death of Alpha below NT

I copied the article title verbatim.

NT on Sparc was even more fun.

Huh? does SPARC even support little endian?

Reply on news.ycombinator.com