Wintel must welcome Androitel and Chromtel into cosy menage – Intel Intel and Microsoft no longer dominate the personal computing industry as the once fearsome Wintel alliance, Intel has acknowledged. Now the chip giant has announced a broad push to get its silicon into devices running Windows' rival operating systems. Intel's PC chief Kirk Skaugen admitted the demise of the ages-old alliance …
This post has been deleted by its author
"They lost the plot when they didn't push itanium properly. "
They pushed IA64 as hard as they could, technologically and commercially.
Still only a tiny subset of the market was interested.
Sensible people buy systems to run software on. People largely bought Integrity servers *despite* IA64, not *because of* IA64. They wanted to stick with the OS they knew and loved. Unfortunately although HP ported the OS, they didn't persuade many application or infrastructure vendors to follow (just look at the HP vs Oracle fuss, where HP have now been shown to be, well let's be honest, LYING about IA64 EOL).
Ditto the Tandem NonStop folk. They want to run their NonStop stuff and the underlying hardware is of little interest as long as it's Good Enough. They got rid of their need for special chip features many years ago. Fortunately for them, they've had a last minute reprieve, their software will be ported onto "yech and spit and horrible amd64" (the one which has, unlike IA64, become "industry standard 64").
Are there specific things about AMD64 you don't like? Or is it just that it showed Intel and HP execs up to be a bunch of incompetent liars and charlatans? How many times did we hear Intel say "an x86-64 is technically impossible, you need IA64" and then out came AMD64 and suddenly Intel had one too.
Intel lost the plot when they acquired Alpha from the wreckage that once was Digital, and buried it as "not invented here". AMD picked up a bunch of talented engineers with lots of Alpha know-how in their heads, produced the AMD64 architecture, forced Intel to follow instead of lead, and almost knocked Intel off its perch. If it hadn't been for an Intel "skunkworks" project that was keeping the original Pentium-3 alive, when the Pentium-4 architecture hit the speed barrier Intel would have been finished.
This battle, which AMD ultimately lost, is probably why Intel didn't spot the threat that ARM and handheld devices posed until it was too late. (I think AMD did spot the threat, but didn't have the corporate strength to respond sufficiently). In another universe, Alpha could have been stripped back to its origins, producing a low-power chip more than the equal of ARM, and with Intel's process technology behind it ....
I still dream of a world where the dominant 64-bit architecture is Intel Alpha, and where x86-32 is ancient history. Intel took one of those wrong turns on which empires totter and fall. I still have the Alpha Architecture handbook to remind me how a really good CPU might have been.
IA64 was killed by closed source software... If you were running all open source code they actually ran quite well, i had a couple of them running linux and all the typical stuff compiled and ran on them just fine.
If Intel were to introduce a new architecture aimed at Android, ChromeOS or Linux it would have a much better chance of succeeding as not only could Intel port these systems themselves instead of relying on someone else, but most of the existing applications would run with little more than a recompile anyway.
Arguably Intel should come out with a new architecture, the legacy baggage of x86 is a millstone around their neck such that even being a step ahead on fabrication tech they are still having trouble competing with arm. If they were to come up with a new architecture designed specifically for power efficient applications they could easily get themselves ahead of arm.
"Arguably Intel should come out with a new architecture, the legacy baggage of x86 is a millstone around their neck such that even being a step ahead on fabrication tech they are still having trouble competing with arm"
Well the question is, can they do this? There is one big advantage of the x86 architecture which kept it alive till now, and that's the PC architecture. It was designed to eliminate porting. You have the BIOS a sorta "minimal operating system" designed to give you a life line to the operating system, so it can go from disk to fully running without having to be ported.
"Are there specific things about AMD64 you don't like? Or is it just that it showed Intel and HP execs up to be a bunch of incompetent liars and charlatans? How many times did we hear Intel say "an x86-64 is technically impossible, you need IA64" and then out came AMD64 and suddenly Intel had one too."
I got the impression from Andy Glew's (working for Intel at the time) posts to comp.arch that he felt that Intel were missing the boat by refusing to extend x86 to 64bits and that he seemed fairly certain it was possible. A few folks tried to draw him on whether he knew of a 64bit x86 skunkworks project in Oregon at the time. Looking back on it, given how quickly Intel shipped a 64bit x86 in the end, there must have been a germ of truth to the skunkworks idea. Itanic generations seemed to take an age to tape out by comparison. :)
"Now the 68xxxx from Motorola. THERE was a processor. Incredibly logical instruction and addressing modes."
You probably wouldn't say that about the 68k if you'd ever seen the NatSemi 32032 family. Don't know if they ever got out in the wild in volume, nothing whatsoever against the 68K family, but the concepts here. Just wow.
OT but yes.
I remember running a 68k system running a Motorola OS up against a Nat Semi 16032, both 16 bit bus, same clock, memory etc. and seeing a 2.5 times performance gain for the 16032 system on our code (which included fixed point arithmetic). Instant sale. But for some reason it never caught on.
This post has been deleted by its author
IBM had a choice between the 8086 and the MC68000 for their desktop system, the Personal Computer. They went with the 8086, well actually the 8-bit bus version, the 8088 for a whole lot of very good reasons. One was that while Intel were delivering the 8086 and 8088 in commercial quantities, Motorola were demoing nearly-functional versions of the MC68000 running at half the rated speed. Another factor was that the 8086 was bus-compatible with 8080 family support chips like the 8271 serial port, the 8259 interrupt controller and the like whereas the 68k chip was going to need a new family of support chips which were still paper exercises at the time. The third, and critical factor was the backwards compatibility in registers and addressing modes to the 8080 which made rewriting existing 8080 code for the new devices a piece of piss. Sure the 68k was a dream to write code for but translating 6800 or 6502 code to 68k was a pain in the arse. Intel delivered code conversion tools along with the new 16-bit chips and the rest is history.
If by "nearly functional" you mean "accidentally run this opcode and soon after the IC will split down the middle with a smell of burning epoxy" - yes, I've seen the remains. I forget the opcode but, as I recall, it turned on one set of gates on the internal bus low, and another set high, and then stuck.
x86 is only beneficial for users who are stuck with a lot of legacy closed-source code...
Linux and other open source is architecture agnostic, we used to run linux on alpha when it was the fastest available, and we run linux on arm or mips now for low power systems.
It was closed source code above all else that killed itanium... Linux runs quite well on it, but windows as a joke - the core os would run but you had virtually no apps and 99% of windows apps dont come with source so you cant recompile them yourself.
For a good while now I've been (anonymously) saying that Intel are only relevant where there's a Windows connection.
I've taken mild flak for it from the Wintel fanbois (and I've been happy to do so).
The surprise here is that Intel are finally admitting the inevitable.
IA64 retired and Windows sidelined (by Intel), both in the same year.
Whatever next? Intel as a licenced ARM design partner again?
The domination of the Wintel desktop is the only thing that's keeping x86 alive. It was a crap architecture from the off, and all the hoops Intel have jumped through to try to improve it would have been better spent by junking it early and coming up with something new. The problem with that was - they came up with Itanium...
That could have been an interesting idea but the execution was terrible.
IMHO it was very sad that DEC came up with one of the best CPU architectures ever (Alpha) just at the point when they were going bust for other reasons. Intel should have bought the IP and the designers from DEC and run with that but by that time they were far too heavily invested in x86 both financially and intellectually.
"Intel should have bought the IP and the designers from DEC and run with that but by that time they were far too heavily invested in x86 both financially and intellectually."
Not to mention that Intel were trying to convince the world that IA64 would become "industry standard 64bit".
DEC's designers knew better, and here's some of what they managed to say about Alpha vs IA64 before they were silenced is in this slightly odd looking (but genuine afaik) 30page white paper [1]. Obviously Intel had deeper pockets, and therefore stood more chance of winning in the marketplace and in the courts.
Some of DEC's chip design folk ended up with AMD, some briefly at Intel working on ARM before Intel decided ARM wasn't their thing (tee hee).
Some of DEC's compiler folk ended up with Intel.
RIP Alpha. Except... there are now more vendors of Alpha emulators than there were of Alpha chips back in the day. HP hasn't quite worked out (or didn't want to acknowledge) that it's that way because some commercially interesting number of people still want to run their VMS setups, despite HP's long term efforts (and Compaq's before them) to ignore VMS.
[1] http://courses.cs.washington.edu/courses/cse548/05wi/files/Alpha-IA64-Comparison.pdf
"So Linux & *BSD don't run on x86?"
Perhaps this needs simplifying for you.
Linux does not need x86 and nor do *BSD. Without x86 they survive quite nicely.
x86 needs Windows. Without Windows, x86 would initially survive for a while, but it would become a niche product rather than the mass market product it has been.
Microsoft's Windows products are distancing themselves from x86. As are HP. Only Dell is still largely Intel-exclusive.
Simple enough for you now?
"IMHO it was very sad that DEC came up with one of the best CPU architectures ever (Alpha) just at the point when they were going bust for other reasons. Intel should have bought the IP and the designers from DEC and run with that but by that time they were far too heavily invested in x86 both financially and intellectually."
Well Intel *did* buy the IP & designers from DEC as part of a settlement of an IP infringement lawsuit, however their aim was to shut up the lawyers and bury one of the other 64bit architectures (Alpha) while they hacked away at IA64 (aka Itanic). At the time Intel planned to keep x86 32bit and relegate it to legacy/low-end stuff while everyone else was migrated to IA64. Thankfully the world was spared from a monopoly shoving an under performing freak-show ISA down it's throat by 'AMD64' and the Opteron. Thank you AMD.
A shortish while ago the Reg published an article claiming that a Chinese uni had built an Alpha ISA chip that yielded some very respectable FLOPS/W figures - I have had a hunt for more info - but I have struggled to find much beyond the initial press release. I think I'll have to learn Mandarin to find out more. :P
Intel did buy the designers and the IP from DEC in 2001. The DEC engineers - who were brilliant - contributed to the last few iterations of the Pentium.
One of my favourite computing stories is the moment when the DEC board realise that their chip guys have taken their top-line seven-figure ECL Monster-VAX and chipped it into something that costs a couple of hundred dollars and runs faster.
Alpha would have been excellent with Intel's fab skills. But that's not what we got.
Still - Intel doing ARM is Not Going To Happen. Not successfully, anyway.
Intel has nothing interesting to offer the ARM community. I suspect too much corporate sclerosis to allow Intel to allow ARM space to use its latest monster multicores - although multicore+optimised OS is probably the only thing that might save a niche for the desktop PC.
Intel has nothing interesting to offer the ARM community.
Not true. Intel's process technology is second to none. If they process-shrank and fabbed ARM chips, they would be the best ARM chips on the planet anywhere in the power - performance envelope.
How else do you think is is that Intel can just about compete in the handheld market with that horrible warty i86 architecture? But I don't think Intel can hold the fort for much longer. Soon, they will be fabbing ARM designs.
Nigel11 wrote:
Intel has nothing interesting to offer the ARM community.Not true. Intel's process technology is second to none. If they process-shrank and fabbed ARM chips, they would be the best ARM chips on the planet anywhere in the power - performance envelope.
But not anywhere in the price/power/performance envelope, that state of the art fab process doesn't come cheap. Having a processor costing thousands of dollars rather than tens of dollars isn't going go down well, even in a top of the range phone or tablet.
If Intel has to reduce margins down to be comparable to other ARM licensees, their lead in fabs wont last long, as they wont be able to fund the next generation.
"There are quite a large number of x86 systems running Linux on the desktop and in the server room, not to mention all the Apple Macs"
Well obviously. PowerEdge and Proliant benefit from being able to offer Windows as well as Linux and BSD and occasionally even (say) QNX. On the other hand, HPQ have been making much noise about ARM kit in the server room. Dell are Intel's poodle; they're going nowhere except down, unless they end their Wintel dependence ASAP. Even their investors know this (and that takes some doing),
Now look at any computer-based kit (ie not just IT department stuff) that never ever has a need to run Windows Mac users like to run Windows sometimes so they don't qualify
Look around you. Is there any Intel kit out there *outside* the Windows market? If there is, I don't see it, from smart TVs to routers to (whatever). Plus a negligible quantity of smartphones.
If it doesn't need Windows, it doesn't need Intel.
Mostly these days it doesn't need Windows.
Times are changing. It's what happens.
"Look around you. Is there any Intel kit out there *outside* the Windows market? If there is, I don't see it, from smart TVs to routers to (whatever). "
Ummm, there's this computer company called Apple, they sell a lot of laptops and desktops and they all have Intel chips inside and they don't run Windows (unless the user wants to). In fact it was a big thing a few years back when Apple gave up on the vastly superior PowerPC architecture and switched to braindead Intel x86 chips for some crazy reason; they swallowed the Megahertz Myth koolaid, after all PowerPC had Altivec, win win! That crazy Steve Jobs guy, what was he thinking!
If it doesn't need Windows, it doesn't need Intel.
Hmmm. How else can I buy a really fast server for crunching numbers or big data using Linux? Say 4 CPUs, 32 Cores, and half a Terabyte(*) of RAM? Sun's gone. HP IA64 systems don't make financial sense unless you want to run VMS. Much the same for IBM Power-based systems. AMD have sadly fallen behind again, their Opteron glory days are behind them. GPGPU computing has its place, but so far, it's a fairly restricted subset of scientific programming for which a GPGPU is the answer. Don't think I've missed anyone.
(*) There's such a machine crunching away a hundred yards or so from where I'm sitting. It's attacking large sparse matrix problems. If you know what that means you'll know why it needs that much RAM.
"How else can I buy a really fast server for crunching numbers or big data using Linux?"
Sadly I agree, We moved from SG workstations to Xeon workstations, gained performance, future proofing and all our specialist scientific software could be readily obtained for Linux, further we then assembled 512, then 1024 then 2048 node commodity-based Linux X86-64 clusters. This all saved a fortune and still allowed us to mix in IBM fileservers & compute servers.
"If it doesn't need Windows, it doesn't need Intel."
With hindsight I should have prefaced with "In the volume market,". HPTC, while fascinating, and clearly important to a variety of folks, isn't really volume market.
"How else can I buy a really fast server for crunching numbers or big data using Linux? [snip] It's attacking large sparse matrix problems. "
Do all roads lead to x86-64 Proliant and SuperMicro etc at the moment? I've lost touch with that kind of hardware.
Anyway it's a very fair question, and yes I did once understand what you mean. It was me whinging the other day about Kuck + Associates Preprocessor (the source to source optimiser for C) having dropped off the market since Intel bought them (if that means anything).
I also remember DXML on Alpha, (again if that means anything - [1] has the details, nothing to do with XML).
It'll be interesting to see where that market goes once ARM starts eating into the volume x86 market. It may damage some of the chip and system level benefits which high end x86 has historically enjoyed from x86 being the volume market leader too. Good job you're on Linux.
"I still have the Alpha Architecture handbook to remind me how a really good CPU might have been."
I assume you mean the relatively lightweight (maybe half inch thick, A5ish) one which was also freely downloadable ages ago ?
The Alpha Architecture Reference Manual is the serious geek's version of the Handbook. The Reference Manual was (is?) a commercially published book, but if you're interested, the 4th edition seems to be freely downloadable online [2], marvellous if that's your thing (published 2002, almost 1000 pages, 8MB scanned+OCR'd PDF).
Enjoy. And have a lot of fun.
[1] http://www.hpl.hp.com/hpjournal/dtj/vol6num3/vol6num3art4.pdf
[2] https://archive.org/details/dec-alpha_arch_ref
WinDOS legacy software allowed x86 to outlast all of it's other rivals. While it sucks when compared to Alpha or PPC, it really isn't that bad when compared to SPARC and that was the leader of the alternatives. Sun was like the Microsoft of the Unix world. They were the big player but they were mediocre themselves.
Now we're down to ARM vs x86 and ARM just doesn't have the horspower. It's nice for low power appliances but still kind of sucks for computation.
I'd much rather see a revial of SPARC then people try to shoehorn ARM into a compute bound server role.
"ARM just doesn't have the horspower."
Have you actually looked at what's in the latest ARM architectures and chips? It may not be what you expect.
There's a similar relationship between today's/tomorrow's high end ARM and yesterday's ARM as there is between an AMD64 and an early 8086. In other words, not that much in common, but a certain amount of familiarity if you feel you need it.
The latest ARM incarnation has a new operating mode (think AMD64 mode vs legacy x86 mode) which doubles the number of registers (often good for performance), drops predicated instructions (good for performance in a system where code density isn't critical) and so on.
Have a read. You might be pleasantly surprised.
> Look at how much they nobbled the atom platform (Netbooks limited to 2GB RAM) just because they were afraid of their low price experiment cutting into the desktop share.
This will continue to be their problem. The question is whether ARM will nobble them or whether they will nobble themselves. The two big phone makers are vertically integrated. Why would they want to hand intel a slice of profit? It's as much a commercial issue as a technical one - even if Intel became better, they will not accept the miniscule margins ARM holdings run on, so they will be more expensive.
... Look at how much they nobbled the atom platform (Netbooks limited to 2GB RAM)
No, it was Microsoft who nobbled the netbook platform by not allowing Windows 7 Starter Edition (the cheap netbook version) to be licenced for machines capable of supporting more RAM (or having a larger display than 1024x600, etc). Manufacturers wanted the cheap licence for the cheap machines , so they nobbled the support hardware to qualify.
It's true that Atom chips, although they do support a full 4GB of address space, are limited by having rather rudementary memory controllers; Intel had to do that to keep the transistor count and so the power consumption low. I wouldn't call that "nobbling" because there's a technical reason for it (albeit one that could have been avoided in a better-designed chip).
It is quite normal that nimble young companies innovate and big old companies follow. But in computers, Wintel managed to lock the market down for decades and destroyed the competition before they could get a foothold. Somehow they missed the threat from really small computers (call them smartphones if you wish) and now there are quite some competitors in software and hardware. Even better, it became obvious that a manageable investment is sufficient to enter the strongly commoditized market.
Expect the new normal with several healthy competitors on the hardware and software side and many small companies trying to get a foothold (witness all those Chinese hardware manufacturers and the flavors of linux-based OSs).
Classic ecological succession...
The small company grew big, then enormous filling it's ecosystem, and making it a habitat specialist with dominance trouncing flexibility. This gives it a huge tail of inertia hanging of it's arse end in the form of legacy hardware,software, customers and business model.
A form of climate\habitat change ocurred in the users world with technical developments allowing practical, useful small devices to be manufactured and wham a whole new user habitat\niche appears.
Genuine marketing guys (rather than salesmen) would see this as an unfulfilled consumer desire to be met by manufacturers. This desire might even be invisible to the customer, Alan Kays DynaBook is 40 yo but it took Apple to make it cool and by selling iPads in bulk kickstart the tablet business.
The Big guy often doesn't have the tech to invade this new niche without making a complete break with it's past and leaving their customer base behind (Windows 8 anyone). So initially they ignore it after all in it's early stages, it's very small and not very useful. Moores law kicks in and wham again, the new habitat becomes the new normal bye bye big guys.
Just like bio-evolution changing habitats means you have to start pretty much from scratch plus you're competing with the new locals in their new normal..
I'm failing to find the article in your archive but several (many?) years ago El Reg mooted that Intel unleash its massive software expertise and deliver itself from the Wintel partnership. Certainly earlier than this 2009 article on Intel and Chrome
The primary reasons that Intel are so uncompetitive in the mobile space isn't because they lack the technological capability. It's essentially a commercial issue. Intel is simply far too big to be sustained by a mobile market which has grown on the back of the incredibly cost-effective ARM eco-system. ARM has a small fraction of the market capitalisation and cost-base of Intel by charging relatively small royalities for use of its technology, and this has driven massive growth. Quite simply, even if Intel did manage to duplicate the ARM eco-system, with all its third party support and flexibility, it would only garner a fraction of the income it derives from the x86 market.
There is a certain irony here. Intel managed to do much the same thing to other processor architectures by offering a more cost-effective option and leaving any survivors with what are niche markets. Intel itself may come to be dependent on just such a niche, albeit lucrative market. The mass market for processors embedded into virtually everything is structurally incapable of supporting a market with the sort of margins that Intel became used to.
The lesson is, once you lock in a high cost-structure company, you are always vulnerable to those that are not. It's no doubt something that the folk at ARM are well aware of.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
