· Content
· News
· Articles
· Mailinglists
· Knowledgebase
· Trouble Tickets
· Files
· Glossary
· Links
· Compatibility Lists
· Forums
Welcome to our website
To take full advantage of all features you need to login or register. Registration is completely free and takes only a few seconds.
IBM mocks Itanium server sales - again
Posted by: duke on: 09/20/2004 03:32 PM [ Print | 23 comment(s) ]
The Register is letting us know that IBM has decided to mock Itanium sales... again.
It's fairly typical for IBM to poke fun at Itanium despite its place behind HP and SGI in sales of servers based on the chip. IBM pulled in a mere $8m from Itanium shipments in this year's second quarter, while its Power-based server line generated billions. IBM sells Itanium boxes because its role as a one stop shop demands it not because IBM is a fan of the processor.For more information, read the rest of their article.
« Microsoft's Chief Linux Strategist Interviewed · IBM mocks Itanium server sales - again
· OpenBSD's Theo de Raadt talks software security »
Comment
|
duraid SMP Qualified Posts: 387 Joined: 2002-03-31 |
#31858 Posted on: 09/21/2004 04:24 AM
IBM is a leading member of the Itanium ecosystem? heh, you learn something new every day on the reg. |
Comment
|
i_wolf labhair dom as gaelige Posts: 2097 Joined: 2002-11-19 |
#31859 Posted on: 09/21/2004 04:37 AM
Personally I always get the feeling that you need to read between the lines with the inq/reg. Much of their journalism seams to be tainted with exageration and a strong bias against anything that does not carry an AMD moniker. In this article, I do not see where IBM openly 'mocked' the itanium. They showed a slide showing that sales were not particularly strong in the Itanium market. Is this not to be expected? IBM is going to favor its own home grown Power series over the Itanium in territories where both processor families overlap. Again I believe a spin was put on this presentation (by the reg) Hung like a donkey. Go like a horse! |
Comment
|
Occupant Registered User Posts: 2421 Joined: 2002-03-04 |
#31860 Posted on: 09/21/2004 05:56 AM
Ya Mom, I planned to have a D average.... |
Comment
|
LRSeriesIII Aspiring Rocket Scientist Posts: 1128 Joined: 2002-08-29 |
#31861 Posted on: 09/21/2004 07:24 AM
I have to agree with you here, and have to wonder if it is not perhaps more the sites in question playing to what will draw more readers. It seems like there is a bit of a knee-jerk reaction against certain products and companies...especially Itanium. IBM pointed out that previous forecasts of Itanium sales have been far too optimistic. This seems to me to be a very valid point, and if I was in IBM's position (pushing the POWER platform in a number of markets which the Itanium is also aimed at) is what I would be talking about. The presentation itself was a "technical outlook" for 64-bit based computing, and a presentation on why POWER is the way to go (according to IBM). The presentation also talked up Opteron and Xeon 64 bit platforms. I think that the claim that IBM is a leading member of the Itanium ecosystem is at least partially justified, however. The article justified the statement by pointing out that IBM is third in revenue from Itanium systems, behind HP and SGI. Of course, the Reg did not mention the complete numbers, which also indicate that IBM, while indeed third in revenue from Itanium sales in the second quarter of 2004, is a very distant third, behind SGI at $40 million (IBM had $8 million), which is itself an extremely distant second to HP at $250 million. Whether or not that qualifies IBM as "a leading member of the Itanium ecosystem" is up to the reader (and given the wide spread of those numbers, should have been left up to the reader by the Reg), but there is at least some sort of justification for it. I suppose this all leads to a question that (I believe) has been lightly touched on before, but which is still really bugging me: why is it that there are such flame wars over the Itanium? The Mac/PC debate is stupid, but I can at least understand what leads to all the evangelism. The same is true of the Intel versus AMD x86 processor debates and the nVidia versus ATi arguments. But what is it about Itanium that seems to get people going in such an irrational manner on both sides? Look, I am not in any position to call the Itanium a success or failure (I have said this before). I have seen various claims that it can produce results untouchable by any other platform in certain situations, and I have no reason to doubt that this is true. I do know enough about it, however, and have seen enough to know that it is not the do-all be-all end-all of processors, or even of high end (non-x86) processors. For that matter, it does not make Itanium necessarily a better platform. I can set up situations in which a 386SX will outperform my AthlonMP's (let's just say it involves some wire splicing and a fuse...  ). The 386SX is clearly the inferior platform of the two, even though it wins that particular performance competition.
In addition, the Itanium has clearly fallen short of sales expectations, both those of Intel and of third part analysts. I am legitimitely curious as to why this is. It seems, however, that a legitimite discussion about the topic is impossible to have without people on both sides making outrageous statements, personal insults directed at anyone disagreeing with them, and a bunch of people coming out and making claims that they, frankly, are in ABSOLUTELY NO POSITION to make. Sorry for the rant, this has just been really bugging me lately. I can deal with the fanboy-ism among the uber-1337 gamers and such (or at least pretend to deal with it), but the discussions of this sort of kit (serious equipment for doing serious work) are not the place for such behavior. Now if you will excuse me I have to go work on my aero homework... <* steps off of soap box and shuffles off... *> ->Computers ->Folding for team 3074 |
Comment
|
Hydra Performance Junkie Posts: 174 Joined: 2001-03-09 |
#31862 Posted on: 09/21/2004 08:31 AM
I knew it, Itanium is crap!  Overclocking is a blast! Do I smell something burning?  |
Comment
|
rmn oh my, it's huge! Posts: 6013 Joined: 2002-01-26 |
#31863 Posted on: 09/21/2004 08:32 AM
The Inquirer (and its more serious brother the Register) is a very democratic site: it bashes everyone equally (or at least in the direct proportion of its targets' boastfulness). If you read their letters (or just hardware forums in general) you'll see people complaining they're anti-Intel, anti-AMD, anti-NVidia, anti-ATI, etc. (and let's not forget their feud with Sun, that lasted until a few months ago). If someone claims to have the industry's best [something], you can be sure they'll pounce on them as soon as the slightest flaw is encountered (which other sites post an article for every item added to the Opteron's or Itanium's errata?). But they do it with everyone, at one time or another, they do it with a sense of humour, and, above all, they have real news (ie, new news). RMN ~~~ |
Comment
|
LRSeriesIII Aspiring Rocket Scientist Posts: 1128 Joined: 2002-08-29 |
#31864 Posted on: 09/21/2004 05:17 PM
This is the kind of comment I am talking about...:rolleyes: ->Computers ->Folding for team 3074 |
Comment
|
i_wolf labhair dom as gaelige Posts: 2097 Joined: 2002-11-19 |
#31865 Posted on: 09/22/2004 02:39 AM
[tongue_in_ cheek!] But it runs doom 3 slower than my Pentium 4 2.4Ghz. Why would anyone spend so much money on an Itanium when it is obviously a much slower chip!!! :P
[/tongue_in_cheek!] Hung like a donkey. Go like a horse! |
Comment
|
rmn oh my, it's huge! Posts: 6013 Joined: 2002-01-26 |
#31866 Posted on: 09/22/2004 03:23 AM
Actually, all joking aside, is there any relatively common, useful, commercial application that is noticeably faster (and faster-per-dollar) on the Itanium than on Opteron / Xeon? RMN ~~~ |
Comment
|
Calum Running on empty Posts: 1916 Joined: 2002-10-11 |
#31867 Posted on: 09/22/2004 04:20 AM
People need to realise that for many things, the cost of the hardware is NOT IMPORTANT AT ALL. It doesn't matter if the CPUs are My Computers|2CPU.com SETI@Home Stats|My eBay feedback|My Heatware|Have you seen the future? |
Comment
|
i_wolf labhair dom as gaelige Posts: 2097 Joined: 2002-11-19 |
#31868 Posted on: 09/22/2004 04:55 AM
You raise an interesting point. Most software (commercial that is) appears to be high end and extremely expensive e.g. Houdini sprung to mind as an app that is powerful , useful , common maybe in high end production houses... however I have absolutely no idea what performance is like compared with x86 in this regard. Just noticed that there are quite a few other powerful apps available (Mathematica etc..). I do however believe that the majority of people that will buy an Itanium are likely buying it for in house applications... these will appear to be the same people that will bother to optimize for it and yield the potential results an Itanium can give you under the ideal conditions. Hung like a donkey. Go like a horse! |
Comment
|
duraid SMP Qualified Posts: 387 Joined: 2002-03-31 |
#31869 Posted on: 09/22/2004 05:34 AM
I could write pages and pages to answer this, but I won't, unless someone asks. Here's the simple, useless answer: Yes. Here's a bit more detail: There are a few such applications, but I'll take just one as a "case study". That application is called Gaussian (http://www.gaussian.com). Before talking about its performance and price/performance on Itanium vs. other sorts of computers, I'll first say a bit about why I think this particular example fits RMN's idea of a "relatively common, useful commercial application." Gaussian is used by chemists and materials scientists, mostly. It is relatively common in that most pharmaceutical companies, materials companies (e.g. DuPont, 3M), related research labs (in university and industry) etc would have a site license for this piece of software: it's hard to estimate, but I'd say that on any given day, there are at least 50,000 people who run a copy of Gaussian. It is so common, in fact, that it is sometimes jokingly referred to as "the chemistry virus" because it seems to sneak its way onto computers, soaking up CPU time, even when the admins (or other users!) don't want it to. 
It is useful in that you can do a wide range of things with it: from basic research (e.g. trying to understand the structure of certain materials or biological processes) to engineering (testing new medicines before trying them in real, living creatures, designing stronger, lighter composite materials etc) there's not much you _can't_ do with Gaussian that you could still call "chemistry". It is commercial: there are discounts for universities but even there a university should expect to pay $10,000 or more for a copy. I'm sure some of the larger drug companies will be paying $20,000/year or more just for their license and ongoing maintenance (upgrades). Right, that's out of the way. I hope that if you've read up to here, you'll think that "Gaussian" is a reasonable example to be looking at (RMN, any comments?) Time to look at the performance. Gaussian includes some tests (for accuracy) that can also be used as benchmarks (they take more than an hour to run, on a single CPU). One such test is #397 (this is for "Gaussian 2003") Well, the figures! 1.4GHz/3M Itanium 2, Intel 8.0 compiler: 6,683sec 1.4GHz Opteron (this is a "240") with Portland Group x86-64 compiler: 17,119sec I haven't bothered re-running the test on a faster Opteron (even though I have access to a few) because it's clear, to me at least, that it will never be as fast. Assuming _perfect_ scaling (which you _won't_ get), the Opteron would need to be running at almost 3.6GHz to match the performance of the Itanium 2. Just a quick mention of prices is probably appropriate here: while Opteron 240s are very cheap, what isn't cheap is a chemical engineer. The average salary is probably something like $60,000+/year, and from what I've seen (I'm not a chemical engineer myself but I have certainly worked with a few over the years), they would spend something like 10-20% of their time waiting for results. So, let's say that _halving_ the amount of time they have to wait will save the company or whatever $3000+/year or more. Bear that figure in mind as we now price up a couple of boxes to run Gaussian: 2xItanium 2 1.4GHz/3M: $2000 (total) 8x1GB PC2100 ECC: $1440 Supermicro 6113M-i (this is a Supermicro i2DMR-iG2 mobo in a 1RU case with a 500W PSU): $1600 ATA disk(s): $200 (say) All up: you should have change from $5,500 for a 2x1.4GHz/3M Itanium 2 server, with 8GB of ECC RAM, 120GB of mirrored 7200 rpm disk, sitting happily in a 1RU case with power supply. Now, the Opteron, **to run at half the speed of the Itanium, costing the employeer an extra $3000/year in waiting time** 2x1.8GHz Opteron (244): $600 (total) 8x1GB PC3200 ECC: $1840 Tyan Thunder K8S Pro: $400 1RU case and 400W PSU: $400 ATA disk(s): $200 All up: $3,440 Right. What's happened here? We've saved $2000 by getting an Opteron instead of an Itanium. But with the performance hit we're taking, we're losing _at least_ $3000 every year in lost productivity. "What about just buying more Opterons with the money you save!" I hear you saying. Well, the problem is this: Gaussian, like many codes out there, doesn't scale well. Adding extra CPUs (e.g. moving to 4-way systems) buys you only 1 extra CPU worth of performance. And if you decide to "cluster" a whole bunch of 2-way systems, the performance scales worse than that again, even if you spend $1000 per machine on interconnect such as Myrinet or Infiniband. Even if you spent $1,000,000 on a large Altix or IBM p690, it is unlikely that you could perform Gaussian simulations more than 10 times faster than you could by spending $30,000 on a 4-way Itanium 2 box with the powerful 1.5GHz/6M CPUs. Now, if you had many seperate users working on different things simultaneously, that's quite a different story and clustering becomes quite useful there. Also, if your _code_ (say 3D rendering, not long simulations of protein folding) scales well, again, clusters can shine. But there are some difficult problems out there that need to be solved, and for which there are no known efficient parallel algorithms. Combine that problem with a need for a solution _fast_, and in any situation where time is money, it makes sense to buy the best hardware for the job. As more and more people figure this out, more and more people are buying Itanium kit (even if they are really like a "blip" in the ocean of x86 and now, x86-64 computers that people buy in vast quanitites every day...) *looks up* okay, i've typed a bit. But trust me, this is just the tip of the iceberg. If anyone wants to hear at length about (IMHO the more interesting story) why the IA64 _architecture_ (not the chip), is superior for scientific research, and why no matter what Intel or AMD do with their x86/x86-64 parts, they will never be able to make a chip that is competitive without basically turning it into an IA64 chip, and having x86 run in some sort of emulation, let me know and I'll get around to writing a proper article (this is the sort of thing i should be doing anyway, for my job) If anyone things I've just made the gaussian figures up, go run the benchmark yourself (anyone at a university should be able to access a copy without _too_ much difficulty) or check out this website: http://www.chemie.uni-dortmund.de/groups/ocb/projekte/mg98b.html (_lower_ is better) If anyone has an issue with the system pricing I've given here, please post your own component details and their pricing here WITHOUT starting a flamewar. If anyone thinks there's something weird about Gaussian, and other programs must run faster on Opteron/Xeon: sorry, it's just not true. Gaussian's #1 "competitor" (it's actually better, but not so commonly used and not _quite_ so general purpose) is called Amber: check out the benchmarks here: http://amber.scripps.edu/amber8.bench1.html (_HIGHER_ is better. 'ps per day' means "picoseconds of reality simulated per day of computer time" (so the more, the better) this last collection is great because the very best compilers (e.g. IBM xlf on Apple G5s, Intel Fortran on Itanium, pathscale on Opteron) are being used. I hope this answers the original question, in some reasonable level of detail. And at least, I hope to have _started_ to convince some of you that for doing chemistry, chemical/medicine engineering or materials science on a computer, Itanium is worth serious consideration. |
Comment
|
duraid SMP Qualified Posts: 387 Joined: 2002-03-31 |
#31870 Posted on: 09/22/2004 05:48 AM
Just to tie my last huge post in with the topic of this article: if you get a quote for a 32-way Altix and a 32-way IBM p690, then compare the performance of Amber on these two systems, you'll see that the Altix offers you about 5x the price/performance of the IBM. On another note, the really nice thing about Itanium 2 is that anyone whether you have a huge 128-CPU million-dollar Altix, or a tiny 1RU box you built yourself with a Supermicro motherboard, will be able to upgrade to the astounding "Montecito", dual-core, "hyperthreading" Itanium 3 processors when they are released next year, just by plugging new CPUs in. If you think things look bad for Opteron, Xeon and G5 on scientific codes now, they are going to get worse. The price/performance gap will widen, in spite of dual-core Opterons and Xeons, I guarantee it. If anyone wants to place $50-or-so bets on this I am MORE than happy to take your money. |
Comment
|
rmn oh my, it's huge! Posts: 6013 Joined: 2002-01-26 |
#31871 Posted on: 09/22/2004 06:36 AM
Gaussian: I had never heard of it. I have no idea how common it is in the chemical industry, but it's definitely not what I would call a "common" application. It's like saying that, since DeWalt's DW621K router is very popular with woodworkers, it's a "common electrical appliance". It's not, outside that very specific niche. I was thinking more along the lines of CAD / DCC / database software. Software that people here are likely to have used, seen in action, or at least heard about. I will ignore the fact that you're comparing it on the slowest Opteron, that the prices are inflated, and the false logic applied to TCO (even if a computer runs at half the speed, your employees aren't costing you twice as much, because they can be doing other things - such as working on a second system - while the first system is busy). All that is irrelevant, because the first issue wasn't addressed: comparing the performance on a common application (i.e., one that people here can run benchmarks on, and compare the results with yours). So, are there any reasonably common programs that have Itanium versions, and where the Itanium version is clearly faster than a x86 / PPC version (comparing the fastest systems available, or systems of the same price)? CAD, databases, renderers, image processing, etc.? Compressors, even? How do things like RAR / ZIP / etc. perform on Itanium systems? Also, I noticed the Lightwave benchmark site has the Itanium listed but no Itanium benchmarks. Do you have any information about Lightwave's performance on Itaniums? I'm sure the Itanium is great when running highly specialised and highly optimised software for very small niches. But that wasn't what Intel had planned for it. Somewhere along the way, they seem to have given up (ex., they helped Discreet rewrite 3DS MAX's renderer to take advantage of SSE2 on the P4 - they could just as easily have helped them port at least the render node to IA-64, but didn't). You wrote in some other thread that there's a 3rd party IA-64 compiler that is actually faster than Intel's. When Intel, that has the deepest knowledge of the CPU, can't even be bothered to invest its resources developing a good compiler for it, how can other companies (or potential clients) be expected to take the Itanium (as a "solution", not the CPU itself) seriously...? Personally, I don't think EPIC is the way to go, in the long run. But I think it could have gone a lot further than Intel took it (I actually considered buying a couple of Itanium workstations ages ago - but there was no software to run on them). Calum: of course price matters. If you can get identical performance for less money, why would you pay more? Especially considering that you'll probably be replacing it in a couple of years? If you can't get the same performance, then the price difference may be justified (assuming your time is worth money). But in most cases, you can; simply by adding more CPUs or more nodes. Most CPU-intensive applications are pretty well optimised for multiple CPUs. RMN ~~~ |
Comment
|
duraid SMP Qualified Posts: 387 Joined: 2002-03-31 |
#31872 Posted on: 09/22/2004 09:05 PM
That's fine.
I disagree. In fact, I'm certain there are more legal users of Gaussian than there are legal users of, e.g., Mental Ray.
Whatever.
Read my post again, carefully.
They are? Where?
Read my post again, carefully.
You yourself said that it must be _commercial_ software: are you suggesting that people here would go out and buy some software just to run benchmarks? This is forgetting, of course, the utter stupidity of benchmarking systems with vendor-supplied binaries and not source code. There's a reason the SPEC guys give you source...
I give up trying to make you happy. I would have thought that 100,000 gaussian users (and that's a conservative estimate) would qualify it as "reasonably common" but evidently not. Image processing flies: how about you suggest a specific filter or something like that, and I'll code it up, and then we can go run it on different machines. Renderers? Mental Ray on Itanium is quite fast, but this is expensive. Itanium is very fast running POV-RAY (and that's free!) - how about you pick a benchmark scene and again, we can all come up with some figures? Databases fly on Itanium when they are large and/or complex: anytime you can get (more of) the indices/frequently hit regions to fit in cache, performance goes up. Compressors don't run so fast on current Itanium chips because they are in-order processors. I mean, they're still competitive with anything else out there clock-for-clock, but for a compressor, clockrate counts for a lot. e.g. the Pentium 4 kills the K8 on a small lempel-ziv encoder.
No idea, sorry.
Actually, neither Gaussian nor Amber are tuned for Itanium at all. They just run fast on Itanium.
Geez, you know a lot! Pray tell, what had Intel planned for the Itanium line?
Oh please, why would anyone buy IA-64 nodes for 3DS rendering when they could get by with dirt-cheap P4 boxes? 3DS is a low-end piece of software to start with, and on top of that it's pirated like all hell. Invalid normal arrays, anyone? 99% of the people who use 3DS wouldn't have the money to sink into _one_ Itanium system, let alone a meaningful quantity of them. Intel knows this. 3D rendering is a "solved problem". Old hat. Yes, new software comes out once in a while and new algorithms a little less often, but it's basically done, and it certainly scales well. Protein folding is not a solved problem. Guess in which of these two applications Itanium finds a home.
There's a tradeoff, of course. Intel's compiler is stable, reliable and builds code rather quickly: IMPACT (I think that's what you're talking about) is more of a research compiler: not so stable, reliable and much slower. Then again, on some programs the Intel compiler will produce faster code anyway. And on others, the HP compiler will give the best performance. And on _some_ programs GCC will actually produce the fastest code (shock, horror).
Now you are talking out of your ass. AMD have the deepest knowledge of the Opteron (according to your logic) - do they have the best compiler for it? Hell no! The best compiler for Opteron is Intel C/C++ 8.1. Look, as far as I can tell you're not qualified to judge the quality of any compiler, much less come in here and simply assert that the Intel electron compiler, which I'm sure you've never used, isn't good. It is a good compiler. OK, can't be bothered writing any more... |
Comment
|
Occupant Registered User Posts: 2421 Joined: 2002-03-04 |
#31873 Posted on: 09/23/2004 12:03 AM
I think intel was planning for, was the death of the x86 instruction set. but now with x86-64 it will be around for several more years... |
Comment
|
darkamage Aspiring Duallie Posts: 93 Joined: 2003-06-27 |
#31874 Posted on: 09/23/2004 01:33 AM
What is the actual price and performance? which actually offered more performance? I am interested because on your previous post you mention that even thought the opteron seemingly has price/performance advantage but losing money in productivity. With regards to the thread, I am sure that Itanium has an advantage in certain applications (a lot of in-house maybe?). Whether these would be enough to justify it against the x86-64 or other 64-bit architectures remains to be seen. My personal opinion (more apropriately a guess) is that somehow the two Intel architectures will merge within the next 5-10 yrs... |
Comment
|
rmn oh my, it's huge! Posts: 6013 Joined: 2002-01-26 |
#31875 Posted on: 09/23/2004 02:00 AM
In other words, "Gaussian" (a program I'm willing to bet 99% of the people here never even heard about) is the most common software you can think of where the Itanium is competitive. I think that says it all, really. RMN ~~~ |
Comment
|
duraid SMP Qualified Posts: 387 Joined: 2002-03-31 |
#31876 Posted on: 09/23/2004 02:02 AM
I am sorry, but for that you will have to click some links. Sorry.
Sorry if I gave you that impression - I didn't mean to.
Given that sales have been rising, one would think there is some justification for having another architecture on the market. |
Comment
|
Occupant Registered User Posts: 2421 Joined: 2002-03-04 |
#31877 Posted on: 09/23/2004 02:08 AM
Itanium clearly has a market. But with entrenched 64bit unix workstation vendors fighting for thier lives. (SUN, SGI IBM) Its going to be a rough to get itanium to any decent sized market share, anytime soon. Ironicly, the 64bit workstation market is exactly where SUN and SGI have been trying to diversify away from. |
Comment
|
darkamage Aspiring Duallie Posts: 93 Joined: 2003-06-27 |
#31878 Posted on: 09/23/2004 02:49 AM
More architectures is always better. Let the market decide and let new architectures spawn... If it's bad than people won't use it and if it's good then *maybe* people will use it... |
Comment
|
Insp. Gadget Aspiring Duallie Posts: 89 Joined: 2003-04-17 |
#31879 Posted on: 09/23/2004 07:50 AM
Still the same old zealot, eh? Sometimes I think you're JEC252's twin, that just replaces "Solaris" with "Itanium".
I'm curious, whre are you getting these numbers? Making them up? How do you know how many Mental Ray users there are? Mental Ray is bundled with several modelling packages (ex, 3dsmax, Softimage), making it pretty common. The student licenses are pretty cheap, too.
I think his point, when he said "software that some people here use" was precisely so that people don't have to go out and buy it. Or do you expect "people" to go out and buy Gaussian ($3000) just to check the numbers you're claiming? Does Gaussian even have an official benchmark page?
Most commercial software does not include the source code. It's not a matter of "utter stupidity", it a matter of the real world (a place that you often seem very distant from).
Again, where are you getting these numbers? making them up?
Can you provide a bit more information about that? What does "quite fast" mean? I can't seem to find the Itanium listed on any Mental Ray benchmark.
Is it? Compared to what, exactly?
Does it? Again, where are you getting these values? Because they seem to be somewhat detached from, you know, reality (where a 2.2 GHz K8 does LZW compression as fast as a 3.2 GHz P4, despite running at 1 GHz less).
Then why do people pay several thousand dollars for them? Surely, if the Itanium is the best platform for that kind of work, the package that sells more will be the one that makes the most of it, no?
Given their (lack of) speed and their (ludicrous) price, I have no idea.
3dsmax currently includes a physics simulator, a raytracer and a GI renderer (not to mention several other pretty advanced procedural modelling and animation tools). It has more 3rd party extensions than any other 3D animation package, including cloth, hair and fluid simulators, Monte Carlo renderers, and so on. If this is "low-end", MPEG4 is RLE.
How does that diminish its quality or relevance? Quite the contrary; if it's pirated that means that a) it's considered useful enough and good enough by enough people to want to pirate it and b) its user base (and therfore market influence) is even bigger than its registered users suggest.
Your use of the word "sink" is quite appropriate.
I have no idea what you mean by that; 3dsmax alone has about 15 different shaders, and new ones are released quite frequently. And whether you're using 50 different shaders in a scene or just a single one, it all boils down to speed. If there was an even remote chance that the Itanium could outperform x86 / AMD64 CPUs, companies like Kinetix and Softimage would jump on it in a heartbeat; in high-end animation, time is (a lot of) money. A system that can cut down render times by 20% or more pays for itself in a matter of months. In fact, Softimage did start porting its software (Sumatra) to the Itanium, financed by Intel. The results were so pathetic they told Intel to keep its money and abandoned it.
Are you talking about AMD64?
The Itanium was released in early 2001, and companies had been porting software to it for over 5 years (even Microsoft managed to have two versions of Windows ready for its launch). It's hardly a "new" architecture. I.G. |
Comment
|
duraid SMP Qualified Posts: 387 Joined: 2002-03-31 |
#31880 Posted on: 09/23/2004 11:52 AM
For the record, Insp. Gadget just pointed us all to some povray figures comparing an Itanium 1 to some rather more (cough) modern CPUs. Some of the responses here have been so underwhelming, I think I'll just give up on this thread now. |