Guest post by What a long, strange trip it’s been. MIPS Technologies no longer designs MIPS processors. Instead, it’s joined the RISC-V camp, abandoning its eponymous architecture for one that has strong historical and technical ties. The move apparently heralds the end of the road for MIPS as a CPU family, and a further (slight) diminution in the variety of processors available. It’s the final arc of an architecture.
Interestingly, MIPS and RISC-V share an architect in Dave Patterson, and MIPS could be seen as an ancestor of RISC-V.
Good news – MIPS lives on, and will not fade to black like the DEC Alpha.
I miss those beautiful SGI Workstations with 64-bit MIPS RISC architecture. Computers today are a bit bland and all look the same.
All in all, great news.
That company is not the “old” MIPS, it’s basically a new company that bought the “MIPS” trademark.
MIPS as an architecture had been in life support since 2001.
FWIW Alpha has lived on in a few x86 microarchitectures. The team that did alpha went on to do Opteron and Ryzen at AMD.
The Alpha isn’t dead either. Whilst there’s very little evidence (for obvious reasons) it’s rumoured that Alpha processors, and descendants thereof, are used quite heavily in chinese military technology
https://en.wikipedia.org/wiki/Sunway_(processor)
Again, the chinese also have a line of MIPS64 processors called Loongson:
https://en.wikipedia.org/wiki/Loongson
Those processors are not descendants of Alpha. They simply used some parts of the Alpha microarchitecture as inspiration. The AXP21264 was a very clean microarchitecture that inspired a lot of designs after that.
So people are running a DEC-descended OS (Windows NT, derived from VMS) on a DEC descended processor, plugging in peripherals using a primarily DEC designed interface (USB) and communicating to the world on a DEC designed interface (Ethernet)
And people wonder why i’m a DEC fanboy, that damned company invented everything.
I wouldn’t use the terms “descended” and “primarily designed” as liberally as you did in that post. Ha ha…
OK OK i might have have been pushing it a bit with USB. But it’s chief architect came from DEC before they joined Intel, and Compaq (which bought DEC) were also heavily involved in the process later on.
Maybe “descended” wasn’t the right word to describe the Zen architecture, but if many of the architects and designers of Zen came from DEC, it’s undoubtable that lessons learned developing the multitude architectures DEC made in the 80’s and 90’s would be carried on into the current products that they’re designing.
Windows NT was designed by Dave Cutler, lead designer of VMS in it’s later years, and Windows NT’s native API is very closely related to VMS’s API, including the kernel architecture, and the way NT handles drives. A bit more loosely related, but DOS was basically a CP/M clone, and CP/M’s architecture and command structure was inspired by DEC’s RT11 for the PDP11, so there’s DEC heritage from both sides.
Ethernet, hands down, was designed by DEC. You can’t argue that one.
Ethernet was developed by Xerox, FYI.
Honestly, to me DEC’s main influence was the PDP which is the architecture that defined C and Unix. In one of those historical “accidents” in the field of software where free becomes ubiquitous over better but expensive alternatives; almost every single microprocessor ISA ever since has been designed to be the target of a C compiler and support the memory management of Unix. Which basically means 99% of processors out there are just very fast 32/64 bit PDPs.
To get the discussion back on track. DEC was one of the launching customers for MIPS CPUs and in turn MIPS influenced heavily the original Alpha.
You mean the PDP-11
The “PDP” series consisted of about 4 separate and otherwise unrelated computer architectures.
There was the PDP-1, PDP-4, PDP-7, PDP-9 and PDP-15, which were iterations of the same basic 18-bit ISA. It was the PDP-7 that UNIX was originally written for. Spacewar! was originally written for the PDP-1. The PDP-2, though very little was known about it, was a 36-bit variant built for some 3 letter agency.
Then there was the PDP-6 and PDP-10, which were an unrelated 36-bit architecture, basically DEC’s “mainframe” range. These were physically quite large and quite expensive. CompuServe built their network on PDP-10’s and derivatives.
There was also the PDP-5, PDP-8 and PDP-12. These were again, unrelated to the other architectures, and were 12-bit. The PDP-12 was actually a PDP-8 with a LINC minicomputer essentially bolted on top. These were very inexpensive (the 8 range in particular) and was DEC’s first big money-maker in the computing industry. They were primarily sold to laboratories and factories, and did a lot of early datalogging and automation tasks.
And the most famous of all the PDP’s was the PDP-11. This was DEC’s major cash-cow, and was sold in one form or another from january 1970 to some point in the late 90’s to early 2000’s. Again, it’s architecture shared very little with the other PDP families, and this machine was 16-bit. This is the machine most people think of when “PDP” is mentioned, and is the machine that C was designed on. UNIX was later ported to the PDP-11, but never originated on it.
Yeah, I meant PDP-11, the impact of the others is rather irrelevant. The 5th or 6th edition (I forgot which) of Unix was done on that, which is what most people associate with Unix.
What do you mean, IBM already invented everything in computing by the 1960s, everything since then has been a re-hash 😉
“..communicating to the world on a DEC designed interface (Ethernet)”
While DEC did promote ethernet as a standard, it was actually one of many foundational technologies created at Xerox PARC.
Yeh ok I’m an idiot
There are two Operating System’s in the world: Unix and VMS.
😉
DEC VAX – invented virtual memory (probably didn’t but it seemed like they did).
I miss SGI Workstations and Sun Ultras too. We had some SGI O2 workstations and some Sun Blade 1550 and Sun Ultra 25 boxes in my university.
And let’s be honest here, the demise of the SGI and Sun workstation is all the fault of those RISC fanbots who convinced large portions of the IT crowd that it was about the instruction set and not the OS. If SGI and Sun had moved their OSes in time to x86 and started targeting more mainstream markets they would have a fighting chance. Imagine Solaris with Project Looking Glass on your laptop, or IRIX updated and sold in a laptop.
Only Apple was able to realise that the thing that makes their machines special is not the ISA but the OS, which is why the Mac line survived, and Macs are the closest thing we have to Unix workstations today, yes even with the recent iOS-ification trend taken into account. We may look favourably to M1 now, but there was a time nothing had a fighting chance against x86.
SunOS/Solaris ran on x86 for as long as they ran on SPARC FYI. And regarding SGI, by 2000 most of the technologies that made Iris special were on linux, but SGI was doomed anyway it was an idiotic company (I worked there).
What made Apple special was the applications and their marketing department, not their OS. Classic MacOS and the 1st versions of OSX sucked big time.
MacOS X Panther was good, and Tiger was so good it actually shocked Microsoft execs:
https://www.informationweek.com/microsofts-vista-had-major-mac-envy-company-e-mails-reveal/d/d-id/1051287
It was around that time (aka Panther to Tiger beta) that Steve Jobs realised they had an excellent OS held back by IBM’s awful PowerPC processors (especially on laptops). Nothing wrong with the PowerPC ISA, it’s just the investment in x86 implementations at the time was so huge that nothing else had a fighting chance, and IBM’s investment in PowerPC in particular (not POWER, that’s a different thing) was literally contractual-obligation minimal.
Sun could have achieved something similar to Mac OS X with Solaris and Looking Glass, but back then Sun had the bad luck of being run by PonyTailDouche™ (Jonathan Schwartz), who was too busy open-sourcing the company’s crown jewels for the behest of RedHat to do anything else. He seriously thought Sun was a hardware company and the software could be given away for free or near-free because they would sell overpriced SPARC systems forever. Instead of viewing Sun as a software company making hardware, like Steve Jobs saw Apple. Things like dTrace made Sun the Apple of servers.
Solaris didn’t always run on x86, not by a long strech. But it did have a short-term affair with it that burnt some people:
http://web.archive.org/web/20180102220610/http://www.sparcproductdirectory.com/view62.html
SGI could have become today’s NVIDIA with their own OS too, but they decided that x86 was beneath them and they would only ship Windows PCs ironically (aka without any SGI silicon in them, just a vanilla PC box with an SGI badge). Trivia: Many 3DFX and Nvidia employees came from SGI.
BTW, I do have to give credit to the PonyTailDouche™ for finally moving Solaris to x86 in earnest, but what’s the point of it when your crown jewels (software) are being given away to Red-Hat as open-source? Did this guy think that SPARC boxes sold to legacy businesses (those which couldn’t move to x86) would make up for the loss of licenses? Believe it or not, Solaris was supposed to be a premium OS you paid for is some way or another, like Mac OS X is today. On the desktop/laptop *and* on servers. Sun actually had the only premium server OS with tons of unique features (like dTrace and ZFS) before the great open-sourcing. All they had to do it port it to x86 and put a price on it. They did the hard bit (porting to x86) but not the easy bit. Unbelievable.
And for bringing about the loss of the premium OS called Solaris, Jonathan Schwartz will forever be known as the PonyTailDouche™.
Thanks for coming to my TED talk.
Sun couldn’t have done what Apple did. Just like how Apple failed when they tried to get into the enterprise.
Operating system don’t sell units (unless for the tiny ultra geek market), applications running on those OSs do. Solaris had zero value proposition for consumers.
SGI also lacked the company culture to do anything that wasn’t ultra high margin products.
Both SUN and SGI fell victim to the comoditization of the technologies they helped introduce. They simply lacked the supply chain and distribution networks needed to address the consumer market.
Also what’s with the insults towards someone who you haven’t meet. Taking these things to such personal level is creepy IMO.
Yes, I am a bit over-the-top, but I consider Jonathan Schwartz the reason Sun got eaten up by Oracle, because he took a company with survival potential and went full-steam ahead with the Gnome Underpants Strategy of giving out things to competitors as open-source. Now Solaris is dead (I know Illumos exists, it’s not very active), Android is getting sued over Java APIs and lots of Sun customers ended up Oracle hostages.
Have you considered that the reality of events was a bit more complex than the silly cartoonish narrative you got there?
No, I haven’t. Jonathan Schwartz took a company with survival potential and ruined it by open-sourcing (aka giving away to RedHat) things that cost Sun Microsystems a ton money to develop. It’s the kind of thing that happens when you read one too many Richard Stallman lectures without taking them with a grain of salt and considering that good software takes money to develop.
Sun was the Apple of servers (and not just webservers, the Linux crowd tends to confuse the two). The kind of companies that relied on Solaris’s unique features wouldn’t go to Linux without those features, so Sun had a niche, a very high-margin niche, to keep them alive. Just like Apple’s Mac OS is on the desktop. While Windows is bringing in >90% marketshare but trapped in the WallMart end of the market (saving for gaming laptop peculiarities), Apple has a single-digit but incredibly profitable niche at the very top. This is what Steve Jobs realised and stopped hunting marketshare and licensing MacOS to other companies for Windows license prices. Jonathan Schwartz should have done the same. Market it as a premium OS to the niche prepared to pay for it. Now imagine Apple giving away their graphics stack, GUI and iMovie under open-source to the behest of Canonical, and you get the point.
I am sorry but you are very badly informed.
It’s silly to compare the enterprise with the consumer markets. Sun wasn’t the Apple of the enterprise and vice versa.
SPARC IV and then Rock were 2 massive disasters that SUN never recovered from financially.
This is what you (and in all fairness Jonathan Schwartz) didn’t realise. Sun Microsystems was never a hardware company. Their first computer didn’t even have a Sun Microsystems CPU inside. They made CPUs because the Motorola CPUs had been getting long in the tooth. SPARC had its moment in the 90s sure, but was nothing too special overall and had no long-term chances against Intel and AMD (and their budgets). By the early-to-mid 2000s, nobody really wanted SPARC CPUs, the bang-for-buck was not there. But a good number of high-margin niche customers paid up regardless because they wanted the software. But if you have watched too much Richard Stallman and Eric S Raymond, you have probably been convinced that the path to solvency involves selling hardware (which is traditionally a low-margin, high-grief business that only the really big boys can play in).
Jonathan Schwartz once publicly stated that the move to x86 undermined the hardware platforms on which Sun depended for revenue – hardware systems that ran only Sun’s Solaris. This guy seriously thought people were buying SPARC for their underwhelming bang-for-bag and the industrial box designs, and not because they wanted DTrace, ZFS filesystem, Solaris Containers and xVM Ops Center, and other things you want in a financial services server where back compact and dependability matters (but don’t want when running in a server running PHP server on a LAMP box, Sun never had a chance there). All he had to do was put a price in the box and say “alright folks, we all know you really want the software, so pay up” or go the Apple route and say “yeah, it’s a PC with a special chip which Solaris needs in order to boot sold at a huge markup, now pay up”.
Look, I understand Jonathan Schwartz is viewed as a hero (or useful idiot, depending on your cynicism) in Linuxland, but he wrecked the company, leading to a trail of layoffs rivalling the dot-com bust and a ton of customers facing the displeasure of becoming Oracle hostages. But since those Sun employees got a fat severance paycheck and were hired by other companies and the Sun’s customers were greedy financial services firms, he served the greater good overall, I guess?
SUN was as much a hardware company as they were software. SPARC was fundamental to sun growth, and most of their revenue came from SPARC-based systems.
Again, I am sorry but I fear your education in this matter may not be very accurate. Open source was very beneficial for SUN, and it was pushed internally and well before their last CEO. SUN troubles predate Schwartz’s tenure. SUN never recovered from the dot-com recession. And their business model couldn’t adapt fast enough to the new commodity-driven datacenter architectures,
Other than a few geeks nobody gave a shit about dtrace or ZFS as main purchase/sale drivers. Those were not the “crown Jewels” of SUN whatsoever.
SUN’s business model was based around high margin integrated systems. Unfortunately, that meant they could only target markets that were growing much slowly than were the big growth areas were happening. Which is why they ended up being bought by Oracle, because in the end, SUN was basically an Oracle VAR.
Opensource helped SUN tremendously, actually.
What did SUN in was SPARC. Period.
I disagree. Linux is the *nix standard bearer these days, for better or worse, and it’s easy enough to find equipment to install it on.
Dell Precisions can be purchased with Xeons, ECC RAM, and RHEL or Ubuntu pre-installed.
System76 makes some workstation level desktops based on Threadripper or Xeon procs.
Raptor Engineering and their Power based stuff.
Intel NUCs now have a Xeon flavored version, and they can be purchased with Ubuntu pre-installed from the correct source.
The Raspberry Pi 4 has 8GB of RAM and dual HDMI outputs now. It doesn’t have internal SATA, but it’s getting there.
GPU compute on Intel or AMD graphics chips is probably the last thing that really missing from Linux workstations. I do realize the old Unix workstations didn’t have GPU compute, but that seems like the kind of thing they would have.
I said “Unix workstation”. Linux is not Unix, it’s Unix-like. Also, Desktop Linux comes with too much baggage for me to consider it the same or a replacement of the hallowed Unix workstation OSes of old, for example having the whole dependencies and respositories nonsense front and center (instead of standalone packages), randomly breaking APIs (for example during the ALSA to Pulse audio transition) and kernel ABIs (drivers), and a generally giving off a an “experiment on the customer, they don’t pay us anyway” vibe.
BTW System76 sells Windows laptops from Clevo with BIOS (EFI) tweaks. There is a certain satisfaction when all the drivers in your system were made for your OS instead of being reverse-engineered or community made drivers, or a minimal effort junk by a Windows-centric hardware vendor, which is something that you can only get with Macs.
giving off a an = giving off a
(sorry)
FYI: Dave Patterson wasn’t an architect of MIPS.
MIPS was a spinoff from John Hennessy’s team at Stanford.
Patterson’s team at Berkeley did the RISC-I/II projects from which SPARC was derived.
Still weird to see MIPS ending up doing the Berkeley architecture. Since it started life as the competing project at Stanford. The circle of life, I guess.
Oops. Trying to do things quickly.
I wonder what this means for embedded systems using mips64, I’m thinking my Ubiquiti Edge Routers… This could be either a strong area for RISC-V or they would move to ARM
The writing has been on the wall for MIPS for a while now. RISC-V is very similar, but more refined. This move makes sense, since they can easily adapt the existing mature core designs and re-target them to the more dynamic and growing market for RISC-V cores.
This what I’m thinking too. RISC-V chips need parts which MIPS has already developed. Marketing boost plus entrance into a growing market for MIPS, and a development shortcut for the RISC-V ecosystem.
It’s a smart play.
I remember doing some assembly on embedded MIPS processors. It was by large margin the most pleasant low level experience I’ve ever had. The arch was clear, logical and very easy to grasp in the matter of hours.