Is it time to stop research in Computer Architecture ?
|
From: nmm1 on 23 Oct 2009 10:25 In article <8o6dnXS7k_mTJnzXnZ2dnUVZ_g1i4p2d(a)bestweb.net>, Mayan Moudgill <mayan(a)bestweb.net> wrote: >> >> No way. Sorry. HP had restarted the PA-RISC line by then, IBM had >> restarted POWER, MIPS was available for purchase (and there were >> many companies with the resources to take it over and restart it), >> SPARC was still very much alive, and doubtless there were other plots >> and plans. > >Don't know about the other companies, but IBM had never ceased POWER >development; it just seemed that way to the outside world :) Yes - I was over-simplifying. They reinvigorated it, and restored its target markets that had been assigned to IA64. >And SGI was catastrophically late. I don't know what the state of their >processor development team was at that point; did they even have anyone >left? No. But MIPS wasn't just SGI, and SGI had already spun off the most critical parts of the technology so that it could have been taken over very easily. SGI couldn't have done it, though. Regards, Nick Maclaren.
From: Bill Todd on 23 Oct 2009 12:25 nmm1(a)cam.ac.uk wrote: .... > The association of VMS and $4 billion with "world domination" in > the late 1990s is amusing :-) You missed the point, as usual. The point was not that VMS per se was a leading player, but that its customers could have made a significant difference to Itanic's perceived acceptance and minuscule market share at a point in time where that difference could have lent credence to the continuing hype machine (a machine which had succeeded sufficiently with such a minimal substantive basis that it's disturbing to think how well it might have succeeded with any more). .... >>> >>> HPC was its most successful area, and something like two sites tried >>> it and rejected it for every one that delivered a service using it. >> The importance of HPC to real-world success of a platform is pretty >> debatable. > > Even in the late 1990s, it had more influence than VMS. Same missed point as above (and you can add the fact that VMS users were only a part of the grassroots reaction against Itanic: as I made clear, I brought them up as examples because I happen to be well-acquainted with the details of how things went there). .... >>> My point here is that, if the Itanic had started to be pushed much >>> harder, the real heavyweights would have joined the opposition. >> What earthly reason do you have to think that? IBM never lifted a >> finger against Itanic but rather got on board early and shipped product >> for several years. ... > > I was informally contacted by one important group, and had contacts > with others. It wasn't quite like that .... It was *exactly* like that: your purported refutation below refers not to IBM but to some of its customers. > > My comment was based on gleanings of throw-away remarks from a large > number of people in very important customers. A few explicitly said > "over my dead body", but one hell of a lot hinted it, often by being > quite open about everything else but clamming up about the Itanic. > This was when they were using publicly available systems, so the > draconian pre-release NDAs did not apply. Virtually none were in > favour. No one in their right mind would have had any use for Merced, which is what those early shipping systems were. Nor would anyone have been that enthused about McKinley, which qualified as 'respectable' for some not entirely unreasonable uses of the term but was hardly the world-beater that one would like to have as an incentive to switch platforms. > >>> It never had an earthly of doing what it was originally hyped to >>> do (i.e. entirely replace x86). >> Save for the grace of AMD it still might have: without a credible, >> inexpensive, and pervasive 64-bit alternative Intel could have just >> waited until desktops began to demand 64-bit processors. > > No way. Sorry. HP had restarted the PA-RISC line by then, IBM had > restarted POWER, MIPS was available for purchase (and there were > many companies with the resources to take it over and restart it), > SPARC was still very much alive, and doubtless there were other plots > and plans. And exactly what does any of that have to do with replacing x86? IBM hadn't 'restarted' POWER: unlike SGI, HP, and Compaq, they'd never stopped development in the first place. The main difference between IBM's and Sun's attitude was that IBM never publicly repudiated Itanic as Sun did (and the value of Sun's repudiation was insignificant compared to what it would have been coming from IBM). The only reason HP restarted PA-RISC was because Itanic was so disastrously late: they had no intention whatsoever of providing any more than an end-of-life kicker, which is exactly what they did. - bill
From: Gavin Scott on 23 Oct 2009 14:11 "Andy \"Krazy\" Glew" <ag-news(a)patten-glew.net> wrote: > I am not aware of an Itanium shipped or proposed that had an "x86 core > on the side". I'm pretty certain that Merced had hard support for x86 and McKinley may well have as well. In fact (IIRC) Merced reset into x86 mode and was fully prepared to boot 16-bit DOS or Windows 3.1 until you came along and explained to it that you had something else in mind. The printed Intel IA-64 Instruction Set manual consists of only about 1/3 IA-64 information with the bulk of the hefty tome being all x86 stuff. You can spot an IA-64 programmer because they have this odd torn-in-half manual with no back cover on their desk. Both of these I'm sure are the result of Intel marketing refusing to believe that customers would accept a replacement for x86 in which x86 execution was not a first class citizen. I'm certain HP came into it with the idea of using dynamic translation for x86 compatibility (since they had such good luck doing this when PA-RISC came out) but that Intel marketing demanded the hardware implementation due to fears about the perception of an "emulation" solution. Of course the x86 hardware was hopelessly out of date and slow by the time the things actually shipped, and Intel later did switch to a dynamic translation solution which allowed them to drop the x86 core finally. I am curious how much the integration of the x86 stuff (and its testing) added to the delays. Of course writing the software translator/emulator might not have been faster. For PA-RISC capability HP had very high hopes for dynamic translation. One slide from fairly early on suggests they expected to get to 50% of native performance using translation. In reality they failed to scrounge up enough cleverness to do it well, and the PA-RISC compatibility on IPF has always been poor enough that the performance is commonly considered unacceptable even for business applications. They also failed to provide any mixed-mode support, so conversion from PA-RISC to Itanium is an all-or-nothing proposition per app. The Itanium project had three legs it needed to stand on. A better architectural and programming model for taking advantage of the hardware, time to market, and lack of competition, and in the end it failed on all accounts. Had any of those things turned out differently we might be phasing out the last of our x86 systems by now. Of course in the end Intel has still ended up with the dominant (x86) architecture, their competition is clinging to life rather tenuously, several competing architectures got eliminated, and HP ended up with their wide word replacement for PA-RISC that they had wanted, so I'm not quite sure the word "failure" is particularly applicable. G.
From: Del Cecchi on 23 Oct 2009 16:17 "Bernd Paysan" <bernd.paysan(a)gmx.de> wrote in message news:7757571.DSDyUQ213Y(a)elfi.zetex.de... > Andy "Krazy" Glew wrote: >> OK, OK, OK. This is not my area. But I would love to understand >> WHY >> something like this cannot work. > > The problem with CMOS is that all the transistors have embedded > diodes, that > need to be reverse biased to make them operable. A transistor > really is a > four terminal device (source, drain, gate, *and* bulk). > > This sort of low-power reversible computation stuff is more for > nano-scale > electronics (using carbon nanotubes and whatever science > fiction-like things > you can imagine ;-) than for microelectronics. > > -- > Bernd Paysan > "If you want it done right, you have to do it yourself" > http://www.jwdt.com/~paysan/ You could use SOI, no bulk. :-) I don't get the point of the AC. Light bulbs and space heaters are AC powered and still disipate power. What did I miss?
From: Bill Todd on 23 Oct 2009 17:25
Gavin Scott wrote: > "Andy \"Krazy\" Glew" <ag-news(a)patten-glew.net> wrote: >> I am not aware of an Itanium shipped or proposed that had an "x86 core >> on the side". > > I'm pretty certain that Merced had hard support for x86 and McKinley > may well have as well. Hardware support for x86 on Itanic existed on Merced, McKinley, Madison, and Madison II (the one with the 9 MB on-chip cache), finally disappearing in 2006 with Montecito. .... > Of course in the end Intel has still ended up with the dominant (x86) > architecture, their competition is clinging to life rather tenuously, > several competing architectures got eliminated, and HP ended up > with their wide word replacement for PA-RISC that they had wanted, > so I'm not quite sure the word "failure" is particularly applicable. It depends upon how you define failure, I guess. Itanic failed - abysmally - to meet its original performance targets, even after also failing by many years to meet its schedule targets. Itanic certainly failed to meet Intel's profit goals, draining billions of dollars from the company over a period of more than a decade before even starting to break even on the margin (and it will be interesting to see whether those sunk costs will *ever* be repaid, even if you ignore accumulated interest and opportunity cost). Concentrating on Itanic and assiduously keeping x86 from competing in the 64-bit space (until AMD forced the issue) also cost Intel significant market share (and thus yet more profit) in x86, though it has since clawed some of it back. Itanic failed to meet Intel's goal of supplanting x86 and thus freeing Intel from the cross-licensing obligations which it had acquired in the past relating to the lucrative market segment dominated by x86. True, Intel did manage to fight back from its neglect of this market after AMD scared its pants off - but dominating that market still isn't as lucrative as having the complete monopoly over it which migrating it to Itanic could have cemented. Itanic failed to dominate the enterprise-level market segment that it was originally touted to take over lock, stock, and barrel. This may have been more an HP goal than an Intel one, since HP was in a position to leverage its co-ownership of Itanic in the integrated systems in that market while Intel would just have been a processor vendor in that area. As a result, HP is moving toward being more of a box-shifter (mostly x86 boxes) in the server market - though given that their original goal was to get out of the processor biz, it would not be entirely surprising if they're OK with drifting out of the integrated system biz as well given that their plans for market domination came up empty (actually costing them share rather than gaining it in all areas save the x86 servers). So since HP didn't sink anything like the money into Itanic that Intel did, and may not be all that unhappy with the way things turned out, you could say that it came out smelling almost like a rose to its investors (having taken a shot at world domination without having had to put up much collateral at all), even if less so to some of its customers and others who were more badly burned by this misadventure. Still, Itanic eventually became a usable (if hardly inspiring) product, so by that particular measure I suppose calling it a 'failure' might be excessive. - bill |