Slashdot Mirror
Itanium Problems
webdev writes "An article in today's NYTimes (free but...) highlights some industry concerns over Itanium. The author suggests the normal "what's bad for Intel is bad for the computer industry". Anyone know the power consumption for IBM's 64 bit effort GPUL?"
I'd venture to say that IBM's processor uses little more power than other PowerPC CPUs. Doesn't it sport SOI and other technologies to limit heat production? Heck, for an--albeit moderately poor--example of this is IBMs 750FX processor vs. the P4. At the same clock speed, the 750FX would consume roughly one fourth the power of the P4.
The Political Programmer
Because this thing really should be called the Itanic...
"It has taken an entire decade, an estimated $5 billion and teams of hundreds of engineers from the two companies to bring the first Itanium chip to market. As the struggles and costs mount for the companies, skeptical technologists say Itanium now has the hallmarks of a bloated project in deep trouble. It is already four years behind schedule, emerging just as companies are in no mood to spend money on technology"
Skeptical? More like, forget it Chachi, it ain't happening.
I guess the larger companies don't get it. Corporations are struggling. Companies are in holding patterns, waiting for the mess, erm, economy, to level off.
Can I have a job now making millions being a skeptical technologist?
Sent from your iPad.
...friend!
I'm part of a team of people working on a largish supercomputer using itanium2. The things are fast fast fast. Much faster than i anticipated. it's special purpose I think, which is why it defies industry logic
-- Who is the bigger fool? The fool or the fool who follows him? --
I just read a story on msnbc.com about AMD's 64bit processor, I close the window, check slashdot and there is the story about Intels Itanium. Anyway here is the link for msnbc. http://www.msnbc.com/news/813950.asp?0si=-
In college, really poor, need a flatscreen.
"Every big computing disaster has come from taking too many ideas and putting them in one place, and the Itanium is exactly that," said Gordon Bell, a veteran computer designer and a Microsoft researcher."
He should follow that up by saying, "Here at Microsoft we have proved this time and time again."
http://www.hp.com/products1/itanium/performance/ar chitecture/speccpu.html
-Kevin
I submitted this a couple weeks ago, but I guess it didn't make the grade:
AMD's x86-64 architecture will allow companies to upgrade individual parts of their software systems to 64-bit without having to replace everything else. That's the key to AMD's future success; it makes the migration path to 64-bit that much easier (and that much cheaper).
Itanium flopped before; chances are good it will flop again.
Non-Reg Link Thanks to Google News.
Slashdot gets worse every day... Pipedot: News for nerds, without the corporate slant
The only problem with the Itanium 2 is that Intel is only offering it in a high end configuration with lots of cache. The chip itself when you normalize for cache costs about as much as the P4. GCC already supports the Itanium and Intel has great code they could give to GCC in terms of optomization (Intel doesn't make money in the compiler business). Apple is looking for a new chip and IBM doesn't work out this is a great place to go. Grabbing Linux, BSD and Apple will put tremendous pressure on Microsoft.
The article itself doesn't mention any problem with the chip other than electricity usage and heat which are both a product of the large amount cache on the current configuration.
Only if you try to overclock it.
"They do not preach that their god will rouse them, a little before the Nuts work loose." Kipling, 'The Sons of Martha'
A freaking 130 Amp Chip?
Even with 220 million transistors in it that is a lot of power. Intel should consider that big companies and small users dint always want the BEST of the BEST, they want something that is cost effective. As the story mentions Google might prefer to use a lower power chip because they could save millions in power costs. This can apply to small users too as that chip alone could cost you up to $100 a month.
Think on the bright side, during the winter when you are on doom 3 you are also heating the house!
Medevo
I don't think most people use datacentres to play Doom, you know...?
The Itanium is not meant to be a desktop chip. The problem is, it can't seem to cut it as server chip either (too expensive, too power-hungry).
You say there's no demand for 64-bit chips? I wonder why Sun and IBM are still in business, then...
RMN
~~~
He's absolutely correct. The most intelligent thing to do is to make insignificant, incremental changes, and charge customers full price for each of them.
Jesus was all right but his disciples were thick and ordinary. -John Lennon
I'd don't know about the common man, but we're
already running x86 linux boxes with 3.5G of
RAM, any more than that and you need 64-bit.
Google for example uses commodity x86 boxes,
and keep there whole internet index in RAM, for
that cheap, big memory, 64-bit
boxes would really come in useful.
We've ported chemistry simulation code to the pre-release ITA-2, and run benchmarks. There's not much like it, performance-wise, and on a cycle/dollar scale, it's in a class by itself. Smokes US-IIIs, walks away from the Alpha, and keeps pace handily with the Power4, at a more academicly-tolerable price. It's a good chip in its second incarnation, and has the misfortune to be introduced during a recession.
As always, the NYT ignored that you'll need the 64-bit address space for large applications, it has excellent memory bandwidth, and those customers requiring such a system weren't explicitly interviewed or mentioned. The heat issue is true, and that's it's one failing, but as with the Alpha, it will get better in time. (I still remember the rumors, pre-release of the Alpha that DEC was going to have to build a liquid-cooled workstation)
The emergence of the 64bit chip market is pretty exciting, even to an ignoramus like me, but this article got me thinking about some things. The whole power consumption issue is really undervalued I think. We've gotten to the point that most chips are fast and powerful (strength) enough to do tasks efficiently. But I've heard that specialized chips are more efficient at lower clock speeds and power consumption but suffer from their rigor and restriction to a certain type of processing. Maybe its time to give specialized chips their due and move flexibility off the chip itself and into multi-proc (using different specialized chips) or even multi-machine situations.
Of course faster is always better in database mining and protein folding and nuclear explosion modeling, but I wonder if the field isn't ripe for a move away from generalized powerhouse chips to more specialized chips that run at lower clock speeds (perhaps) and have lower power consumption (a must). Personal computing made advances due to cheap general use chips, but as our computers become specialized appliances, a move towards specializing the insides makes sense to me.
Itanium seems to me to be too late to the party. Its an old school chip and probably/ perhaps a badassed one at that. But computer users, from desktop to database, are likely to appreciate specialized chips in multiprocessor or multimachine configurations that express the flexibility. I don't know if its possible, but on the desktop side, rather than have a 3 Ghz general chip, maybe two cheaper and less power hungry 2 Ghz chips each with a unique specialization for certain types of tasks might perform better. One chip to rule them all is so last century.
Regardless of the feasibility of what I've said, lower power consumption is really cool (no pun intended, honestly). Just because it doesn't have an exhaust pipe port doesn't mean that the computer doesn't pollute.
Etc, etc, ad nauseam, and so on and so forth.
...can you imagine a beowulf cluster of these?
In fact, I know from a reliable source that tomorrow the president of the USA is going to reveal that the Iraqi army has managed to get hold of 2000 Itanium chips and is threatening to turn them all on and melt the Earth.
RMN
~~~
Heaven knows they have a copy of MS's book on corporate behaviour when it comes to competitors.
Bad for Intel probably means good for the industry, as we won't have another half-assed chip shoved down our throats.
* Winners compare their achievements to their goals, losers compare theirs to that of others.
using as many 10 watt bulbs as that thing would light...
Seriously, is efficency no longer even considered?
I live in a giant bucket.
There's something even above compatibility (migration path) - namely Moore's law. The goal #1 goal of a CPU company is staying on Moore's curve. Now the problem with x86 is that it is a f*cked up instruction set architecture, and because of its monstruosities (8 registers ? stack-based FP ?) it has become a major hurdle in staying on Moore's curve. Good luck to AMD with their 64 bit thing ... I seriously doubt that their 64 bit chip will be any faster than their own Athlon (going from 16 to 32 bits registers is a big deal, from 32 to 64 not so much)
The Raven
The Raven
We have an early model of the Itanium ( given to us free by HP ;-).
The beast has a 220V power line coming into it, and we've decided that the reason its so heavy is that if it was lighter, the fans would propel it across the room like a jet engine.
I think he's very right. Take for instance SMP. A single threaded application running on an SMP system has no advantage over the same app running on a single processor system.
In the same way, most applications aren't even aware of 64 bits. So they will continue adding, multiplying, and addressing memory in 32 bits -- whether they be binary ports, or actually recompiled versions.
For the lazy man's migration path of using the same apps on a 64 bit system, there will be no advantage whatsoever of using a 64 bit system.
On the other hand, if you are recompiling, you might as well switch to the EPIC instruction set (Itanium), and get a defacto performance boost -- even if you don't port the code to be 64 bit aware... that's something you won't get even if you recompile for 64 bit CISC opteron.
And last, if you are refactoring, or re-designing your app for 64 bits, there is no migration path per se.
So I think it all boils down to: power consumption (for google), marketing strategy (ie. hyping strategy), and economy.
As an ex-Itanic designer, I can't help but get a warm fuzzy feeling every time I read bad news about Itanic. I sat there for years and watched upper-middle management screw over the project (and each other) in order to advance their careers. The only escape (especially after they froze internal job transfers) most of us grunts got was a job at a new company.
I went into Merced with all the hope and excitement of a new engineer. I left hating the profession and the management that controls it.
Regardless of how much Intel stock makes up my portfolio, I hope Itanic crashes and burns. I hope Yamhill (64-bit x86, designed in Oregon) succeeds flawlessly. I am way too cynical to believe it'll happen but, I hope the success of Yamhill forces Barrett to realize the uselessness of Santa Clara design, causing him to shut it down and rely on Oregon design to do it right. But, considering that Gary Thomas was "punished" for his failures on Itanic by being given a ton of options and a cushy job in Intel-Folsom, Itanic and Santa Clara "mis-design" will just continue along.
Of course, I am just a bitter old engineer taking cheap shots.
Long live Itanic, Intel's Verdun!
The Itanium relies heavily on exceedingly good compilers that will perform for the IA64 the same level of optimization that regular, on-the-fly predictive optimization do in RISC chips.
The main obstacle with this method is that Turing's theorem says static compile-time optimization will never work as well as dynamic optimization. This is because, roughly, the only way to guess what a program will do with a given set of input data is to execute it with its actual data set. Here is a link where a reader of The Register addressed this concern in 1999.
Is anyone aware of how well the limits predicting by Turing can apply to the compile-time IA64 algorithms?
--
Mad science! Robots! Underwear! Cute girls! Full comic online! http://www.girlgeniusonline.com/
NYT contrasted I2 with AMD's upcoming 64 bit offering quite prominently.
The number one value of a 64 bit CPU, to my mind, is the ability for it to address more than 4G of RAM, without destroying locality, like the PAE does on 32 bit processors.
PAE, for those of you who are, as yet, unaware of it, allows you to access more than 4G of physical RAM, by reviving an old technique called "bank selection". It's fairly useless for most of the applicaitons for which you would want more RAM in the first place, since it doesn't increase the allowable size of the kernel or process virtual address space at all, so the only thing it lets ou do is use RAM instead of swap, and not run lots of applications at the same time, without a lot of VM changes.
Intel keeps trying to sell us Itanium on performance, when, in fact, we don't care. What we care about is the ability to operate on larger data sets.
Intel: just because your delivery of access to larger amounts of physical RAM on 32 bit processors, via the PAE, was not welcomed (mostly because it was implemented in a way that was totally useless to software engineers and OS designers), doesn't mean that access to more RAM *by a single kernel or process* will not be the major selling point for Itanium: it will.
Get your crap together, and quit concentrating on clock rates.
-- Terry
I'm glad to see that Macs finally support multitasking.
The nytimes needs google *much* than google needs the nytimes. Without the nyt - google *still* has thousands of news sources - without google, the nyt looses probably 20 to 30% of the page views they would get otherwise.
Besides, all that is being "subverted" is the moronic registration process, something that the nyt willingly gives up for google news readers
Personally its not God I dislike, its his fan club I cant stand (bash.org)
GPUL the stupidest name for a processor? What were they thinking?
A deep unwavering belief is a sure sign you're missing something...
...and because of its monstruosities (8 registers ? stack-based FP ?) it has become a major hurdle in staying on Moore's curve...
Could have fooled me. It seems like just yesterday that MIPS said they would change the world. Not buying it, this time around.
C//
Allegedly large data centers such as Google are sensitive to power consumption. Of course we are not just taking about the power consumption of the processor. We are also taking about the power needed to keep the boxes cool as well as the power that is needed run the air conditioner to cool the data room at about a 20% efficiency. What this means is that several watts of energy must be used to cool each watt used by the computer equipment.
I agree that Itanium may have misjudged the market for this chip. If AMD can produce a chip that is almost as good, but much more efficient, it may well be more economical to buy three AMD based machines instead of two Intel based machines. This becomes even more possible as a box becomes a single disposable commodity component in a very large networked array. Much like the auto industry, it may be practical to build inefficient cars when energy prices are low, but it is nevertheless a risky venture.
"She's a scientist and a lesbian. She's not going to let it slide." Orphan Black
Now the problem with x86 is that it is a f*cked up instruction set architecture, and because of its monstruosities (8 registers ? stack-based FP ?) it has become a major hurdle in staying on Moore's curve.
Huh, that's really interesting. I'd say Intel and AMD have been doing a pretty good job. If what you say is true how come we aren't all running RISC computers now? Well, in a way we are. Today's AMD and Intel chips are not truly CISC anymore. Might wanta read up on the features of CISC and RISC and then read the specs on a K7 or P4.
There's one thing Inever understood about Intel's and AMD's design for 64-bits CPUs. Intel seems to aim for simplicity, that is, 64-bits code should be clean, as compared to current x86 code. AMD, on the other hand, seems to be mainly concerned about downward compatibility (which is a huge win). But why not have it both ways? The CPU could just start out in 16-bit stone age legacy mode, and then be switched to 64-bits mode, similar to how today's x86en are switched to 32-bits mode. The 64-bits code could then be clean like Intel proposes, and we'd all be happy. Of course, it would effectively mean having two CPUs on one chip, one for legacy code, and one for modern code, but isn't that what's happening anyway? Last thing I want to say: clean 64-bits code makes me think MIPS.
Please correct me if I got my facts wrong.
Saddest sentence in the whole article:
"There are other benefits for Hewlett-Packard. The Itanium allows the company to eliminate both of its current 64-bit chips -- the H.P. PA-RISC and Compaq Alpha. That alone should save the company $200 million to $400 million annually in development and manufacturing costs, according to Steven M. Milunovich, an analyst at Merrill Lynch."
Yeah, HP and Compaq have been fine stewards of their engineering legacy...
"How to Do Nothing," kids activities, back in print!
Nerd: Derogatory term typically directed at anybody with a lower Slashdot ID than you.
I'd give Intel engineers just a bit more credit than the average /. poster. Intel has been right at getting the trends for awhile now. Take the Pentium 4 for example. Everyone thought it would flop cuz it had crappy IPC. It sucked in the first several iterations (less than 2 GHz). But its quite the speed demon now, ain't it?
As for Itanium, there are quite a few ways it could succeed. It has the potential for serious performance. The super-wide architecture is perfect for code like scientific processing, image processing, and 3D graphics that are nice, regular, and easy to optimize and parallize. And what kind of processing do you think is going to be popular in the future?
A deep unwavering belief is a sure sign you're missing something...
What's bad for Intel is bad for the computer industry? Intel may have their fingers in a lot of things, but if Intel (and for that matter MS) disappeared tomorrow, the computer industry would survive. AMD would love that, I'm sure... they would not only be the de facto standard on x86-64, but on x86, in general. And hopefully AMD would hurry up and release a mobile Duron or XP with really low power consumption, enough to be put in a PDA along with plenty of AMD's flash memory too (come on, ya know many of you would love an x86 PDA that you could run windows, freebsd, linux, etc. on with minimal changes)...
And of course, Apple would love that too, hehe
IA64 will have the edge for about 6 months. After that Power4 (next rev) will leap over IA64 with a minimum of disruption because it is already 64 bit.
Then Intel will go back to their day job of manufacturing chips in incremental 25% improvements. Intel will reach the limits of power consumption before they reach the manufacturing tolerance limit.
google for itanic, and you'll begin to see why.
the continuing campaign is just throwing good
money after bad. now is AMD's time to shine.
i'm considering doing my next project closed source
just so that i can release it exclusively as
opteron-only, because i love being right.
-I like my women like I like my tea: green-
Yes, but the radical change was in the 20 years from 1960 to 1980. That was when computing made the leap from punchcards and discrete logic to ICs and magnetic storage. The last 20 years have been spent making incremental improvments to the same x86 architecture.
0 1 - just my two bits
Dynamic optimization is not restricted to hardware. Java Hotspot will do well with Itanium (if Sun survives), and I believe Smalltalk and LISP have dynamic optimization as well. The way I see it, Virtual Machines are the future of high performance computing. And yes, .NET is important for Microsoft to prosper in the non-IA32 world. (Although I hate it when the wicked prosper.)
I always wondered why Sun was putting a great deal of emphasis on power consumption on their new line of processors. In retrospect, I see why. Smaller blade servers, which allows you to pack a lot of servers into a small space. And power consumption, which if it is very high, eats into the TCO. Oddly enough, it looks like the SPARCs may be playing the game better than you'd think.
They should expect help from Intel because open source GCC users aren't tied to Intel in any meaningful way. Offering better GCC performance effectively ups the performance of Intel processors for open source software. Especially since open source software is being used more and more for processor benchmarks this is useful for Intel. Sales of ICC are worth almost nothing to Intel.
Just as a quick point Compaq offered Gem for free to Linus about 5 years ago. They wouldn't agree to licensing terms so Gem didn't become system compiler but Compaq's willingness to give away a crown jewel to woo the Linux crowd proves I'm not entirely out of line.
While you may think that GCC should not expect anything from Intel, Intel disagrees; Intel has provided documentation as well as money for Red Hat (and Cygnus before them) to get free software to run decently on their hardware. AMD has done the same, it is simply good business.
GCC is a portable compiler; ia64 is a radically new architecture that needs special treatment from compilers. It will take time to get things working well, and problems with compilers may be the factor that makes AMD win in the long run over Intel. If the ia64 is theoretically faster, but compilers generate better code for the less radical AMD 64-bit processor, AMD wins the performance battle. If you have to buy a compiler from Intel to get the same performance you get with AMD with the free compiler, same deal. For that reason, Intel will have a strong financial motivation to help GCC do better, even if this cuts into their compiler business.
>just makes you wonder if we'll still be using x86 compatible chips in the year 3029..
Yup, and IPv4, and people will still not buy a PC without a 1.44MB floppy drive, despite the fact that the last floppy disc was finaly destroyed in 2589...
Advanced users are users too!
Yes, but what about AMD users?
Perhaps. But I don't expect them to give as much information to make GCC >= icc. That would be self-defeating from Intel's viewpoint.
Usually because of VM mapping software is limited to 2GB. With Windows you can up that to 3GB by buying Enterprise Edition. Mind you there are already "high memory" style hacks (just like the good old DOS days with EMM, etc) to access extra memory in servers.
Nerd: Derogatory term typically directed at anybody with a lower Slashdot ID than you.
code from Intel's compiler frequently performs very well on the Athlon..
there's one test that Tech-report are fond of (sphinx speech recognition) that's faster on the P4 using a Microsoft compiler, and faster on the Athlon with Intels
The P4 was an initial disaster for Intel--the cpu hardly anybody wanted. But it wasn't just because of its low performance and IPC, it was because of its dependence on Rdram in the beginning. A mistake which Intel has since remedied.
/. posts from that time!
>>>>>>>>>
Um, no. The P4 was initially aimed at the high-performance market, to whom RDRAM's cost really wasn't that much of an issue. The real problem was that even with RDRAM, The P4 was slower than a cheaper Athlon. The RDRAM factor is arguable (given that RDRAM is still the fastest memory for the P4) but the P4 really took off when they jacked up the clock-speed and overtook AMD.
The fact is that for the past three years Intel has done a lot more wrong than right, stretching all the way back to the infamous re-called 1.13GHz P3--it's the first time in my memory that a shipping cpu was ever recalled by the manufacturer.
>>>>>>
Wow. Obviously, somebody doesn't remember the fdiv pentium. I'd hardly call the 1.13 GHz P3 infamous. They were so rare that the recall affected all of the five people who actually bought one. Besides that, and the trouble with the P4, which I referred to when I said they have had some initial problems with new products, what else have they done wrong?
In fact, it wasn't until the Northwood P4 2.53GHz variant that Intel started doing some things "right"--and that's been for only a few months now.
>>>>>>>>
Just because AMD was a good competitor doesn't mean that Intel wasn't doing the right things. They were working on jacking up the speeds on the P4, and that'll pay of significantly now that they've got a handle on it.
Everybody knew that the low IPC in the P4 would be made up for, eventually, in sheer clock speed--that wasn't debated as far as I can recall.
>>>>>>
Read up on the
What hardly anyone suspected was that AMD would be able to extend the Athlon architecture so well against Intel's Pentium architectures. Indeed, with a new stepping of the Thoroughbred core which started shipping only last week, The Athlon holds its own against the P4 and will do so up to the 3GHz level and maybe beyond. After that comes Hammer, which supposedly will start shipping at close to the MHz range where Athlon XP leaves off, ~2.4GHz.
>>>>
Err, most of the stuff I've seen pegs the Athlon XP at around 2 Ghz, not 2.4.
Only thing is that Hammer will be at least 25% faster than Athlon XP clock for clock, which makes it considerably faster than NOrthwood clock for clock, yet it will have no trouble scaling up in MHz.
>>>>>
I doubt it will have "no trouble." Due to the architecture, it simply won't be able to scale to the kind of clock-speeds the P4 will. Intel is gunning for 5+ GHz, real soon now. AMD will have a hard time keeping up.
A deep unwavering belief is a sure sign you're missing something...
Hey, at least you get 16 GPR with x86-64.
So let me get this straight, the new Intel's require a complete hardware shift in order to be useful, just like Apple. Both have 64 bit chips in the works. For the first time Apple, Sun, IBM et al will be on a level playing field with Intel. If Intel succeeds with Itanium then none of the software owned by any company will run, necessitating purchase of a new OS, programs, ect. Doesn't this realy put Apple, Sun and IBM in an interesting position? For the first time companies will see a level playing field. I would hope companies see this as a golden time to dump x86/Intel architecture and go instead towards more open solutions. After all, they have to switch hardware and software anyway. Why not think different?
This is the fellow who, before World War II, was concerned about the coming conflict so he converted his savings to precious metals and buried them
So let's see: Gold was $20 an ounce in 1934. Hmmm.... he could have kept the $20 instead of an ounce and would have had all that cash instead of gold, which is worthless per ounce today.
Obviously Turing was a complete idiot.
It's all well and good to be able to execute 4 instructions at once, but most systems spend a large portion of their time in library routines (strlen), function prolog/epilog, and so on. Even assuming that you are running some pretty hard number crunching code that can parallelize the inner loops, you are still starving all of the other threads/processes that could be running.
Why not just work on n-way SMP, so that an application can monopolize one or more processors and still have cycles to spare for mundane work.
Matt Slot / Bitwise Operator / Ambrosia Software, Inc.
"The way to make money during the boom is to have built good products during the preceeding bust, and have them ready to sell once there is a market for them."
But is Itanium a good product? That was the question of this article. Even during a good economy there will not be a big market for Itanium because Intel just went into the wrong direction with it's design (bloatware). At least I believe so. And Intel agrees with predicions of a 10% market share of the server market.
Even in a good economy, people will just buy from competitors as Google is going to do (and Google has good economics already). With other X86 compatible processors or platform independent programming, it's a buyer's market and Itanium just doesn't seem to be the best buy.
I can applaud the decision to make a break from the old X86 architecture, but why did they design it as structurally complex bloatware?
First they head into the direction of more simplicity (switch to RISC core inside the CISC Pentiums) and then they double back into the complexity trap with Itanium.
Humans are just much better at improving simple things than they are at improving complex things. Why didn't they just go multi-core or something? I guess it's their CISC cultural heritage.
And if I may go slightly offtopic for a bit. I think there's something unelegant about those extremely power hungry chips. Something just doesn't feel right about the fact that your solid-state chip's continued existance is dependant on the oil on the ballbearings of a spinning bit of plastic, and that it's just a matter of time before your PC/server breaks.
A PC should be as solid-state as possible, just make sure electricity keeps going in and it runs. I think server farm cowboys/girls agree with me. They have better things to do than replace fans all day.
For this reason I like the Transmeta Crusoe, Via C3 and IBM G3.
However, even though it's power hungry, I do like the Intel Pentium 4's ability to survive the removal of it's heatsink, and continue running Q3 like nothing's happened when you put the heatsink back on. Could you underclock and undervolt a P4 3GHz to 1.5GHz and run it using a giant heatsink without a fan? I bet you can! At least it would survive.
- -- Truth addict for life.
Intel isn't targeting the consumer market here (AMD targets everyone with opteron) but eventually that's where 64 bit CPUs will be sold as well. They target the business market. Here is a totally new 64 bits chip that is created to compete with IBM and Suns offerings. And IBM and Sun support their beasts. Intel, Dell and Microsoft are still noobs at giving that type of support. That's going to take time too. And it's going to take lots of time and impressive stats to convince CTOs and IT heads to run wintel in their datacenter.
What the itanium will be great at is to run as a single server for small workgroups that need computing power. Universities and the like.
Of those to whom much is given, much is required.
This wouldn't be the first time Intel has screwed the pooch. Do a google on "iapx432" for something that turned out possibly even worse than Itanium might. (It was a nice chip design, on paper -- and it eventually met all its design goals, of which performance was not one, alas.)
-- Alastair
Intel Corp. said it failed to reach an agreement in a $250 million dollar patent lawsuit by computer- services company Intergraph Corp., which already was paid $300 million by the world's biggest chipmaker to resolve an earlier dispute.
some info can be found here:7 0,146182,00.html
http://www.intergraph.com/intel/legalpic.asp
and
http://straitstimes.asia1.com.sg/money/story/0,18
Today, Intel and intergraph anounced a break down in cour ordered mediation to resolve a quarter billion dollar patent infringement suit against the ITanium.
In July last year, Intergraph (www.intergraph.com) brought a lawsuit against INTEL alleging the basic design of the Itanium violates ateleast two patents they had held for ten years. Intergraph alleges the concept of software based instruction routining in highly parallel architechtures was developed for their C5 (aka clipper) chip.
Itanium basic design is based on a HP concept for highly parallel processing in which the order of execution on the chip can actually create race conditions for dependencies in calculations. This allows performance enhancements and simplication of handshaking harware, since basically the chip does not have to wait for the slowest operations. INstead the job of preventing race conditions falls to the compiler. The compiler must model how the processor will execute an instruction in the context of the other instructions the chip will be executing in parallel and then re-order the micro-code to prevent erroneous computations.
It would appear the methodology for achieving this was patented by intergraph for the C5 chip. The C5 chip project was eventually abandoned and intergraph parteneres with intel to replace the CPU in their workstations with pentiums.
We all know that intel was previously accused of stealing the ALPHA processor designs and that law suit was "settled" by intel buying out the impoverished ALPHA (dec).
This law suit is for 250 million dollars. which is about 5 % of the entire 5 billion dollar development const of the Itanium. Mediation talks have broken down so the Suit will presumable go ahead. If you are interested try a google search, there's lots of info out there as this trial has dragged on for over a year.
Some drink at the fountain of knowledge. Others just gargle.
Correction, it does have legacy support, it's supposedly awkward and unnatural. Anyone have heads up on how this compares to the Opteron?
Of those to whom much is given, much is required.
Err, I meant the AMD Hammer.
A deep unwavering belief is a sure sign you're missing something...
Such a processor is of no use, unless you're a gamer. That's why the PC market is down now: no one cares about having a 3+ GHz processor.
That's quite true for some architectures. However, note that the PowerPC CPU, for example, does a lot of optimizations at execution time with branch caching, speculative execution and other predictive techniques. This, on a code that has been somewhat optimized at compilation.
The question is not whether the IA-64 is the only processor to do these compile-time optimize. The question is whether it's wise to rely mainly on compile-time static optimization when you hope to be a performance leader. Turing says that you cannot, because static optimization, obtained by guessing the execution code path, is always inferior to dynamic optimization generated from the actual code path with the actual data.
Do you have pointers regarding the amount of dynamic optimization in the IA-64? In other words, if the compiler in only run-of-the-mill, can the IA-64 still perform?
--
Mad science! Robots! Underwear! Cute girls! Full comic online! http://www.girlgeniusonline.com/
2^36 = 2^4 (16) * 2^32 (4G) = 64G.
...bank selecting it in for the *entire duration* of the operation.
Also, the use of the extra 4 bits to get the full 36 bits requires use of the PAE.
Which, if you had ever read the processor manual, you would know requires *bank selection*.
Bank selection is incredibly useless. It doesn't let me have more than 4G of virtual address space split between user space and kernel space, combined.
Since I already have that with only 4G of RAM, and since the people who build servers that use that much memory aren't *stone stupid*, and tend to *dedicate* them to tasks, the extra RAM is basically useless.
The best you can get out of it is *a RAM disk*, and that still requires you copy crap all over.
You can't use the memory for *DMA buffers* or anything useful on which *the user process or the kernel operates*, without
The reason anyone want more address bits is to access more memory *simultaneously*, in a single program. If I were bumping my head on available RAM, I'd implement demand paging, and be done with it, wouldn't I? But that only works if I have the virtual address space to back the real memory *plus* the virtual memory, and I'm still limited to *32 bits* there, aren't I?
I guess maybe you are one of the Intel engineers who thought *bank selection* was a good idea?
Let me give you a clue: it's not.
Let me give you another clue: Segments are for worms: segment registers were a bad idea, too, which is why no one uses the damn things, except as extra registers ro hold things like thread contexts.
Try naming one modern OS that uses segments; the last one to do it was medium model Windows 3.11 and SCO Xenix 2.x, a full *decade* ago.
PS: If I wanted to use bank selection, I'd be writing software for the Commodore 64.
PPS: While we are at it: ring 1 and ring 2 were also stupid ideas; no modern OS uses anything other than ring 0 (kernel) and ring 3 (user). At least on the VAX architecture, you could use ring 2 to implement asynchronous system traps to call user functions from the kernel that ran as if they were in ring 3, only on a different stack. Too bad Intel's ring 1 & 2 aren't even useful for that.
PPPS: Try building a processor that's friendly to how people are actually going to use it, instead of pretending we live in some imaginary hardware designer universe where Spock has a beard, for a change.
-- Terry
Hey regarding compilers.... Does anyone have any info about how .NET will go on Itamium, or even Java?
.Net and Java are all about JIT compiliation of some intermediate bytecode to native machine code as needed.
.Net assemblies at runtime will have a higher performance overhead on Itamium, if the JIT compiler wants to extract the best performance out of the chip?
Think about it,
Itanium is all about moving the complexity of moving out-of order execution and stuff onto the compiler at compile-time instead of doing it at runtime with silicon.
Doesn't this imply that JIT compilation of Java Bytecode or
Has anyone here seen/done any benchmarks for this?
The difference is you're saying what should be, and Grishnakh is saying what is. Intel *should* be doing whatever it takes to make the code that gcc produces for ia64 optimal, but they won't because they're being greedy and believe that they can sell the compiler as well.
Maybe they're right, and maybe they're wrong, but given the lack of uptake on ia64 so far, and given the relative amounts of money that they stand to make on the compiler software versus the ia64 hardware, I think they're making a big mistake.
It rather looks like a case of company politics, with nobody able or willing to see the big picture and "do the right thing" regarding "giving away" the compiler technology.
Tim
Hold it right there.
Didn't Intel introduce the Northwood-core Pentium 4's starting with the 2.0 and 2.2 GHz versions first? I think increasing the L2 cache on the CPU die to a generous 512 KB did wonders for the CPU. It'll be very interesting to see what the Prescott-core Pentium 4's with its 1024 KB L2 cache does in terms of performance.
Right. My point is though that this remains an easy option for Intel at any point. That is there aren't technical reasons (like problems with GCC) that prevent adoption of Itanium. For example if Apple made it a condition of switching this wouldn't be an issue.
This is besides the fact that I think Intel is likely to help out GCC regardless. They are going to want to make sure they stay ahead of Hammer and Power4 for Linux BSD.
Sounds good to me. Winter is coming, I'll just slap an Itanium in and turn off my space heater this winter.
Seriously though, if the chips take this much power, is this peak or average? If the average load is anywhere near this, anybody using this is going to see a rather nice jump in their power bills.
Every time I turn on my PC the lights dim - phorm
With PAE you still cant use more than 3GB per process. And, unfortunately, it's not too uncommon that you need more than 3GB.
You only have to look at it's (itanic) performance under gcc to see the answer to this question. I seem to remember performance figures 20% of that obtainable with intel's compiler.
Stick Men
Itanium is a step backwards for software. It make the tradeoff of giving you somewhat better performance for a few languages and benchmarks, with complex compilers, while being even harder and more problematic for anything that deviates from the canonical benchmarks. That locks new kinds of software even more into a straightjacket than it already has been.
If Intel sees dynamic compilation as the solution to the complexity of Itanium, they should do the same thing Transmeta does: define a simpler instruction set for compilers to target and make the dynamic compilation and optimization software effectively part of the chip.
At the rate things like Microsoft Windows and X Windows and KDE and Gnome and all that other bloatware are growing, in just a few years almost everything will need 64-bit just to load up.
now we need to go OSS in diesel cars
Shouldn't that be the DEC Alpha? So quickly these things are forgotten .... I'm still waiting for the day they start talking about the Compaq VAX!
Alison
"It is a miracle that curiosity survives formal education." - Albert Einstein
Hint: He is not some random Microsoft employee.
Mea navis aericumbens anguillis abundat
You _have_ to compile with GCC. If you're writing C++ code at least. Or you wont be able to link with any C++ libraries. Doesnt seem like the C++ abi is going to get standardized ever, does it.
I have the money to spend, but what I dont have is the time to dink around trying to get whatever code to compile with whatever compiler to link with whatever library. Either it works with everything compiled in GCC out of the box or it's a useless pain in the rear.
And at roughly 1/40th the price of the smallest POWER 4 box you can buy, my wallet's pretty happy too!
Where did you get an Itanium-2 machine for$424.93 ?
At that price they might just displace the Athlon for price/performance!
The Itanium isn't Intel's first attempt at a much more powerful non-x86 architecture. In the late 1980s, Intel was pushing the i860--eventually succeeded by the i960--which really were amazing compared to the x86 line, but they flopped.
And who can forget Motorla's 88000 line, which was meant to be the follow-on the the 68K?
You're right, it is utterly straight-forward, and that's why AMD has done it. You can see this from AMD's documentation about the Hammer architecture at www.x86-64.org. Intel can't do the same because it has bet the farm on itanium and it needs to persuade the corporate world to buy millions of itanium machines to recoup its R&D spending. They may have backed the wrong horse because 10 years ago when they started the project it seemed like the best thing to do. They will try every marketing, slaes and PR trick in the book to noble AMD's architecture simply because the cannot afford for Hammer to succeed.
Stick Men
My first desktop UNIX box at work was a DG/UX on 88K. Extremely slow (it was outdated by the time I got it). Basically used it for an XServer, and every once in a while used gcc on it just to check compiler warnings (because no one else used it, I could give it the latest gccc, 2.7.2 at the time, without disrupting anyone else).
As I understand it, the Opteron will act exactly like a regular x86 processor until you start using the 64-bit instructions. That's it's claim to fame, that it will be extremely fast for 32-bit apps, and will also support 64-bit extentions. For a practical example, having MMX or 3DNow! or SSE doesn't affect the processor when it's doing a regular integer divide, and I expect that the Opteron will be the same.
Though I have been known to be wrong...once...and I was drunk at the time...
It's been a long time.
Yep, that's what you'd think. Apparently Intel doesn't think that way.
Aren't PS2 and GameCube 128 bit?
Not true - the NYT came to Google to work out a deal so that google would spider the NYT's news articles (which would have been unavailable to google's web spiders because of the NYT's registration). This info comes from this column which is referenced by this /. article. It's a good read on how the NYT needs google more than the other way around.
Other search engines would love Google to make stupid decisions to censor some of the better news content. A partnership with NYT would just give them one leg up on Google. Google can't afford that.
Okay, time to drop the crack pipe. Nobody said *anything* about censorship. What *is* being discussed is the NYT's decision to allow google news readers to view NYT's article without having to register (a topic you completly screw up in the next paragraph).
On another note, regarding the "moronic registration process", I'm sure NYT has had plenty of time to re-evaluate its registration process to determine if it is counter-productive. I'm sure the money they save on consulting and profiling studies because they can link their stories to users far outweighs the $.005/hit of advertising money lost due to the 5% of its potential readers that are turned away.
I seriously doubt the *main* reason the NYT has registration is for profiling studies. They want email addresses to sell (IMO) - an email address from a NYT reader is worth more than one from, say, etoys.com (or wherever). They decided it's worth it to them to go without the email addr in order to get the page view from google (and more importantly, a possible daily reader and/or subscriber). Besides, with the referrer-google, you onl;y get to view the *one* article without registration - if you then go to another NYT article linked from within NYT, you have to register.
Personally its not God I dislike, its his fan club I cant stand (bash.org)
They have a concession speech already written -- it's called Yamhill. But of course if they run with that they're essentially admitting that AMD was right all along and they've officially lost control of the ship. /Brian