The Vax (2005)
Posted4 months agoActive4 months ago
yarchive.netTechstory
calmpositive
Debate
40/100
VaxComputer HistoryCisc vs RiscDec
Key topics
Vax
Computer History
Cisc vs Risc
Dec
The post shares a historical document about the VAX architecture, sparking a discussion about its history, design, and the CISC vs RISC debate.
Snapshot generated from the HN discussion
Discussion Activity
Light discussionFirst comment
2h
Peak period
3
1-2h
Avg / period
1.6
Comment distribution21 data points
Loading chart...
Based on 21 loaded comments
Key moments
- 01Story posted
Sep 25, 2025 at 4:17 PM EDT
4 months ago
Step 01 - 02First comment
Sep 25, 2025 at 5:56 PM EDT
2h after posting
Step 02 - 03Peak activity
3 comments in 1-2h
Hottest window of the conversation
Step 03 - 04Latest activity
Sep 26, 2025 at 8:28 AM EDT
4 months ago
Step 04
Generating AI Summary...
Analyzing up to 500 comments to identify key contributors and discussion patterns
ID: 45378413Type: storyLast synced: 11/20/2025, 2:36:48 PM
Want the full context?
Jump to the original sources
Read the primary article or dive into the live Hacker News thread when you're ready.
Supnik, Robert oral history
https://www.computerhistory.org/collections/catalog/10273826...
If you have any interest in VAX or DEC or chip design in the 80s. This is a must watch.
Later also goes into how Alpha was created in Part 2.
https://en.wikipedia.org/wiki/DEC_Alpha
and Intel’s own
https://en.wikipedia.org/wiki/Itanium
I think it’s as simple as a mass market architecture that sells more units can justify more investment than anything that sells fewer units. If DEC had been able to steal market share from the 386 with a VAX based product it might have been able to take the 386’s place but from a business perspective they didn’t want to cannabalize sales of higher margin minicomputers. The transition from bipolar to CMOS was also difficult because it did mean a regression in performance, IBM addressed this in the 390 by introducing a clustering solution but it was a bold and risky move.
Comp.arch was something really special. Guys like Mash, John McCalpin, Bob Colwell, Eugene Maya, Mitch Alsup, Terje Mathisen...folks who really, really understood computers, what the real tradeoffs are, did actual homework instead of guessing, and were generally able to discuss the issues without being jerks. Good times.
https://www.computerhistory.org/collections/catalog/10273826...
Its in Part 2.
If I remember correctly, it was basically 64bit because Alpha was basically a (good) virus inside of DEC. Or maybe like secret society that revolted against the leader of the company.
Olson had killed 32 bit PRISM and they already had VAX that was 32 bit and people making processor for it. To get people all over DEC to buy into Alpha (Alpha barley had any budget of it own) it had to be something new, and winning 64 bit did make sense.
Olson really killed basically everything that make sense, that DEC survived so long with Olson as CEO is kind of crazy. The amount of horrible decision starting as early as the early 70s is kind of crazy.
VAX had also been early, driven by Gordon Bell not Olson, very few of the competitors had 32 bit processors then, and people like Data General and Prime struggled to develop them in response. Funny enough a hardware guy on the VAX team basically proposed RISC-like architecture but it was rejected because they optimized for code size. To bad that they didn't hit on the idea of compressed instructions.
DEC was (slowly) on its way to making alpha mass-market. Around the time of the Compaq acquisition, they started offering the CPU to 3rd parties to design their own boards. One example is the API UP1000, which had an AMD irongate chipset and an EV6 CPU. (I had an early sample that I ported FreeBSD/alpha to). Alpha was the only somewhat popular non-x86 platform that Windows NT ran on. FX!32 made it possible to run x86 apps, like Rosetta on Apple M4. Alpha was the first non-x86 linux port.. I was at the USENIX where John "Maddog" Hall gave Linus an Alpha (I think it was a Multia) Alpha was also the first 64-bit FreeBSD port (which I contributed to), and paved the way for amd64.
Alpha had a HUGE amount of momentum when HP summarily executed it. I still hate HP with a passion, and I still won't buy anything from them, 20+ years later.
What I also know is that DEC chose ARM for low-power applications, because the design of the Alpha was simply not capable of scaling to lower power usage.
"According to Allen Baum, the StrongARM traces its history to attempts to make a low-power version of the DEC Alpha, which DEC's engineers quickly concluded was not possible."
https://en.m.wikipedia.org/wiki/StrongARM
Power efficiency was soon to become a chief concern. This alone would have ended the Alpha.
Interesting. One of the main design principles for the Alpha was for longevity, and to have a thousand-times increase in performance over 25 years.[0]
[0] https://danluu.com/dick-sites-alpha-axp-architecture.pdf
"ONE LAST TIME: it wasn't politics that ended the VAX, it was engineering judgement by excellent (IMHO) DEC engineers, and the economics of doing relatively low-volume chips [low-volume compared to X86]. Nobody could afford to keep the VAX ISA competitive at DEC volumes...."
In contrasting this against Intel, what x86 could always obtain was and is performance. A large base of buyers will enable architectural improvements.
As an example, Bob Colwell was brought in to implement out of order opcode/microcode execution in the Pentium Pro. This was contemplated for the VAX, but never seriously entertained.
What could not be added to the Alpha was power efficiency, and x86 finds itself regularly beaten in this race as well.
It seems that money can buy only some things. Performance can be bought, but efficiency cannot.
___
EDIT: I have read this before, but it's always funny.
MIPS was the most provocative and frightening thing in high performance computing in the 80's and early 90's.
"IBM, DEC, and HP were among the few that actually had the expertise to do CMOS micro technology, albeit not without internal wars. [I was invovled in dozens of these wars. One of the most amusing was when IBM friends asked me to come and participate in a mostly-IBM-internal conference at T. J. Watson, ~1992. What they really wanted, it turned out, was for somebody outside IBM politics (i.e., "safe") to wave a red flag in front of the ECL mainframe folks, by showing a working 100Mhz 64-bit R4000.]"
Most importantly, outside of packaging the core in SoC-like setup, Digital essentially could only ship one microarchitecture at the time with Alpha.
But I think generally speaking you are right, the size of the teams that all the RISC people had compared to the amount of resources Intel had was not really comparable.
> The transition from bipolar to CMOS was also difficult because it did mean a regression in performance
DEC already had the VAXCluster by 1983 and they were working on a bunch of implementations at the same time. They did a lower performance VAX already when the did a AMD 29000 bit-slice implementation in 1982. So they did know the value lower performance. By 1985 they had matched the 780 in a single chip. Basically by 1988 the CMOS version was already faster then anything else.
The transition was really just difficult because some people (who happen to be CEO) didn't want to see the reality.
At the point Alpha got killed, it was the fastest option, and then HP had to resurrect it anyway because the supposed Alpha killer (Itanium) was slower in practice. That said, the killing was done by Compaq, not HP - and done fully on politics/vibes not actual thorough thinking. A lot of late Alpha design ended up being "inherited" in ways by later x86 cores - Athlon reused chunks of EV6 architecture which also enabled cheap Alpha motherboards (but made Athlon-MP extra expensive, unfortunately). K8 heavily drank from experience of the team members who worked on EV7 which exhibited the same fabric structure.
[1] https://simh.trailing-edge.com/dsarchive.html