What's Next in CPU Land after Itanium?

"I work for a major research organization. Of late a lot of the normal big computer companies have been visiting and preaching the gospel of Itanium. My question to them, and to the assembled masses here at Slashdot is what happens next when Itanium is real? My world view is that Itanium based systems will become commodity products very quickly after good silicon is available in reasonable volume. At that point, why should one spend $8-10k for that hardware from the likes of HP, Compaq, Dell and others when one can build it for $2k (or even less)? In other words, has Intel finally done in most of their customers by obliterating all the other CPU choices (except IBM Power4 [& friends G4, et al] and AMD Hammer) and turned the remainder of the marketplace into raw commodity goods? Lest you defend the other CPUs... Sparc is dead, Sun doesn't have the money (more than US$1B we'll guess) to do another round. PA-RISC is done, as HP has given away the architecture group. MIPS lacks funding (and perhaps even the idea people at this point). Alpha is gone too (also because of the heavy investment problem no doubt). Most other CPUs don't have an installed base that makes any difference, especially in the high end computing world. So what's next? I don't like the single track future that Intel has just because it is a single track!"

Comments

compilers

[avandesande]

Itaniums will become commodities when people figure out how to write compilers for them. That will be in about 10 years.

Re:compilers

[jgerman]

Not likely, it would take a couple of weeks max for the first compilers to appear. Wish I could afford one, I'd love to hack away on a compiler for a new machine.

Re:compilers

[shitfit77]

You seem to miss the point on this a little bit. Although there will be compilers available, there is an extreme difference between a compiler and a good compiler. A compiler works, a good compile is able to utilize an architecture to its fullest (or at least close).

Re:compilers

[spencerogden]

first compilers != useful compilers

Re:compilers

[jmv]

Not likely, it would take a couple of weeks max for the first compilers to appear.

Sure, but the problem is how long before there are good compilers? That's one of the main problems with architectures like Itanium.

Re:compilers

[jgerman]

First compilers are the MOST useful compilers since they are the ones that will be used to create the later ones.

Re:compilers

[bmajik]

You overestimate the brilliance of compiler writers.

LIW and VLIW were tried before. They flopped, because compilers were dumb then. Compilers stayed dumb until midway through the RISC era. Now RISC and CISC are the same, compilers are reasonably bright, and intel is trying its own hacky LIW thing. The compilers are smart enough for a generation 1 LIW design to work, but there may or may not be any indication that they'll be smart. And as each successive subarchitecture of IA64 happens, the compiler will need to change or the chip will need to handle previous generation instructions. Intel is not true LIW in this regard - you should be able to run unmodified IA64-1 bins on IA64-2 chips.

So, some brains are still in the IA-64 chip, meaning the compiler wont have to be _as_ smart, but they'll still need to be smart, and you'll still need a new compiler for each IA64 implementation to get max performance.

Re:compilers

[Zathrus]

Not likely, it would take a couple of weeks max for the first compilers to appear

You obviously know nothing about Itanium, EPIC, VLIW, or pretty much anything else on this topic.

The issue isn't whether or not there's a compiler available. The issue is how GOOD the compiler is. In the case of a Very Large Instruction Word (VLIW) CPU like the Itanium, the compiler is the bottleneck for system performance. Why? Because the premise of these CPUs is that while they have a low clockspeed (750-800 MHz for Itanium), they execute many instructions per cycle - 10 or more. So while "slower", they get more done per cycle, resulting in a faster overall execution. It's up to the compiler to properly structure the executable machine code to take maximum advantage of this layout and keep all execution units of the CPU busy at all times, as well as reduce disseparate memory accesses and so forth.

The intial compilers that are released with these machines do it, but not as well as they could. In fact, compiler writers are still trying to grasp the issues with pipelining on modern CPUs and their much lower number of execution units, and this is without utilizing special instructions that explicitly do non-conflicting operations at once. We're still years away from writing fully optimized compilers for contemporary CPUs. And while there's been a great deal of work done on VLIW already (prior to Itanium), there's even more yet to be done. A decade for a "good" compiler is probably optimistic.

You may be wondering, what's the point anyway? If VLIW is so damn hard, why bother? Just ramp up that clock speed and get more CPU power! Well, that's nice, but it doesn't work in reality. We're starting to bump up against physical limitations in CPU speeds. Electrons are not magical particles that travel instantaeously. They are limited to slightly under the speed of light, which means roughly 1 cm per nanosecond. This doesn't seem to be a big deal until you realize that a 2.0 GHz CPU means each clock cycle is 0.5 nanoseconds. So if you have to fetch an instruction or data from main memory, and that memory is a mere 5 cm away, under optimal conditions you've just sat around for 10 clock cycles waiting on that memory to be fetched. This is ignoring the fact that there's propogation delays, latch delays, and other things. So go ahead, pump that CPU up to 10 GHz and waste even more clock cycles waiting on data. That or redesign the entire thing, expect the compiler to do the work and properly feed you data and instructions such that you can do 10x as much in the same amount of time, and all with no wasted CPU instructions.

That's the theory at least.

Reality is that not only does the compiler have to properly organize the machine code, it also has to have some idea of what the code is doing to do so. Compile the code w/ profiling, run the code against a "realistic" data set, then recompile it again feeding it the profile data. Many compilers can do this now, but it's rarely done. Because it's hard to guess a "realistic" data set, it's hard to acquire the same, how you expect the code to be used and how it actually is used are rarely the same, and there's more development time involved in all of this. So most companies don't bother. And despite what I said above, 2.0 GHz still hasn't reached the point where the CPU is sitting on it's ass more than it's doing work. Until we start approaching that point there's little incentive to put in the R&D time necessary to switch to a new CPU archictecture.

And, of course, on top of all of the above is the issue that Joe Sixpack will invariably see 2 GHz as faster than 750 MHz no matter what. Have fun with that one.

Re:compilers

[RollingThunder]

I can code what I can because I stand on the compilers of giants?

Re:compilers

[maraist]

So basically you're saying that computers are magical radio-wave transcievers? Funny, I thought computers were based on capacitively switched [Bi]CMOS transistors. This means the "logical operation" travels at the speed of the capacitor charge / discharge times. After the ramp-up, ramp-down time (further delayed by theinnefficiencies of junctions), then the signal travels at the drift velocity of the electrons trapped within the conduction-band; significantly slower than a stream of free-flowing electrons, much less a single electron going full-tilt.

In fact, when electrons start going close to the speed of light within a silicon, there's typically an avalanching effect (utilized in zenor diodes). Channel break-down can easily occur under such situations (caused by relatively high voltages).

To my understanding, the single biggest speedup in the past several years was the introduction bipolar transistors into the CMOS frame-work. Bipolar are very fast (non-capacitively switched), have high current, high amplification, but are power-hogs and require difficult geometries to manufacture. My understanding of BiCMOS is that FET's are used everywhere, but when a FET needs to be charged quickly (or generally requires high current output), a bipolar device is attached on the output as an amplifier. You get the best of both worlds (with the possible exception of the geometry limitations).

Wiring obviously was an issue because new copper based CPUs can run cooler and faster.

I only have an undergraduate understanding of the processes, but the simple point is that there are paracitics all throughout the architecture, and we're discovering efficincies everyday which provide percentage increases in overall performance. Thus it's not the speed, but the sophistication of the design.

There's lots of work going into light-based computing, but I don't think this will ever win out because they're plagued with even bigger interconnect problems and thus paracitics.

-Michael

Re:compilers

[maraist]

I thought that this was the crueso. They don't exactly offer performance boosts as a selling point.

If you have a target CPU, and a target OS, then you can make highly optimized compilations.. Just look at Quake. They've always optimized their code for the intel line (P2, P3 being best lately). A VM can't make several types of optimizations.. Once you're down to assembly, things like load-order are important (and x86 is horrible at dissassembly-based optimizations; no hints, and few registers).

For consumers, most only purchase on evolutionary leaps in performance; thus 50% increase or more (going from 486-66 -> P100 -> P2-300 -> p4 2,000). Theoretical gains of 5-20% based on timing isn't impressive to this realm. And for server-class software, they're going to demand hard-compiled and optimized code, which a VM could never achieve.

-Michael

Re:compilers

[blair1q]

Itanium has been around for nearly two years now.

The compilers aren't completely done, and they'll never be perfect (cf. any compiler for x86; the bug lists are shocking), but they're good enough for a high-end server chip where the integration engineers know how to talk to the compiler vendor to get bugfixes turned in time for the next rollout compilation.

--Blair

Re:compilers

[MicroBerto]

Hehehehe

What's after Itanium? That's easy

[wrinkledshirt]

"Anadium"

That's probably only funny to chem majors.

Okay, maybe not even chem majors.

Re:What's after Itanium? That's easy

[freeweed]

"Adamantium"

There, we'll make it funny only to comic book fans :)

I don't think we need to worry just yet

[Indras]

Think for a minute how long we've been using 32-bit processors. If (and when) 64-bit becomes mainstream, I imagine it will be around for a LONG time, as it becomes standardized and slowly takes over a majority of the market. Also, we'll have the other contenders butting in with equivalent and cheaper options, like Cyrix (tried) and AMD (did).

Just because Intel will pave the way for mainstream 64-bit processors using the Itanium doesn't mean it will monopolize the market until it comes out with a 128-bit processor. No matter what, it will probably be years from now before we have to worry.

Re:I don't think we need to worry just yet

[SoftwareJanitor]

The only problem with AMD's 64 bit line is that it isn't going to be compatible with the Itanium. That is both good and bad. Good in that it is an alternative, bad in that it is going to cause a lot of confusion.

I think a lot of people are too overconfident that Itanium is going to be successful, let alone quickly. It is going to require a lot of changes to software in order to take advantage of it because it isn't just a 64 bit x86, it is a whole new architecture, one more closely related to HP PA-RISC than x86. It also may not do a very good job of running existing 32 bit code, which could slow down its acceptance, particularly in desktop systems. The last time Intel made a big push (with the i432) to create a whole new non-x86 processor family, it was less than successful. Although to be fair, the i432 was a radically different proposition and the Itanium with its more proven PA-RISC roots looks a lot more sound.

AMD's Hammer architecture, on the other hand, is more conservative, being a x86 family processor extended to 64 bit. It should require less modifications to existing software to take advantage of it, although an argument could be made that it won't have as much advantage to take having more legacy issues with the aging x86 architecture. It also may perform a lot better on existing 32 bit code than Itanium. And if AMD's track history holds true, it will probably be significantly less expensive than the Itanium.

A lot of whether it is Intel or AMD that paves the way for 64 bit mainstream CPUs will probably have to do with which of them is the first one that offers a price attractive product that runs existing 32 bit software well while being marketable as a 64 bit chip. Unfortunately for AMD, the marketable part is, as always going to be tough. While AMD has been hugely successful in "white box" sales where customers can choose their CPU, they've had a much more difficult time penetrating the big name PC markets, particularly in higher end systems. This despite the fact that in many cases an Athlon or Duron would offer a better performance than a PIII or P4 at a better price.

Re:I don't think we need to worry just yet

[stripes]

Think for a minute how long we've been using 32-bit processors. If (and when) 64-bit becomes mainstream, I imagine it will be around for a LONG time, as it becomes standardized and slowly takes over a majority of the market. Also, we'll have the other contenders butting in with equivalent and cheaper options, like Cyrix (tried) and AMD (did).

Not really. A good rule of thumb is the address space increases about a bit a year. Partly because the amount of memory goes up (as price goes down) so anyone looking at size/speed trade offs, or even size ease of coding trade offs go more towards size.

I'm typing this on a laptop with 1Gbyte of RAM. In two years it'll be 4G (well whatever laptop I have then). While you can address more then 4G with only 32 bits, plus some hackery it is just a big pain.

If this trend continues to be true we have about 32 years of 64 bit address space to look forward to...

...of corse that discounts the use of anything like a capability based OS which needs much much larger address spaces (I think the AS/400 has a 256 bit address space).

Re:I don't think we need to worry just yet

[stripes]

I think a lot of people are too overconfident that Itanium is going to be successful, let alone quickly. It is going to require a lot of changes to software in order to take advantage of it because it isn't just a 64 bit x86, it is a whole new architecture, one more closely related to HP PA-RISC than x86.

It isn't really any closer to the HP-PA then it is to the x86 (unless you count "only" having 16 times as many GP registers rather then 64 times as many...).

[...] The last time Intel made a big push (with the i432) to create a whole new non-x86 processor family, it was less than successful.

Intel has done a lot of non-x86 CPUs since the i432 (and "less than successful" is an understatement). They did the i960 which while I didn't like was a real winner in the embedded (esp. military) market. They also did the i960 which I did like and was only modestly successful, I think mostly because they end of lifed it so fast (it was doing fairly well in the market until they EOLed it). They are also doing ARMs, but they didn't design the ISA there.

Re:I don't think we need to worry just yet

[SoftwareJanitor]

It isn't really any closer to the HP-PA then it is to the x86 (unless you count "only" having 16 times as many GP registers rather then 64 times as many...).

I dunno, the Itanium looks more like a HP-PA relative than part of the x86 family to me... But I mainly say that from the standpoint that HP's PA people were involved in the design as well, and the Itanium looks like it has a fair amount of influence from the PA. The x86 compatibility in the Itanium looks like more of an add-on than it being a part of the x86 family.

Intel has done a lot of non-x86 CPUs since the i432 (and "less than successful" is an understatement).

I don't know of any that were in the general purpose CPU market that were successful though.

They did the i960 which while I didn't like was a real winner in the embedded (esp. military) market. They also did the i960 which I did like and was only modestly successful, I think mostly because they end of lifed it so fast (it was doing fairly well in the market until they EOLed it).

The i960 was more targeted towards the embedded controller market, or at least that is where it had most of its success. I don't know of any general purpose computer that used it as its primary CPU.

They are also doing ARMs, but they didn't design the ISA there.

Well, as they didn't design the ARM, I don't really count it. The ARM has been used in some general purpose desktops and servers, but especially in the US has had limited commercial success outside of niche markets like handhelds.

Re:I don't think we need to worry just yet

[stripes]

But I mainly say that from the standpoint that HP's PA people were involved in the design as well, and the Itanium looks like it has a fair amount of influence from the PA.

It doesn't seem to be any more like an HP-PA then anything else. Actually I would say it is more like a Multiflow then anything else. Ok, I mean yes it is a three register ISA like the HP-PA (but also like most other modern CPUs!), but almost none of the HP-PA addressing modes, and very few of the HP-PA quirks (the bit mangling instructions may have been from there, but I'm not sure, and they are a pretty small part of the IA64 anyway).

I don't know of any that were in the general purpose CPU market that were successful though.

OK, I didn't realize that you were concerned with how they were marketed rather then designed. The i860 was in a modestly successful Unix machine in Japan, and a woefully under successful one in the USA (same box, the OkiData 8500 and one or two other models). It had better SPECfp numbers then most of the other Unix boxes at the time too (I think it predated the Alpha, but can't remember, it didn't beat the RS/6000, but it beat the DECStation and was way way way faster then the SPARCs of the era). Hmmmm, come to think of it maybe it was (a little) after the Alpha, it was slower then the Alpha.

The data books definitely read like it was intended to be a GP CPU. Mine are packed so I can't say for sure if they used the phrase "General Purpose CPU" or not though.

The i960 was more targeted towards the embedded controller market, or at least that is where it had most of its success. I don't know of any general purpose computer that used it as its primary CPU

Me neither, they could have though, some of them had MMUs, but none of the fast ones did. Closest to being in a GP computer was being the controller on the I2O bus (and I think it might have made a decent PDA CPU though, but the StrongARM makes a better one, as does the CPU32+ both for different reasons).

Re:Single track

[grammar nazi]

The itanium already has not-an-instruction bits attached to each operation. Perhaps, they need not-a-thread bits for processes and not-a-bug bits for features.

Next?

[jgerman]

In a word, quantum. Or maybe that's two words, actually it might only be a word when you're looking directly at it.

Re:Next?

[jgerman]

Yep and we'll have be careful not to jostle the keyboard, otherwise we won't be able to find the home keys.

Re:Next?

[emmons]

Not really. Quantum computers aren't very good at adding, subtracting or a lot of other things that most programmers find come in handy from time to time these days. Boolean logic will still be prevalent for a VERY long time to come. It may not happen on silicon for that long, however.

There's a lot of information in this thread: http://slashdot.org/article.pl?sid=02/02/10/001620 4&mode=thread,
specifically, this post: http://slashdot.org/comments.pl?sid=27736&cid=2980 985.

Logically, it should be Anadium

[PD]

Itanium is Titanium without the T, so Anadium is Vanadium without the V.

I can't wait until they get to Hassium. They could name their chip Assium!

Itanium vs. Hammer vs. All Others.

[Talonius]

AMD's newest chip is supposedly fairly remarkable (don't have specifics, see Tom's Hardware's search engine). What about the Crusoe? VIA's purchase of (I believe) the M3? I wouldn't look at companies that are currently in the business only - I would tend to look at companies that might move into the business, either via investment, startup, or outright purchase.

I'm not too worried about Itaniums, and I don't see them becoming prevalent for quite a while. While the Pentium II, III, and IV moved through the marketplace fairly rapidly they all offered compatibility at some level. If I recall correctly 32 bit programs that are not rewritten for 64 bit run SLOWER on the Itanium than they do the equivalent Pentium line.

In essence consider this: it's like a brand new operating system attempting to break into the monopoly that Microsoft has. (Parallels drawn out of necessity.) While it may be better, faster, superior in every way it doesn't have 20+ years of legacy code behind it - and that will end up being what drags it down.

Only time will tell. Remember the Pentium Pros..

Talonius

Re:Itanium vs. Hammer vs. All Others.

[Skirwan]

If I recall correctly 32 bit programs that are not rewritten for 64 bit run SLOWER on the Itanium than they do the equivalent Pentium line.

When Apple transitioned from the M68K line to the PPC, they were in the same situation - 68K code would run faster on a 40Mhz 68040 than on a 40Mhz PPC 601. The reason consumers didn't mind was that the the PPC 601 started at 60Mhz (approximately the break-even point to the emulation layer), and (to the end user) didn't cost significantly more.

Until Intel gets the Itanium cost down to the point where they run 32-bit code at equivalent speed to a Pentium at the same cost, Itanium probably isn't ready for the consumer market.

--
Damn the Emperor!

Re:Itanium vs. Hammer vs. All Others.

[Mr Z]

When Apple transitioned from the M68K line to the PPC, they were in the same situation - 68K code would run faster on a 40Mhz 68040 than on a 40Mhz PPC 601. The reason consumers didn't mind was that the the PPC 601 started at 60Mhz (approximately the break-even point to the emulation layer), and (to the end user) didn't cost significantly more.

While that's a valid point, it also bears pointing out that Pentium IV is at 2200 MHz whereas Itanium is at 800MHz -- about 1/3rd the clock speed. That ratio is going to remain for awhile too -- McKinley will come out at 1000 MHz, while Pentium IV continues its mad march toward 3000MHz and beyond. You acknowledge this fact implicitly with your next statement (re: Itanium not viable until approx same speed at approx same cost), but I felt it'd be interesting to point out just how large a gap there is.

These ratios spell doom for hardware-level emulation of the Pentium on the Itanium. Unless Intel has some serious magic, having a 100% cycle-for-cycle perfect emulation of the Pentium III or even Pentium IV on the Itanium die will never run better than 1/3rd the speed of the real thing, since the fundamental clock rate is so far off. The only real way to get close is to do a software-level translation and get a boost from scheduling for the native hardware.

It's interesting to note, BTW, that HP's Dynamo project does a software translation of PA-8000 code targeting (guess what) a PA-8000 CPU, and rather than slowing things down, it actually gets 20% speedups! Ars Technica also did a piece on this. Perhaps that's why HP doesn't have hardware-level translation from PA-RISC to Itanium on the die like Intel does -- they (HP) are in a better position to just translate the PA-RISC code to IA-64 when needed. (Also, in the UNIX world, it's just simply less necessary.)

--Joe

Re:Itanium vs. Hammer vs. All Others.

[Myxorg]

PPC 601 started at 60Mhz (approximately the break-even point to the emulation layer)

Actually the break even point wasn't reached until about 100 Mhz or so, not sure. But I do remember when the first ppc came out they were definatly slower than the old 040's. Still don't know how Apple pulled that one off (selling new computers that were essentially slower than previous models)

Re:Itanium vs. Hammer vs. All Others.

[Locutus]

> Only time will tell. Remember the Pentium Pros

the ONLY reason the Pentium Pro didn't catch on was because Microsoft released a 16bit OS and told everyone it was a 32bit one ( Windows 95 ).

SCO Unix, OS/2, and to some degree Windows NT ran quite a bit faster on the 32bit optimized PPro when compared with the same clocked Pentium.

Because of Microsofts great PR, even Intel was caught off guard and scrambled out a hack called MMX to give the appearance of progress in the CPU market. While the MMX based Pentiums were getting press/air time, Intel was hacking at the Pentium Pro core to get it to run THE 16bit OS (Windows) faster. That was the Pentium II.

IBM did some speed tests of OS/2 on the PPro and in some cases they saw a 100% speed increase on the 32bit optimized PPro.

This reminds me of the 7degrees from Kevin Bacon reference. It seems that many failures in the computer industry are only about 3degrees from Microsoft. And never is the failure do to competition but more likely, marketing and market control. IMHO.

The PPro was a darn good CPU. It finally took 32bit-ness seriously though about 10 years after the 32bit i86386 was released. As much as I like the simplicity of RISC, Intel will never get the Titanicium off the ground and AMD/Hammer will force Intel to follow their lead with an extension to the i86 instruction set into 64bit land.
IMHO.

LoB

Re:Itanium vs. Hammer vs. All Others.

[Ryan Amos]

Well, eventually, that will happen, without a doubt. Moore's law pretty much assures it, in fact. The big question mark is whether or not the Itanium can match the price/performance of the Pentium like before someone else does. Seeing as Itanium is currently running at clock speeds around 800 mhz when it would need about 1600 to be equivalent to a P4, even Intel's not betting on this (hence the Yamhill) and they're seemingly relegating the Itanium to high-end servers (to take over where the Suns and Alphas left off) which seems to be where they're best suited. At least for now, it looks like the x86-64 (Hammer/Yamhill) is the platform of the future, and Itanium will be just another expensive non-consumer platform.

The luxury Apple had in this situation was control of the operating system, which Intel doesn't have. Ironically, Apple will also be moving to a 64-bit architecture within the year (conservative rumors say Q3/Q4 2002.) The transition is supposed to go very smoothly, as developers are being told to prepare their programs with the 64-bit OS X libs and OS X-64-bit is being developed in concurrence with the 32 bit version. FAT binaries helped immenseley in the 68k-PPC transition, and probably will again for the G4-G5 transition.

Though honestly, if Microsoft gets what they want with the entire .NET plan (not the framework, the entire plan) then architecture will become largely irrelevant. In any case, I doubt that many people will need frequent execution of their old 32-bit apps much more than 2 years after any sort of major switch happens. It happened with Mac OS, and it'll happen with Windows. Linux is irrelevant here, as most Linux software can be easily patched and recompiled.

Re:Itanium vs. Hammer vs. All Others.

[boopus]

While 800/2200MHz is a large difference, you fail to mention something that everyone here should know by now, that clock speed does not equal performance.

Clock speed does not equal performance. This is a fact of life, especialy with 20 stage pipelines and the like. AMD and Apply have been trying to teach this to the world, and on the surface most geeks understand, but they don't beleive it in their hearts.

Now, I'm not saying that the PIV won't be faster than Itanium for a good while here, and I honestly have no idea if it will be or not. We just need to stop using Mhz as our comparisons unless we're comparing the same chip.

Re:Itanium vs. Hammer vs. All Others.

[sean23007]

While it may be better, faster, superior in every way it doesn't have 20+ years of legacy code behind it

If you are drawing a comparison between Windows and the Unices, you may find it interesting that Unix came before Windows and thus has more legacy code than Windows does. The reason for Windows' success is not all their old, crappy code, but because of their excellent marketing department and their vision of what the consumer wants.

Re:Itanium vs. Hammer vs. All Others.

[Chris Burke]

Um... Who would ever do a cycle-for-cycle perfect emulation of anything, when the point is just to mimic the -functionality-? It'd be stupid. Computer architechts are the only ones who need that accuracy -- every one else settles for just functional correctness.

Will IA-64 processors ever be able to -functionally- run 32-bit code at a speed comparable to a Pentium?

The answer is: Only if Intel stops developing their 32 bit chips. Then, sometime in the future, an IA-64 chip will run 32-bit code as well as -todays- P4.

But here's the problem: The moment Intel does that (stop developing 32 bit chips) they hand the desktop market to AMD. AMD is already positioning itself as a way to get 64-bit code capabilities without needing to sacrifice 32 bit performance. In that period between when there are no new Intel 32-bit chips and when Itanium gets fast enough to catch up there will be one option -- AMD. So obviously Intel isn't going to do that.

Their other option is to switch everyone over to IA-64, even the consumer market... But there really isn't any incentive for anyone to do so. Unlike the 16->32 bit transition, 32->64 really holds no benefit for the average consumer. 640K may not be enough for everyone, but 4GB will be enough for the consumer until the issue has already been decided.

Re:Itanium vs. Hammer vs. All Others.

[swb]

I thought that '040 was all-around competitive with PPC until much later in the Apple product cycle, primarily due to the dependence on 68k code in the OS.

This was from a work-and-feel mindset. Once everything was all PPC based, '040 ceased to be competitive, but it wasn't until far too late in the product cycle.

I always thought Apple would have been in a better place if they had focused a lot more on making MacOS available in 100% PPC version at the time they released PPC machines.

If using a PPC had actually been much faster than an '040, PPC would have had more momentum, especially as the clocks ramped up. As it was, it didn't seem like you got ahead with PPC until way later in the game.

Re:Itanium vs. Hammer vs. All Others.

[Phroggy]

FAT binaries helped immenseley in the 68k-PPC transition, and probably will again for the G4-G5 transition.

From what I understand, FAT binaries shouldn't be necessary, since the PowerPC instruction set is already completely 64-bit clean (it was designed that way from the beginning, and IBM has made 64-bit versions such as the 620 already). So, the same binary executable should work just fine on either a G4 or a G5.

It's entirely possible that I don't know what I'm talking about; feel free to correct me.

Re:Itanium vs. Hammer vs. All Others.

[Have Blue]

They had a PPC version of Photoshop from day 1.

Re:Itanium vs. Hammer vs. All Others.

[homer_ca]

"the ONLY reason the Pentium Pro didn't catch on was because Microsoft released a 16bit OS and told everyone it"

I wouldn't say ONLY. There was also the slight problem of the double chip package (separate cache and cpu dies mounted on one substrate) being horrendously expensive to produce. Looks like Itanium will have thesame problem.

Re:Itanium vs. Hammer vs. All Others.

[Locutus]

The PentiumPro had a double chip package? I know the Pentium II was a double chip package and was also that dumb SLOT format but the PPro? From what I remember, Intel said the PPro was too expensive to make with the cache onboard so they HAD to change from a socket to the slot that accomodated the CPU and cache on a PCB.

IMO, Intel came up with the SLOT format to put a major dent in the clone chipset manufacturing that was going on. When Intel released the SLOT CPU, they only licensed it out to very very few companies. Intel motherboards ruled the day for quite some time back then.

All just more smoke and mirrors and no real progress. IMHO.

The Itanicium having the double chip package is most likely because preformance would suck without a huge cache. They needed it if anyone was to even LOOK at the chip. A huge cache onboard would kill the yeild/wafer count and made the chip even MORE expensive to make. IMHO.

LoB

Re:Itanium vs. Hammer vs. All Others.

[ksheff]

gIDE (which seems to have disappeared) and the like. Buy kylix. That's the root to acceptence. Make things real people can develop with.

According to a recent GNotices, gIDE and Anjuta are going to be merged for Gnome 2. So if 'real people' can't develop with vi/Emacs, what do you call those that are productive with those tools? Personally, I can't stand GUI IDEs.

Re:Itanium vs. Hammer vs. All Others.

[sjames]

Regressing from visual studio to emacs is a painful wrenching experience. Regressing to vi is simply intollerable.

Actually, of the many IDEs, etc I much PREFER vi, make, etc over all of them. Screen is also helpful. The thing I most like about using X for development is.....more text windows!

As for Windows, I despise the API, and the big catchall event loop, and the 'wizards' with their exercise in obfuscated code generation. It seems a lot more intuitive to connect an event to a function (such as in GTK).

I do hope the various Linux IDE systems work out since some genuinly prefer that, but give vi etc. 1 month and see if it's not actually much more efficient.

Re:Itanium vs. Hammer vs. All Others.

[homer_ca]

The PPro was more like 2 dies in one chip. The CPU and cache were mounted in one ceramic package which had to be tested together. If either one was bad the whole package was rejected. This process wasn't very efficient with chip yield, but it was probably still better than trying to get a CPU core and 512K cache on one die at .35 microns. The P2 had the CPU + SRAM cache chips on a PCB.

Re:Itanium vs. Hammer vs. All Others.

[Locutus]

Too bad they couldn't have tested them before glueing them together in the carrier.

Thanks for the explaination and I'll be sure to take one of my 2 180MHz PPro's apart when/if they fail. They've been running like clockwork for years in a SMP system running Linux ( so it's still OK to use ).

LoB

Re:Itanium vs. Hammer vs. All Others.

[sjames]

The event-based programming model isn't impossible in Windows either; in fact it's how visual basic is done.

Absolutely. GTK has an event loop as well, it's just internal to the library. The last time I saw MFC, you could see the event loop, and it auto-generated calls to methods from within the loop. The programmer was supposed to hunt down the method definitions and fill in everything between the comments with real code. Analysing someone elses code can turn into a real nightmare!

In general, event driven and procedural approaches can be mapped one to one. Should is another matter!

I much prefer (and some others don't!) GTK's approach. The last time I saw VB, it seemed to tie visual appearance and method calls too closely together (obfuscating the opportunity to assign a generic handler to multiple objects).

Recurring problem

[colmore]

This seems to be a recurring problem in a number of technology based industries. Once you get to a certain lever of high-tech, only the (very) big boys can even compete.

So here's the question: how do you keep competition alive when an initial investment costs in the billions of dollars. For any company less than Intel sized, a single bad product cycle spells complete doom. That's no kind of market to be in.

Also, wasn't this inevitable. There are a few Beowulf jokes being posted, but that's really what's going on. Increasingly high performance tasks (Google, render farms etc. etc. etc.) are using massive arrays of low-power CPUs. It costs a lot of money to develop big iron chips, and if people aren't buying them then there's no point in investing that much money.

What I'm worried about are the isolated markets that still require massively powerful, low processor number architectures. Not everything splits into nice Distributed.net packages.

Re:Recurring problem

[Waffle Iron]

Speaking of badass mainframe processors, I was an intern at IBM in the mid 80's. The top-of-the-line mainframes used a central processor comprised of about 100 custom ECL chips mounted on a 4-inch-square 100-layer ceramic substrate.

The whole thing was cased in a shiny metal module. Each chip had its own sping-loaded heat slug that transferred heat to the cooling liquid sent through the module's plumbing. (100 ECL chips == major kilowattage)

They told me each CPU cost about $50,000. On a factory tour, I saw an entire pallette of these sitting on the floor, kind of like gold at Fort Knox.

These things may not perform like today's chips, but they gave meaning to the term "Big Iron"

Re:Recurring problem

[joib]

Yes, very true. As I see it, there are currently 3 high end cpu architectures with a future, sparc, ibm:s power and IA-64. Producing (and designing) these chips gets more expensive all the time, and the upcoming extreme ultraviolet lithography is again going to cost even more. So barring any big surprises, these 3 companies are those which I'd bet on to have the funds to play the game till the end of the current silicon technology. Hopefully whatever comes after silicon will have a cheaper entry-to-market so there can be lots of new innovative companies driving innovation and competition forward. And I think it might be very feasible. Look at all the nanotechnology startups, research on molecular transistors, self-assembling systems and of course quantum computing. Very interresting stuff, albeit it will probably be >10 years before anything like that gets out of the research labs.

Regarding supercomputing, you are quite correct in that there is little incentive to develop new vector supercomputers, taking into account increasing costs of design and fabbing the cpu:s. Cray:s upcoming SV2 seems to be an exception, though. Also keep in mind that most supercomputers, like the Cray T3E and the IBM SP, are just commodity cpu:s connected with a bad-ass proprietary interconnect. So essentially, they are just beowulf clusters with a superfast and low-latency interconnect. They are programmed through MPI, the same thing used for beowulf.

While there are few scientifically interesting applications which require essentially no communication between the nodes, like seti@home and distributed.net (if you want to do something scientifically useful with your spare cycles, check out folding@home), most get by with a quite modest amount of communication between the nodes.

What I personally see as interresting in the supercomputing area in the next few years is not the maximum flops of the DOE:s newest lets-simulate-nuclear-explosions-big-ass-fastest-s upercomputer-in-the-world-thingy, but that the increasing affordability of commodity clustering will allow projects with somewhat more down-to-earth budjets to run some really interresting simulations. Also commodity clustering is improving. Things like Infiniband and whatever they are called will probably offer higher bandwidth/lower latency than the current Myrinet / Gigabit ethernet stuff. Also the improved SMP scalability of Linux 2.4 will perhaps make it economically feasible to have, say, 4-cpu nodes in the cluster instead of the current practice of 1-cpu nodes. The downside of this is of course that applications must be able to take advantage of this, either through directly using a combination of MPI and threads/SHM/whatever or through using frameworks like Petsc or POOMA.

As a final note, remember that the tools (i.e. supercomputers) and the way we solve the problems at hand are not developed in a vacuum. The numerical methods that are developed today certainly place a greater emphasis on solving the problem via scaling through clustering than, say, 20 years ago when vector supercomputers where in the same price class as the parallell ones.

According to Mr. Pearson...

[bobetov]

Quantum computing, 2007.

Bet on it. ;)

SPARC is dead?

[bconway]

That's news to me. I could swear a friend of mine just jumped in on the UltraSPARC 4 project.

Re:SPARC is dead?

[Anonymous Coward]

Actually, I was just transferred to the UltraSPARC 4 project at Sun in Burlington, MA. I don't know of the official release date, though I've heard rumors of early 2003. I'm amazed at the quality of FUD in this "article" and that it actually made it to the front page of Slashdot.

Re:SPARC is dead?

[supabeast!]

Go take a look at Sun's sales numbers for 2001 and Q1 2002. Given that they have X86 machines ready to hit the market in June, the chances of Sun being able to convince already reluctant buyers that Sparc systems are still worth the money are rather low, especially now that big iron is being replaced with clusters of cheap systems. Sparc may not be dead, but Sparc's future as a commodity item is dim at best.

Re:SPARC is dead?

[elmegil]

When has SPARC ever been a commodity item? At least on the same level as the Intel processors? NEVER. When has that been necessary for SPARC to be a strong competitor in the marketplace? NEVER. Get over yourself. SPARC is not dead. It won't be dead until Sun itself is dead.

Re:SPARC is dead?

[HiredMan]

the chances of Sun being able to convince already reluctant buyers that Sparc systems are still worth the money are rather low, especially now that big iron is being replaced with clusters of cheap systems.

But this argument works against ALL servers and 'big iron' makers not just Sun. You could replace "Sun" with "Intel" and "Sparc" with "Itanium" and the argument remains the same.
What you're arguing here is not that Sparc is dead but that big iron is dead. And I think you're wrong.

=tkk

PS UltraSparc/// is stands up well against Itanium and probably even against McKinley. And who has the history, and client base, and code base and, and, and.... Sun does.

Re:SPARC is dead?

[Multics]

But this argument works against ALL servers and 'big iron' makers not just Sun.

That is the poster's point exactly. ALL big iron people will be irrelevant if Itanium based systems make it deep into the volume curve and hence are very inexpensive. I don't see how Sun will be able to keep up with the economies that Intel brings to the party. Obviously Compaq and HP thought similarly and that's why they're out of the CPU business.

So let's say that that happens... where will the marketplace go then?

-- Multics

Re:SPARC is dead?

[HiredMan]

That is the poster's point exactly. ALL big iron people will be irrelevant if Itanium based systems make it deep into the volume curve and hence are very inexpensive.

What?!? The IA-64 project IS big iron. The McKinley cpus will cost several hundred dollars a piece to produce and will cost several thousand each in boxes that are at least 5 figures - and I'm sure Intel's hoping for 6. They are not consumer chips.

I don't see how Sun will be able to keep up with the economies that Intel brings to the party.

Again - What?!? The P4 outperforms the Itanium. No seriously, you can buy a personal computer that outperforms it RIGHT NOW. Even if the McKinley then out-performs the current P4s the roadmap has McKinley's getting to only 1.6Ghz through 2003. The P4s that will be shipping then will again outstrip the McKinley (probably) before the second gen ships. Also the IA-64 does not run x86 code! (At least any speed you ever WANT to run your M$ Word on.) It runs IA-64 specially compiled code.

The IA-64 is NOT a consumer effort. This whole discussion is about "what's the next big iron processor" going to be? Big Iron has it's place and always will - a 32 processor Power4/UltraSparc3Cu server machine swings some serious frickin' power. The jobs that this machine can do will NOT be replaced by any desktop machine. At least not for a few years. ;)

=tkk

SPARC's death *greatly exagerated*

[AtariDatacenter]

Having recently participated in an NDA from Sun regarding the SPARC processor (and even with the knowledge I had walking into the meeting), SPARC is not dead or dying. In fact, I'd say that Sun squarely recognizes it as a strength. Their competition (HP for example), however, is wishing they didn't knife their baby.

As far as money to go another round, remember, Sun doesn't fab CPUs. What Sun does is design them, and they turn it over to Texas Instruments for production. And TI has their own reasons to keep up-to-date with the latest production technologies, so Sun doesn't eat that cost.

BTW: I really wish that I could talk about the SPARC presentation. I liked it a whole lot better than the NDA I attended with HP talking about their Itanic future.

Re:SPARC's death *greatly exagerated*

[AtariDatacenter]

Sun's strength isn't in the performance of its servers. You don't buy a Sun because you want the fastest thing out there. You buy it for the support, reliability, software base, and probably a number of other things. As long as Sun's processing performance is "on par" with competitors, it isn't going to be a liability.

When I talk with management about servers, they don't ask me which one has the fastest CPUs. They've got a "short list" of hardware vendors (IBM/Sun, then further down HP/NT).

Re:SPARC's death *greatly exagerated*

[schlach]

BTW: I really wish that I could talk about the SPARC presentation. I liked it a whole lot better than the NDA I attended with HP talking about their Itanic future.

Itanic. That's really funny.

• Itanic rituals and sacrifices
• Itan worshippers

then I ran out of ideas and had to search for 'satanic' on google

• Itanic Sysadmins
• The First Itanic Church
• The Itanic Verses
• Itanic Hampster Dance

...

(this post is obviously the set-up. now I just need someone to supply the punchline)

Re:SPARC's death *greatly exagerated*

[elmegil]

I like the fact that Itanic rhymes with Titanic.

Re:SPARC's death *greatly exagerated*

[elmegil]

Sun's never had a Fab, period. Just because Intel has the power of brute force (maximal transisters per square unit) doesn't mean their tech is superior. It's the architecture of the processor that makes it superior, not the bells & whistles at the Fab.

Re:SPARC's death *greatly exagerated*

[wysoft]

Sun doesn't have to worry about raw CPU power because their machines are not designed to write Word documents or play a game of Wolfenstein. Compare a Sun machine to almost any PC out there and it will smash the PC's memory and system bus bandwidth. For the kind of tasks that Sun machines usually accomplish, that is much more important when it comes to the throughput that people buy Sun machines for.

Hell, most PCs don't even have enough PCI bandwidth to fully saturate a gigabit ethernet connection unless you have a totally bare PCI bus or a system which provides each PCI slot with its own dedicated bus, as most Sun PCI systems do.

Let's not even compare the stability, scalability, and worksmanship of PC and Sun hardware. That would just be unfair to 99% of the "business" PC workstations and servers on the market.

Re:SPARC's death *greatly exagerated*

[rcs1000]

The problem with discussions of Intel vs every other chip maker is they ignore the extraordinary differences in scale between the players.

Let's compare: Sun is a company that produces operating systems (Solaris), computers, CPUs, motherboards, and a host of peripherals. (Plus it has to invent Java, J2EE, etc.) It's R&D budget was $2.0bn in 2001.

Intel is 95% CPUs. It spent $3.8bn on R&D in 2001.

Intel has the world's most productive fabs. It's capex budget is so huge, it can order the lithograohy companies and the like to build to order inside its factories. Result, it's yields are 25% better at start; and still 10-12% better after 6-9 months.

It is incredibly difficult for anyone to keep up with the Intel machine. I wish it weren't so; but it is.

*r

Re:SPARC's death *greatly exagerated*

[Paul Jakma]

indeed... and he'd be $8k better off than you.

heck, he can buy /2/ $2k AMD boxes and be redundant, and still be $6k better off than you.

presuming revenue is not so huge that $6k is a miniscule sum (reasonable assumption if his business uses only 2 AMD boxes, and yours uses only one SPARC), then he's /way/ ahead of you.

given these are the only 'facts' to hand, i know which one i'd invest in if i was a shareholder.

Itanium

[crumbz]

Given the tremendous capital requirements in building a state of the art fab along with the incredible amount of enginnering man-hours required to leap to the next level, I think we are seeing a situation similar to the one for airliners: Airbus or Boeing. They are the only two that matter because the cost of entry into the airliner market is so prohibitive. This does not necessarily apply to Microsoft and it's OS monopoly as the Linux community has illustrated. Mindshare and marketshare are not always linked.

I have hopes for Intel producing the worlds best microprocessors as that would benefit s all. Simply advocating a move to Itanium for marketing reasons or to meet revenue targets does a disservice to the computer industry.

Then again, they are in business to make $$$....

The newest chip will be called...

[RobL3]

The Unobtainium

It's release will follow the distribution pattern established by Transmeta.

Itanium will be Hammered

[Brian Stretch]

The huge die size of the Itanium and its upcoming successor make the chip far more expensive than the Pentium series, so I would not expect Itanium machines for $2K. So far, the CPUs alone are several $thousand. I also haven't seen where its performence is that impressive. x86 code performence, since its emulated, is poor. Recompile or else. Intel has sold, what 500 Itanium CPUs?

The upcoming AMD Hammer series, OTOH, is supposed to be about 30% faster clock-to-clock than the current Athlon XP series (which is considerably faster clock-to-clock than the Intel P4) and start at 2GHz. Sun's recent announcement of Linux x86 platform support, with details to come midyear, suggests that they'll be moving to the Hammer (to ship Q4). Sun would certainly love to take a swipe at Intel, and Sun has made positive comments about AMD's x86-64 Hammer architecture.

Speculation: Intel gets Hammered in the second half of this year.

Re:Itanium will be Hammered

[sheldon]

The huge die size of the Itanium and its upcoming successor make the chip far more expensive than the Pentium...

This is a short term problem. I mean go back and look at the original Pentium 60 and where that went... Given time with the new architecture they will work on these issues and improve it. The original P60 was quite a heat generator compared to subsequent models.

I also haven't seen where its performence is that impressive.

Take a look at the floating point scores. The few benchmarks I have seen have it blowing away most of the competing RISC processors.

x86 code performence, since its emulated, is poor. Recompile or else.

The same was true when 64 bit RISC processors were first introduced. I remember getting our first DEC Alpha back in '94 and having to rewrite large pieces of our research apps to take advantage of it. Or look at the Mac when the PowerPC was introduced. The same will happen with the Intel world of applications. Given time, things will evolve and you won't look back.

The upcoming AMD Hammer series

Well we'll have to see. AMD might be able to cut themselves a nice niche in the desktop market. But Intel may eventually rule the servers where recompiling your code is not an issue if you can achieve greater performance.

Speculation: Intel gets Hammered in the second half of this year.

I wouldn't take that bet. My speculation, especially in light of your other comments with regards to Sun is that McNealy ends up on the Today show complaining about liquidity and having to sell the cottage in Aspen. Sun is between a rock and a hard place, yes they would love to take a swipe at both Intel and Microsoft. But by backing away from Sparc and Solaris and adopting Linux and AMD they do more harm to themselves than the Wintel monopoly.

On the other hand, by doing the "smart" thing, and focusing on Solaris and Sparc they also harm themselves. This is the Innovator's Dilemna that guy from Harvard wrote about and that's why I say rock and hard place. So we'll see how Sun plays this.

I still don't see AMD as a serious threat to Intel as they don't have the reputation for reliability to be widely adopted by business. Mitigating risk is nearly all about reputation.

Just...

[JonWan]

name it P-51 and use the 'nickname' Mustang.

No, no, and no.

[hotsauce]

No, Itanium will not become commodity as soon as you foresee because compilers and software do not exist to make good use of it (some argue nothing can make good use of it [derogatory]).

No, Intel has not killed the competition. AMD is alive and well. The PowerPC family is on the verge of The Next Big Thing (G5). And the reports of Sparc's demise have been greatly exaggerated.

No, other vendors are not irrelevant. Hitachi makes killer chips for big iron, and looks set to increase that trend. If anything, the CPU market is looking less and less like a monopoly than before.

*cough* PoerPC *cough*

[S-prime]

Now that the G4 has finally gotten past the 1GHz mark, and Apple has a brand spanking new Unix based OS running on it(and if you don't like it you can run others), this opens a whole new choice for the researcher looking for a new platform.

Re:*cough* PoerPC *cough*

[TeknoHog]

I'm drooling for an open-design PPC system. Something as generic and clonable as the current x86 family, but using G4 or better, and modern interrupts, interfaces etc. Current machines from Apple have nothing to do with this idea - but one of these from Merlancia is on my wishlist.

Intel CPUs will be killed by Microsoft's CLR

[eyefish]

It is my opinion that once Microsoft makes its Common Language Runtime a forced deFacto standard, and once they manage to implement it on other CPU architectures, they'll essentially have a hardware-independent Windows platform. Once that happens Microsoft will have sole leverage on the PC business. That means that Intel will NOT be needed at all for running future versions of Windows-compatible programs. Who knows, maybe this could spell a revival on new and innnovative CPU architectures, since they all will now be able to run the CLR. Side note: We *could* do this today with Java, but sadly Sun doesn't have the leverage Microsoft's monopoly does on the PC business.

Re:Intel CPUs will be killed by Microsoft's CLR

[javiercero]

They already tried that. Guess what? NT was supposed to be multiplatform! And geez do you see any of the non-X86 versions out there? Nope....

In fact, NT was developed on MIPS. And M$ is in no way interested in having the CLR running on non windows based platforms. CLR is not designed to make code machine-independent, but rather location-independent. M$ still wants you to be using Windows, it just wants to have a tighter grip on you no matter where you go.

Why would anyone even think about adopting .NET is beyond me.

What will drive Itanium price down?

[sphealey]

First, you are assuming that Itanium will succeed and drive all other choices from the market. At the moment, this is far from clear, and even Intel is said to be hedging their bets with a P4 follow-on.

Second, what will drive the price of the Itanium down? Historically, Intel have announced that their latest superchip is "targeted at servers, not desktops" about a week before releasing a flood of them into the desktop marketplace (usually the ones that didn't pass spec at the higher speed level), thus driving down the price of the server chips to where no one else could compete. What will be the driver this time? Businesses aren't buying desktops, and when they do start buying again it will be pure commodity: there is zero appeal for Itanium on a business desktop. And treble for home desktops.

Which leaves high-end servers. I don't think that any datacentre manager worth his pay is going to pull out $100,000 HP N-Class boxes in favor of $2,000 Intel clones. There's a bit more that goes into a server than the CPU.

sPh

Re:What will drive Itanium price down?

[sphealey]

So are you saying that you could move more people with one Cadillac than 50 Yugo's

Excellent analogy. Yugos were cheap, and it appeared that they could replace more expensive cars. Then they started coming apart on the highway at speed, leading to a less than 100% chance of arriving at one's destination alive, much less on time, if using a Yugo. So they disappeared from the market in favor of more expensive, but more reliable and serviceable, cars.

sPh

Re:*cough* PowerPC *cough*

[joe_n_bloe]

Also featuring stinking fast floating point.

Dead? I doubt it.

[BlackStar]

SPARC dead? I'm not sure where you come across that idea. Having listened to a few talks down at JavaOne and chatted briefly with Marc Tremblay (head chip dude down there, father of MAJC and one designer of SPARC) they've already got design down on the next two levels of SPARC as the IV is experimental, and the V is the next production level as I understand it. MAJC seems to be the experimental platform they are using for smaller implementations and alternative ideas to be tried, based on some of Tremblay's theories.

I may be off base on some of the details, but Sun has a unified approach from top to bottom, from tools to silicon for the systems they plan to deliver. I doubt it will just throw in the towel. Ultimately, Sun ships iron, and they lead the market in their segment.

I don't see the basis for your assertion, and where you pulled 1B out of for cost I also don't know.

Alpha is AMD now, as that's where a good chunk of the people went. MIPS is still kicking, with the 14000 so far, but I won't speak to the future of that chip line. There's a lot of chip heads on this site with much better info than I on many of the lines.

One decent, although dated summary is here

Please tell me there's more information you're basing this on than consumer workstation marketshare....

64bit code

[geekoid]

before everybody starts saying (too late, i'm sure) there is no 64bit software to support this chip, I'd like to point them to here.

Re:64bit code

[lkaos]

Bah! Bah!

You should of pointed them here.

It was first after all.

Re:Innovation in the CPU business

[GGardner]

Nice idea, but keep in mind that static compilers are extremely difficult to create for Itanium. Performance results I've seen show that while the theoretical maximum for IA-64 is pretty impressive, the actual results static compilers are generating are not so hot.

Now, try to write a dynamic, JIT compiler for Itanium, which is even hardware than a static compiler. I haven't seen any java or CLR performance numbers for IA-64, and suspect I know the reason why. :-)

More importantly...

[Kerne]

A fast CPU is nice, but how about upgrading the rest of the standard PC architecture and peripherals to the same level?

Weren't we all suppose to be using high-speed serial connections by now instead of a cocktail of SCSI (1/2/3, wide, fast, hold the mayo), IDE (ATA-33/66/100), parallel, 8 bit serial, USB, Firewire, PS/2, PCI, ISA (which is finally disappearing), etc. Heck, I'd be happy if the motherboard ran at even half to a third the speed of the cpu. :P

Using a 20 year old peripheral port on last weeks multi-gig cpu is like sucking a McDonalds shake through a coffee stirrer!

Re:More importantly...

[iomud]

Agreed. We're disproportionaly favoring cpu when the real gains would be seen in high speed interconnects especially with storage devices. Most of my cpu's time is spent waiting for instructions and even when sent could stand to recieve them both faster and in greater numbers.

Re:More importantly...

[boopus]

Yes, In one of my CS classes we were told the statistic(which was probably made up, and i've since forgotten) about how long it takes to read the entire contents of the hard drive. If current trends keep up, it'll soon take us weeks to just read everything we can store on one hard drive. Anyone have "hard" figures?

Re:More importantly...

[Bishop]

The real kicker is that the 8259 is long gone, but modern chipsets emulate cascaded 8259s. The other joke is the A20 address line. All because of two decade old legacy software.

NVidia, the next player

[Animats]

My own guess for the desktop is that NVidia will put a CPU core, probably from AMD, in the next generation of their nForce part. That puts CPU, graphics, networking, sound, disk control, and the motherboard logic on a single chip. Their current nForce part already has all of that but the CPU.

If you look at the transistor counts, NVidia's graphic chips already are more complicated than most CPU parts. This is quite do-able.

Re:NVidia, the next player

[maraist]

if you look at the transistor counts, NVidia's graphic chips already are more complicated than most CPU parts. This is quite do-able.

There's more to [CG]PU complexity than transistor count. Look at the 512Mbit memory cells that run for only a couple dollars a chip.

The trick is inter-related logic complexity. To my understanding the existing GPUs have no issues with backward compatability (so much of the x86 overhead is avoided), the core itself is pipelined and modular, so the complexity is spread out across the whole chip (independent teams can work on their own components with little concern for sistern components, whereas every ounce of performance is being squeezed out of x86's which require complete coordination). Further, graphics acceleration is simply the application of graphical algorithms into silicon. While I'm not quite sure which algorithms there are, the possibilities are endless. Imagine a fast-fourier transform implemented as a SIMD floating point instruction. You create an array of floating point logic units, and interconnect them. The floating point unit is pretty much a common-off-the-shelf design, so the only real logic you apply is the interconnectivity.

I'm not saying that GPU's are easy to design, I'm just saying that hardware filters are designed this way all the time, and I would'nt be surprised if a large percentage of the nVida chips weren't stock logic modules.

-Michael

Re:NVidia, the next player

[Hektor_Troy]

"2. So you have a 8 million transistor chip (CPU), and a 7 million transistor (graphics) and 2 million transistors (sound, disc, etc.) on one chip. That's 17 Million transistors- that's gonna be a huge fucking die size. Whatever you save by selling an all in one system you eat by the cost of the huge @#$#ing die, higher failure rate, heat issues, etc. A big expensive chip is not condusive to low-end markets. "

The AthlonXP has 37.5 million transistors and has a die size of 128 mm^2, the Geforce4 has 63(!) million transistors (don't know the die size of that one), but hey - just because I spent 3 minutes checking the facts at the manufacturers websites doesn't mean that I'm right - I'm sure that that 8 million transistor chip is correct along with the 7 million tranistor GPU is correct as well, I just don't know what cpu and gpu you are refering to. Of course the two forementioned chips get quite hot, but hey - they still work, and die shrinks tend to solve some of the heat issues (I think).

"3. The NForce isn't a single chip, it's a set of chips. Yeah, they could put a CPU as part of the package, but why would mobo makers put a surfacemount cpu on a board that would obsolete inventory every 6 months? Why not put a socket on and let the distributors and retailers worry about what CPU to ship with it?"

The Hammer is supposed to contain the northbridge on die, so you could probably do the same for the southbridge as well (put it in/on a socketable chip) giving you the option of upgrading that as well.

Sure, build your own...

[twoflower]

Sure, build your own box for $2k instead of buying one for three times that much -- if you don't mind being fired.

You don't pay $6k or $8k for a server just because there's high markup on the parts. A lot of it is due to tighter tolerances required for high-availability or high-reliability equipment. There's greater consideration for issues of heat, RF, power consumption and stability -- and then there's the built-in redundancy for many components (power supplies, fans, etc).

It's not as simple as you think.

Twoflower

Re:I am confused

[sphealey]

If I worked for a major "research" organisation which couldn't even find out simple stuff like the short term future of the microprocessor industry then I would want to remain anonymous too.

One sees this comment in every Ask Slashdot thread. It is not only tiresome, it is wrong as well. Seeking multiple opinions from diverse sources is certainly part of research, similar to skimming all the current trade and academic publications.

Sorting out the meaningful comments from the slush is part of good research.

sPh

Re:Just wondering, not a troll.

[putzin]

It's different because they haven't signed exclusive deals and used marketing to force other competitors out of the fray. Essentially, they will have priced the competitors out of the building. I'm not saying they aren't a monopoly, but realistically, it's harder to argue they did it illegally or unjustly.

However, I still think that there will be room for others. AMD will probably succeed doing what they do best, outpace Intel in quality and lower the price by ~10%. This has been successfull (I hope it continues, I own stock) and will probably continue. And I doubt Sun is out. There maybe changes coming, but I figure McNealy would sell his baby prior to using Intel chips. As for the others, they fell and never recovered. You can't charge super high premiums when your competition is charging super low premiums. A lot of corps assumed you could and get away with it and look what happend.

The future is unwritten, so any sort of prediction is just fantasy for the most part. Step back to 95 and tell me who predicted 2000 or 2001? Reality is far more interesting than any professional opinion from the Gartner group et. al.

You are nuts if you think ...

[joe_n_bloe]

... that a runtime environment where "Hello World" will require, let's say, several GB of disk, a few hundred MB of RAM, continuous online updating (also requiring continuous hardware updating), and hundreds of old and newly-arriving security holes and exploits, is going to "take over the world."

Granted, it's going to be popular for a while. But isn't what's popular *always* sucky?

Re:You are nuts if you think ...

[kubrick]

... that a runtime environment where "Hello World" will require, let's say, several GB of disk, a few hundred MB of RAM, continuous online updating (also requiring continuous hardware updating), and hundreds of old and newly-arriving security holes and exploits, is going to "take over the world."

Sounds just like Java in the 90s, hey? :) (although the security holes are MS's special addition to the experience, I imagine)

I wouldn't count out everyone else yet.

[jtshaw]

You talk alot about Sparc, MIPS, and Alpha in that question of yours. Yes, those are all relatively low volume products, yes they do cost a lot of money. However, the Itanium is almost like Intels version of those products, done in a slightly different way. Even though they are made in lower volumes they are still profitable because the people buying them will pay a lot more for a system. Sun can sell a 64-processor UltraSpac III system for in the realm of a million dollars and more. If you don't think they are making a nasty profit of of that you are nuts. That is why they keep advancing the technology.

People love to through buzz words like 64-bit vs. 32-bit and stuff like that but when it comes down to it what do you need on your desktop? If you are using your PC for basic development or coding there is not much to be gained from a 64-bit core at all. You don't really need anymore precision. If you are talking about scientific applications then maybe you do need the 64-bit core.

I am not saying that desktop PC's won't eventually go to 64-bit cores. However, even if you were to get a cheap Itanium right now it would perform no better, and possibly worse then your high end AMD and Intel x86 processors because few of your applications would take advantage of the core.

This question will be better asked for when Intel puts a processor on there desktop timeline that utilizes IA-64 technology.

Sparc dead? And what about SGI?

[wizzy403]

Umm... Given how well Sun is entrenched in the financial world, I think you saying the platform is dead is just plain FUD. Check with the IT department at any major financial company and ask them how many 4500 or better systems they have. (I know, I used to work for one) And yes, a lot of them are upgrading to the new UltraSparc III machines.

And for those folks doing hard research (or special effects companies with lots o' money) SGI is still king. Despite what nvidia would like us to believe, SGI's not going anywhere anytime soon for big 3d rendering projects.

Address space requirements

[Florian Weimer]

At the moment, Itanium systems are worth their money only if you have large address space requirements. Intel seems to focus on optimizing the Pentium 4 compiler, and not the Itanium compiler. I doubt that the Itanium architecture will surpass IA32/x86 on the desktop (where 4GB is enough for everyone ;-) anytime soon.

That's why I doubt that we are going to see affordable IA64 systems soon. After all, the transition is quite rough, thanks to Itanium's abysmal IA32 emulation (performance-wise), so there isn't even much market demand.

In the future, Intel may well decide to switch to the IA64 instruction set before it is really time for it, just to make things a bit more complicated for AMD.

Re:Address space requirements

[Florian Weimer]

IA32 can currently handle up to 64 GB in one node, with some kind of EMS-like hack. This means that you can put more than 4 GB in your machine (actually, PCI devices need adress space, too, so you hit the barrier at 3.9 GB or so), and still use all of it.

On the other hand, the per-process adress space is still limited to 4 GB. I don't think this is a concern for the pro user who wants to show off his RAM size, though.

It depends on what you need

[jayslambast]

You have a very valid question, but you're statement,

"At that point, why should one spend $8-10k for that hardware from the likes of HP, Compaq, Dell and others when one can build it for $2k (or even less)?"

Is missing something. HP, Compaq and Dell provide more than the hardware. They provide services that go along with the HW. They use the hardware to suck you into to using their services. While small companies can build these systems on their own for cheaper, the larger companies are the ones that need to outsource some things that HP, Compaq and Dell's services provide.

Also its kind of silly to think that these IA-64 systems will be able to be built for $2k each (given the cost of similiarly performance) Sparc's and IBMs. Intel is hoping for their backwards compatibility and clout to push ISVs into programming for their systems. Once they have those vendors in their camps, the chip and server prices will go up again.

And finally, most people that would need a 64bit solution will probably need multiproc systems. OEM's will be able to provide the small systems, but once you go past the 4-8 way space, there really isn't a cheep way of scaling up any higher (, and btw, clustering is really only a solution for tasks that don't involve large sharing of data between processors that is time sensitive.) Which is where HP, Compaq, Fujitsu, NEC, and IBM will be with their high-end systems. I doubt I will ever see Dell release a system with more than 8 IA-64 processors.

Of course only time will tell what will happen next. OH, one last thing. The guy who posted should be informed that HP did not sell any processor guys, they sold some chipset guys to Intel. I'm surprised that someone that is in a processor research group would not know this. Checkout:
http://slashdot.org/comments.pl?sid=22319&threshol d=0&commentsort=3&tid=118&mode=thread&cid=0

Parallel machines.

[Christopher Thomas]

Also, wasn't this inevitable. There are a few Beowulf jokes being posted, but that's really what's going on. Increasingly high performance tasks (Google, render farms etc. etc. etc.) are using massive arrays of low-power CPUs. It costs a lot of money to develop big iron chips, and if people aren't buying them then there's no point in investing that much money.

The problem is that a massively parallel computer is only useful for certain classes of problem. There are many types of problem where communications load goes up very rapidly with the number of processors, which makes a cluster (with its relatively poor communications bandwidth) impractical. This is what Big Iron is designed to be useful for.

Re:Parallel machines.

[foobar104]

There are many types of problem where communications load goes up very rapidly with the number of processors, which makes a cluster (with its relatively poor communications bandwidth) impractical. This is what Big Iron is designed to be useful for.

Also, some problems are inherently serial, and can't be easily parallelized. When single-threaded performance is important, you need the fastest single CPU you can get your hands on.

My 2c

[UTPinky]

I had a professor last semester that worked at Intel, and several things he told me, reminded me of somthing: It's still a busisness. In my opinion Intel will not make any huge move, until they KNOW that they will profit off of it. This means that they won't make any major move until the consumer market is there. For example, he was telling us that there have been times where they have come up with ideas that would in fact increase performance, HOWEVER due to their wonderful job at brainwashing the entire public into thinking that clockspeed is THE measure of performance, they scrapped the ideas because they noticed that they would cost too much to implement, and would result in no frequency increase. (Thanks Intel)

I also think that while AMD has shown that they can provide an honest competition in terms of performance, it is going to be stuck following Intel's every move, for the mere reason that Intel is "sleeping with" so many big OEMS (*cough* Dell *cough*), leaving it as the CPU for the hobbyist

Well, anyways, that's just my 2c...

Itanium? in $2k systems?

[orz]

You're not going to be getting an Itanium based system for $2000 anytime soon.

First of all, Intel has said ever since the Itaniums much-delayed release that it couldn't really compete and is primarily released to get some infrastructure ready for when the McKinley is ready (IIRC, it's scheduled for about 3 months from now...).

Secondly, the die size for the McKinley is HUGE. On todays top-of-the-line .13 micron process, the manufacturing costs are likely to be too high for this chip to make it into high-end workstations, let alone $2000 consumer computers.

Thirdly, the competition isn't dead yet. Sparc and PA-RISC may be dead, but Sun offers competition, and IBMs Power4 will be a decent competitor. Alpha does indeed look to have disappeared, but I thought I heard something about some Japanese company buying rights to some Alpha stuff, and planning on a big die shrink and integrating a large cache (which is all the Alpha really needs to compete, for the near future).

Fourth of all, the performance of even the McKinley is questionable. Compilers for it's IA64 instruction set are still quite poor, with little sign of the anticipated improvements. It's predecessors, the Merced/Itanium, was dog-slow at most tasks (though good at floating-point). The most recent benchmarks show the McKinleys 32-bit performance as terrible, though it's floating-point performance is supposed to be stellar, and its integer performance decent (when combined with an enormous on-die L3 cache...).

Anyway. Intel just likes the Itanium because the the instruction set is sufficiently complex that the prohibitive cost of designing a compatible would raise the cost of entry to the market enough to give them a more secure monopoly for the next decade.

64-bit isn't necessary - and Itanium may suck

[camusatan]

The implicit assumption that the author is making here is that 64-bit CPU's such as Itanium will be the 'next big thing'. I'm not sure - 64-bit CPU's really only are necessary for machines that need more than 4 GB of VM space - and with various x86 addressing extensions, some IA32 CPU's can address up to 16 GB (I think).

Now don't get me wrong - 64-bit filesystems are great, and necessary - being limited to 2GB or 4GB files is terrible. But no 64-bit CPU is necessary for that kind of thing, the filesystem just has to be written as 64-bit (which is easier said than done, and could easily sacrifice backwards-compatibility with various API's, but I digress...).

That being said - Intel might very well be moving down the wrong path - the Itanium is a huge, expensive, hot, completely new chip. Even Intel is hedging its bets on whether or not Itanium will take off - and AMD is poised to eat Intel's lunch with their new Hammer design.

Who knows, perhaps all CPU's from now on will be compatible with x86 IA32, and innovation will be in the various processing units that sit behind the instruction-set decoder. Take a look at AMD or Transmeta for examples of that, already.

Re:64-bit isn't necessary - and Itanium may suck

[dlapine]

Both Linux with 2.4 kernels and Windows NTFS support large files (files greater than 4 gig) on 32bit CPU's. Yes, there is a need for apps to be upgraded to support this, but RedHat Linux 7.2 supports this out of the box.

It's possible that updates to the Itanium (McKinley and so on) may prove to be more useful than the Itanium itself. Anyone remember when the 60MHZ Pentium was released? It was hot, expensive, huge and completely new, just like the Itanium. It wasn't really worth the purchase price, in comparison to what was available in 486 at the time. You have to start somewhere...

64bit CPU's will provide a much better range of addressing, both for RAM and filesystems. They will also be faster than their 32bit cousins. How do I know? Well, we have 2 large-scale linux clusters here at NCSA. The IA32 cluster has 512 dual PIII nodes at 1.0 GHZ. The IA64 cluster has 160 dual Itaniums at 800MHZ. The Itanium cluster is about 10% faster for its top500 run. Check out entries 34 & 41. Yes, the Itanium cluster has more RAM per node than the IA32 and it was running RedHat 7.1 versus RedHat 6.2 for its older sibling, but you're still looking at a factor of 3 in CPU to CPU comparisions.

To sum up, Itanium doesn't quite suck :) Let's hope the AMD 64bit CPU's are even better!

Re:64-bit isn't necessary - and Itanium may suck

[Guy Harris]

from what I remember, the vax(yes, old, but still 36 bit) could handle a max of 64 Gb.

You're either trolling or misremembering or think "N-bit" refers to the size of physical addresses. VAXes were 32-bit processors; I no longer remember what the page table entries looked like, so I don't know how large the physical addresses were (but the VAX-11/780 was designed in an era when 4GB was a lot of disk space, so VAXes may well not have had even 32 bits of physical address).

32-bit machines can handle more than 4GB of physical memory. (In fact, the Pentium {Pro, II, III} and successor x86 processors can support 36-bit physical addresses.) They just can't handle it conveniently within a single process (no, segmentation on x86 doesn't make things better; it's not even necessary).

There are some people who might be delighted to see 36-bit desktop machines, but I doubt you're likely to see any new 36-bit general-purpose processors any time soon.

Consolidation is a Fact of Life

[dso]

Just look at the auto industry. GM, Ford, Chrysler began the North American market by consolidating all the smaller auto companies and dominated for years. Then along came Honda, Toyota, Nissan and now they have made huge gains.

The fact is that even though it looks impossible to overcome Intel at this point, someday someone will.

Re:Consolidation is a Fact of Life

[Peyna]

In 1999, here's what the market share for automobiles in the USA looked like:

General Motors: 29.2%

Ford: 18.1%

Honda: 9.8%

Toyota: 9.1%

Chrysler: 8.5%

GM has acquired Saab since then, and I'm sure a few other small companies have changed hands, but their numbers are small enough to not really matter.

Apparently Chrysler is no longer 1 of the Big 3.. but GM and Ford look pretty untouchable right now. (Doesn't mean it will always be that way.)

The killer is custom-made systems...

[Kjella]

Rewriting standard applications to take advantage of the Itanium is one thing. However, companies that need a $10k+ server usually have programs that are specialized. After 20 years of the x86 standard there's a large codebase, although given a few improvements along the way. If you read the FreeDOS article a little while back companies were still running DOS in production systems, because it *works*. Porting it to Itanium will be a lot worse than porting it to x86-64 and Hammer. Let's face it, the hardware cost is usually minimal today. Software programmers however, are not cheap.

Kjella

you're thinking too far away

[Anonymous Coward]

You won't see anybody building an Itanium for $2K, since the chips cost more than that when you buy 1000 of them at a time.

Maybe 10 years from now, but that's too far off.

1) HP's PA-RISC is as dead as Intel's x86

2) Alpha should regain the speed crown with the EV7 for a while, so they aren't dead yet. They've just announced they'll be dead in a few years :)

3) IBM's POWER4 is the current speed king and is likely to be around for a long long time.

4) MIPS.. Aren't these popular RISC chips in the world due to their embedded use? (N64, Playstation, networking) At 500Mhz in SGI's machines they are pretty dead, but various MIPS chips are doing quite well in emerging areas. Infact AMD just bought a MIPS company.

5) Sparc has never been that great CPU vs CPU with the other companies, but I expect them to be around for a fairly long time still, just based on their installed base. Their customers never really bought on performance (otherwise ALpha would still be around!), but on service and reliablity. As long as they can provide good enough performance they'll be around.

The next Itanium is HUGE making it very expensive to produce (meaning you won't ever build a system for under $2K with one!), requires a LOT of optimization in software to get accepable perfomance (meaning it'll suck unless you run active profiling optimizations and I doubt most game companies will even do that), it uses a lot of power and creates a lot of heat (it makes the Athlon/P4 look like embedded chips!), and it isn't really compatible with existing software. Nobody is going to run Win98, WinXP, or even GNU/Linux on it on the desktop.

The next Itanium will be more popular than the last, but it won't even register on people's radars as it won't provide the best performance, it won't have a bunch of software written for it, and it'll be expensive. Apple will sell more iBooks than Intel sells Itaniums for the next few years.

Itanium will take YEARS to commoditize...

[barfy]

There is little compelling need for desktop users (the ones that create the volume for commoditization) to move to 64 bit systems.

Until there is breakthrough brought on by computing speed, we will see a stall in computer upgrading as we have seen in the past.

I expect we will see more things like the Imac (very cool computers), before we see a press for new computers for speed.

The two things I think will create the next level breakthrough.

Real Time CGI imaging at Toystory/Mosters INC/FF, level of quality. We can probably predict precisely WHEN that will be possible by mapping the development speed of 3d hardware, memory, software breakthroughs, and polygon density to date, and where the predictable bottlenecks will appear. (My suspicion is that we are 5-8 years away).

The other breakthrough which I think would do it, and right now it is very difficult to predict when it will happen, but I suspect that adoption would be pretty rapid, is real time voice interaction that is 5 9's accurate. This is likely to appear after a certain speed level of computers, and a breakthrough understanding/algorithm for speech recognition.

However, I suspect the AMD x86-64 solution may be adopted much faster than the Itanium solution. Likely there is an app out there that may have a large enough niche to require 64 bit apps, and the rest of the apps on the computer would be 32 bit. I suspect that the app will be imaging or video related, and that will create an adoption around the AMD solution, before the Itanium moves out of the server market to the desktop market where it will be commoditized.

Re:Itanium will take YEARS to commoditize...

[donglekey]

You are missing a big issue for computers that will emerge, and that is real time compression of video codecs like JPEG2000. JPG2k has seemed to go every which way but sober, but it is a proof of concept. Imagine a codec 2 or even 4 times as good as DiviX. Enocode it in real time and you have a very good video stream on your hands. Two people with broadband could video conference very comfortably.

Mismatch?

[90XDoubleSide]

except IBM Power4 [& friends G4, et al]

While the Power4 will no doubt compete with the Itanium in the server space, since many people are talking about when 64-bit chips will hit the desktop, you should note that its "friend" the G4, which has been out since before the P4, is by no means meant to compete with new Intel offerings; the Goldfish PowerPC 8500 ("G5") is aimed squarely to dominate the desktop space before Intel can get to it with 64-bit chips. It's ability to run 32-bit code at much better speed than the othet 64-bit offerings makes it much more appealing to people looking to transition to 64-bit on the desktop, and if they can pull off the .13 SOI, 500MHz RapidIO bus, etc. it should reassert A.I.M.'s competitiveness in high-end desktops. Now when it will actually ship, how much of this will get implemented, and at what frequency it starts at is anyone's guess.

The Answer

[sharkey]

A boring movie titled "Itanic", starring an effeminate man-boy and a chubby love interest.

Re:Innovation in the CPU business

[Courageous]

Virtual machines rely on things like delayed compiling that are fairly antithetical to the whole idea of Itanium, where they push enormous amounts of work previously handled by the CPU out to the compiler. Personally, I believe that VLIW for general purpose processors was a really bad idea that was disproven a good decade ago. Intel is in the middle of giant train wreck, and the market doesn't even know it yet.

Consider the downside of pushing the majority of your branch prediction to the compiler. For example, the compiler doesn't know about multiple processes and how they interact with eachother! This means that it's likely that Itanium boxes won't even serve transactions very well. This begs the question of what Itanium will be useful for. If it's not for the desktop, and it's not for transaction service, what the heck is it for? High end scientific computing? Competing for Alpha's market share is a big mistake, in my mind.

C//

Re:Just wondering, not a troll. SUN IS OUT

[jproudfo]

Give me a break.

Anyone who sees the recent Sun announcements (re: Linux) as the end of SPARC or Solaris, clearly doesn't know anything about the business world or about Sun.

Yes, Sun has made an announcement to start supporting Linux. This is no big surprise, especially after the Cobalt aquisition.

This doesn't mean that they are switching to Intel or giving up on the SPARC architecture.

SPARC is far from dead. All you have to do is talk to anyone within Sun to see the U4 and U5 roadmaps. Sun firmly believes in their architecture and has/will spend the R&D to to continue to develop it.

Plus, the install base of these technologies is much too large for them to just give up on them.

Look at HP, for example... Here is a company that is part of the engineering process for Itanium. They've already committed to use Itanium on their higher end servers, but they aren't completely giving up on their PA series CPUS (yet). All of their new systems take both.

No company wants to alienate the majority of their install base. :)

Re:Dead? I doubt it.

[elmegil]

I don't see the basis for your assertion, and where you pulled 1B out of for cost I also don't know.

I dont' know either. If it were a valid cost though, it's worth nothing that according to the annual report, Sun has $18B in assets, and I'm willing to bet at least 1/18th of that is "in the bank" and available. SPARC is the foundation of the company, moreso even than Solaris, and it's hardly dead.

Wrong point of view guys...

[MrPerfekt]

This is a typical example of someone lacking clue and claiming to be authorative. I can admit I know nothing of most of the arch's there, but I can tell you about the SPARC.

The UltraSPARC for workstations has always kinda been a niche market. For the simple reason, that you can get an Intel box with far more hardware options and software support and for far less money.

However, in the server market (which I doubt the submitter has ever had any experience in) is a different story. For the most part, hardware support is irrelevant if it does what you want it to do. Which in most cases is just be some type of Internet server.. be it oracle databases or web servers or whatever. People that run critical servers and need the UltraSPARCs stability and Sun's support (or this can go for some other alt. arch. like IBM and an AS400) almost always do buy something other than Intel for their mission critical stuff.

Anyway, my whole point is, just because you don't use it in your workstations (or your webserv0r on your dsl line) doesn't mean its dead. Workstations and Servers are and hopefully always will be very very different to actual companies that need a different level of service from their servers. I suspect because the submitter has a lunix server with a mandrake enterprise kernel, he thought he was an enterprise business.

Re:Wrong point of view guys...

[sql*kitten]

The UltraSPARC for workstations has always kinda been a niche market. For the simple reason, that you can get an Intel box with far more hardware options and software support and for far less money.

I think you may be confusing "workstation" with "high end PC". Sun's equipment has always been popular with workstation users, in engineering, finance, oil and gas, scientific research, academia, etc.

Remember the Pentium Pros

[Pharmboy]

Yes. Yes I do. I still have 4 servers (IBM Personal Server 325) that each have two of the Pentium Pro 200's in them. One is running Win2k, the others are running Linux (rh6.2). They each have 4 SCSI drives (ufw/40mbps/2.1gb). They each have around 35000+ hours on them (4 years x 24/7). I have not replaced them because I have only had one go bad (now is spare parts).

I kept them because the quality of the ppro is UNREAL. I have not replaced them because the quality of PIIs and PIIIs are, well, OK at best, and Xeon's are simply overpriced for what they are.

Yes, I am just strengthening your point, to make a point. Those of us in the "smaller" world of serving will take durability over speed, reliability over clock ticks. What will get me to switch to AMD or Itaniums is that warm fuzzy feeling you get when you go to sleep, and don't have to worry about driving into town (30 minutes away) at 3am to reboot (or switch over) a server.

I did just order a Dell dual p3/1000, but it wont replace any of those machines until I have 90 days with it in place. (average uptime on the Linux IBM's is over 6 months)

Build an $8-10 server for $2k - um, no.

[sirwired]

No, you can't build something like a Netfinity (oops. er - xSeries eServer) in your garage for $2k. Built into a high-level xSeries is:

1) Hot-pluggable power supplies, drives, and PCI - slots.
2) Built-in hot-plug SCSI
3) Integrated service processor for diagnostics (essentially a computer within a computer)
4) Extremely well-tested box. (Very important to do integration testing on high-end units.)
5) Very nice, serviceable, rack-mount chassis
6) Crap-load of PCI slots
7) Light-path diagnostics. (Lets somebody without training figure out what's broke.)
8) IBM Director
9) Well-designed cooling that would be impossible to achieve with a garage box. (Do you know how to do airflow modeling?)
10) Support.

The list goes on...

Yes, they will become a commodity, in that you will be able to get them from multiple major manufacturers, but don't expect to build it yourself in your basement anytime soon.

SirWired

Re:Build an $8-10 server for $2k - um, no.

[cabbey]

Exactly! It always amazes me how many people don't take this into account! The $6K you saved on parts will be lost the first time you have a power supply failure and crash the whole machine, or have a drive go bad in the raid and have to take the whole box down, pull it apart and swap one drive rather than just walk up, pop a clip, slide out the drive, slide in a replacement and snap it back, then login as root and issue a command to put it back into the raid. Compare no down time versus at least a half hour (if you rush the job, thereby risking doing more damage). The savings just aren't worth it for any business, let's face it that's why IBM exists and makes the kind of money it does.

Re:Nano-technology

[Com2Kid]

Hmm, full home computers?

A mid 1980's home computer could EASILY be reduced to fit down to the size of my pocket.

Actualy size in itself is not the problem. The bleeming screen is.

Until we get some direct to retina or direct to optic nerve display technology, the size of a computer is always going to be limited by what the smallest display the user will stand for is.

Well that and keyboard sizes.

So you should also add direct mental input to your list of features that are needed.

Quite frankly, if you give a modern day computer manufacturing facility the technologies that I have outlined above, and in a decent sized package, a computer could EASILY be made that fits in the palm of your hand.

Hell, lets see now. Use a modified form of Sony's memory stick technology, they have gotten in packed down quite dense now days, so you really do not need such a large package if you are just going to store 4 or 8 megs of data.

The CPU should be no problem. Since this is a business computer we are dealing with here, no FPUs are needed, and 66mhz or so should be enough to run a highly optimized operating system along with some standard business applications.

Hmm, actualy, have you seen those MP3 players that they sell in the stores now days? Yah, those ones, the ones that are about the size of two of my thumbs next to each other. (or to put it in references that mean something, heh, about an inch tall by a bit less then an inch wide)

That is what we can do now days.

It is just the friggen display technology that is holding us back.

Everything else until then (new display technologies coming out) is just a stop gap measure designed to keep the technology sector alive.

Why else do you think that the latest office applications require 500mhz+ to run? Seriously? (this is MS Office of course, bleh. POS. . . . )

Wrong math, was Re:compilers

[HuguesT]

Hi,

Speed of light is 3.10^8 m/s

In a nanosecond (10^-9s), light travels 30cm,
not 1cm like you wrote.

Re: Wrong math, was Re:compilers

[Alan Partridge]

so let me get this straight, you're asserting that the SPEED OF LIGHT is 30cm per second?? has the whole world gone crazy here?

Re: Wrong math, was Re:compilers

[Doctor Bill]

Doh!

(open mount insert foot, chew, spit toes)

That would be approx 30cm/ns, *not* second.

What's 9 orders of magnitude between friends, eh?

-b

Re:Actually, 4gb is not enough

[Florian Weimer]

I was careful to talk about desktop machines. ;-) There are no applications which require 4 GB per process on the desktop right now. Even databases do not require that much address space all at once (although it might simplify coding---but who runs large databases on a desktop computer?). Number crunching is a different issue, but you usually have specialized machines for this application if you really need it.

Oh no he isn't

[Mike Greaves]

All of those little guys make much smaller planes and/or a smaller range of models than Boeing and Airbus. Boeing and Airbus are the *only*, I repeat *ONLY*, two effective competitors left in the world-wide market for large airliners.

Lockheed abandoned the market; McDonnell Douglas was bought by Boeing, and so on...

Ultimately, only IBM and Intel can stand on their own when it comes to state of the art fabs. It was the cost of continually upgrading the fab that caused Compaq to abandon Alpha. Sun will be very fortunate if it continues to succeed in it's traditional fab arangement with TI. I think that Sun, TI, AMD (and maybe others) should try to pool their resources some way here; maybe with jointly owned/operated fabs or something...

Re:Just wondering, not a troll.

[the eric conspiracy]

We are approaching the end of Moore's law for silicon

Based on what? We have seen plenty of reports of technologies that are in real development that will keep Moore's 'Law' valid for at least another 10 years.

Some more SPARC news...

[Anonymous Coward]

I heard SPARC chips are so fucking scared of the multi-GHz x86 clones that they are running their instructions out of order! Some of the Sparc instructions think they can even hide in a delay slot (under a jump) so the x86 clones won't find them and kick their sorry out-of-date asses!

Re:Sorry...you're wrong..

[Jeffrey Baker]

Whoa there. You just pointed out three HUGE companies entering the aircraft market, which strengthens the point that the barriers to entry are high. Bombardier manufacturers at least part of every passenger rail car currently deployed. They have been in the aircraft business since 1989 IIRC, their revenues just on aircraft exceed $10bn Canadian annually. I believe they are the thrid largest civil aircraft company after Boeing and Airbus. Their R&D efforts are funded significantly by Canadian tax revenue. So, all it took for Bombardier to break into the aircraft industry was to leverage a strong position in the rail market and get the backing of a major national government.

Embraer is a $6bn company that has been around for over 30 years. Again, it took no small resources to break into the commercial aircraft business: they funded that effort with their cash cows in suppyling and repairing military aircraft in the world's forgotten air forces.

It is also strange to say that Fairchild is some kind of up-and-coming civil aviation company. Both Fairchild and Dornier have been around since before WWII. Their chief product was developed with R&D money from Daimler Benz, and they are now owned by the largest insurance company in Europe. Also I believe they have not delivered any aircraft to major customers since their reentry into the market.

It may be that Boeing and Airbus are not invincible, that the age of the large jet has passed, and that a company with few resources and a big idea could come along and knock those Goliaths over. But, the three huge companies you mentioned don't qualify for the role of David.

Re:Remeber what about the Pentium Pros?

[Talonius]

Fact is that the Pentium Pro itself was a marketing failure. While you go on to state that the future lines of Intel chipsets were based on the Pentium Pro that is and isn't true. The PPro when considered alone, was a dismal retail failure.

Consider me an MBA if you'd like. I'm fairly close to that.

Re:Get that new heat sink ready...

[be-fan]

Umm, the first Hammer proc (clawhammer) is a laptop CPU with 100mm^2 die size. Should be pretty cool...

Re:When will we finally have CLOCKLESS CPUs?

[sn0wcrsh]

Have you ever tried to do large scale clockless design?

Imagine a ballet without music.... Very hard to
get things right.

Peripheral communication.

[Christopher Thomas]

A fast CPU is nice, but how about upgrading the rest of the standard PC architecture and peripherals to the same level?

Weren't we all suppose to be using high-speed serial connections by now instead of a cocktail of SCSI (1/2/3, wide, fast, hold the mayo), IDE (ATA-33/66/100), parallel, 8 bit serial, USB, Firewire, PS/2, PCI, ISA (which is finally disappearing), etc. Heck, I'd be happy if the motherboard ran at even half to a third the speed of the cpu. :P

The good news is that USB is well on its way to completely replacing serial and parallel ports, and that PCI has been the One True Bus for the past couple of years now. Everything south of the southbridge is slowly fading away.

IMO, if we'd switched to 66 MHz 64-bit PCI years ago, we'd have no further problems on this front. In practice, PCI-X may finally be pushed through by Intel, and that will serve most internal communications needs. Motherboard chipsets are modular enough that it doesn't really matter what flavour of IDE/SCSI/firewire your drive is hanging off of; the drive controller is just another PCI device to the processor. You have enough bandwidth and DMA functionality on PCI bus to handle it.

The only peripherals that are currently bottlenecks are RAM and the video card. RAM is handled by upgrading the memory bus every couple of years. This is easy to do, because peripherals don't care what happens on the other side of the northbridge. The video card was handled adequately by the hack that is AGP (64-bit 66 MHz PCI would have been a much better idea, but that wouldn't have given Intel its nice AGP port to license).

The only peripheral that *might* be a problem in the future will be the network card (when gigabit cards finally come into vogue), and that will probably be what forces motherboard makers to put wider/faster PCI on to midrange boards and not just high-end boards.

In summary, this is less of a problem than it first appears to be.

The only serious bottleneck for performance is RAM latency, and that's not because of legacy peripherals.

Re:Peripheral communication.

[Christopher Thomas]

4X AGP is a 32-bit 266 MHz bus. That's more throughput than possible with PCI.

Unless you buy into Intel's PCI-X, which is 64/133.

And most graphics cards are not limited by bus bandwidth with *any* flavour of AGP (see the various Tom's Hardware benchmarks). The usual limit is fill rate for new cards, and lack of geometry processing for old cards (assuming you're playing a new game). Textures are stored on-card by any sane game, so the only thing going across the bus is lists of triangles.

AGP doesn't have contention with other devices on the bus so it doesn't have to do any logic for mastering or controlling and can allocate all its clocks to doing a data transfer.

While this would be an issue for very short data transfers, graphics cards will likely be transferring large batches of data. This is done in burst mode, which gives one transfer per clock.

Why would you want PCI? The only advantage PCI gives is that you can hang multiple devices off of it. But while that lets you get multiple monitor support easier, it will really kill your limited bandwidth.

You have bandwidth to spare; all you'd be doing in a multi-monitor setup is sending the same triangle lists over the bus, not cutting and pasting image data or doing texturing. Have one one dominant card and leave the others snooping traffic, and you have zero extra overhead for this.

The real benefit of having multiple video cards is that it lets you easily do render farming for things like games. Have each card render half the screen, and copy all cards' partial renderings to one card's frame buffer. 32/33 PCI is too slow to be practical for this, but 64/66 has more than enough bandwidth. I studied the feasibility of this at one of my past jobs.

Of course it's Jtanium! (Also, Ptanium, Netanium)

[leonbrooks]

Then instead of VLIW it could run Java bytecode, and Java pages would load under Netscape in less than an hour.

Alternatively, as Intel seem to like skipping letters, you could have a Pytanium that ran Python bytecode - or just ran the ASCII Python source.

Naturally, Microsoft would like a Netanium that ran CLR. The licencing costs would quadruple the cost of the CPU, which from Microsoft's POV is ideal since they can give Windows away, live off the CLR licencing costs built into the CPU, and claim to no longer be driving up the price of commodity systems with an expensive OS.

As in, you're lucky if it does stuff?

[leonbrooks]

They could name their chip Assium!

The ultimate in speculative execution? If it turns out right every time, it's ``pure ass''? Sounds very quantum to me...

Fat binaries

[leonbrooks]

When Apple transitioned from the M68K line to the PPC, they were in the same situation - 68K code would run faster on a 40Mhz 68040 than on a 40Mhz PPC 601. The reason consumers didn't mind was that the the PPC 601 started at 60Mhz (approximately the break-even point to the emulation layer), and (to the end user) didn't cost significantly more.

There's a bit more to it than that. Apple also strongly encouraged developers to ship ``fat binaries'', that is, the executable component of programs came in two ``forks'', one for the 68K and one for the PPC, made together from one set of sources. This meant that you could write stuff for the PPC that would still run on the 68K, or from another POV, simply recompile your 68K stuff and be able to also run it on a PPC.

Linux (or at least the bulk of the gcc-related stuff) is set up to enable this kind of thing fairly simply. The same piece of loader that decides what to do with Java executables and the like would be able to select code for different CPUs from one binary. Careless application implementation would involve a significant memory and performance hit, but it would still work. Windows, on the other hand...

It would be kind of neat to have a truly asymmetric multi-processor system and be able to run x86/Alpha/MIPS/Sparc/PPC binaries on the one Linux box without blinking. I wouldn't want to be the one sorting out the data structures common to all of the kernels, though.

BTW, does anyone remember the Dec Rainbow (PC-100)? Imagine how much better that would have fared if .COM (Z80/8080 on CP/M-80) and .CMD (808[68] on CP/M-86 or MS-DOS) files were able to be run seamlessly on the appropriate CPU.

Re:*cough* PowerPC *cough*

[ToLu the Happy Furby]

Also featuring stinking fast floating point.

Completely wrong. The G4 has some of the slowest IEEE floating point this side of a StrongARM.

Presumably you're confusing AltiVec with "floating point". AltiVec is a vector unit, not a floating point unit.

True, many AltiVec operations operate on vectors of floating point values (don't think they're IEEE, though), but that is most certainly not the same thing as normal floating point performance. Only a small subset of all floating point calculations can be effectively vectorized, and doing so requires extensive reprogramming (not just recompilation).

In any case most real-world floating-point applications have heavy bandwidth requirements and large datasets, which brings us to another glaring weakness of current Macs, namely their paltry PC-133 memory bus. Sticking two 1 GHz G4s on the same shared 64-bit PC-133 bus is almost comedy. (Tragicomedy if you want to run serious HPC workloads.) Sure the L3 helps if your dataset is 2MB and well-behaved, but if you think PC-133 is still ok for a modern desktop PC, perhaps you ought to think about *why* Apple and Moto had to start adding L3 cache and a backside bus to each processor when every other desktop CPU gets by with much cheaper on-chip L2 and a modern memory bus.

(Yes, I fully expect that DDR G4s will be available real soon now, but until then the situation on the high-end of Apple's desktop is just embarrassing. As for floating point performance, ditch the 64-bit PC-133 shared bus for a 128-bit PC2100 memory system and P2P interconnect and then that dual-GHz G4 might start looking credible...if only the G4 had decent fp number-crunching power.)

Bwa Ha ha ha ha ha ...

[BadlandZ]

Sparc is dead, Sun doesn't have the money (more than US $1B we'll guess) to do another round

Someone remind me to post a link back to this story in a month or two when Sun announces their faster processors with solved ecache solutions...

Who's paying for this researcher?

[spinlocked]

Fud, fud, fud. I can't speak for the other companies but Sun can easily afford to fund R&D on the next generation SPARC chip, they've got 6 billion $ cash in hand. Let alone investments, and have done for over 2 years. BTW the current generation is UltraSPARCIII, UltraSPARCIV is just a fabrication improvement. Work is already underway on UltraSPARCV's design. Sun's crown jewels are SPARC/Solaris, when Sun stops working on their own OS/CPU/Server platform it's time to stop investing in them.

Re:Who's paying for this researcher?

[Anne_Nonymous]

1. You're linking to data that's six months old.
2. Even if you look at that data, you're high by over $4 billion.
3. As of 12/31/01 they had $1.205 billion in cash and near cash, and $0.648 billion in Marketable securities for a total of $1.853 billion in cash-esque assets.

http://www.sec.gov/Archives/edgar/data/709519/00 00 89161802000605/f79125e10-q.htm#003

Re: Wrong again

[pdp11e]

The speed of the signal propagation trough the medium is not equal to the speed of electrons. The actual speed (group velocity) depends on properties of the "transmission line". For the good old coax cable it is about 0.66% of the speed of light. It is obvious that electrons in the coax cable do not even remotely approach that velocity but, the fact remains that signal travels ~20 cm during period of 1 ns. The actual fraction of c for the lines on silicon chip is very similar to the previous example.
When you discuss physics it helps if you know basic facts.

Re:Dead? I doubt it.

[BlackStar]

Very true. I type corrected, and thanks for the clarification!

Re:NDA Schmen DA

[AtariDatacenter]

I'll go ahead and take this.

No ecache problems ever again... because they're going to fully mirrored. That was really a black eye for Sun that they don't want to repeat. BTW, they're saying one of the bad batches of L2 cache came from IBM.

Over 5Ghz running inside Sun? I suppose it is possible, but not an everyday event. They will absolutely break the 1Ghz barrier this year if they haven't already. 1050 should be the clock speed (150mhz bus x 7 multiplier).

Sun playing dead? Yes. We should have seen the 1Ghz processors a long time ago. Soon after the ecache problem surfaced, new technology enhancements almost completely stopped. Only now are we seeing stuff again (and at a cautious pace) now that the UltraSPARC III is shipping.

None of this is really inside stuff if you've got a finger on Sun's pulse. (Like a major customer.) Here are a few more things that are easily deduced:

The UltraSPARC IV should be compatible with the UltraSPARC III. Same line of servers and all. Even numbered CPUs at Sun are improvements over the odd numbers. Not a core redesign.

The UltraSPARC V will more than likely require new hardware, just like the UltraSPARC III did.

Not to rain on your parade...

[Chris Burke]

This is exactly why 'virtual machines' (VM) or 'Just In Time' (JIT) compilers will eventually replace the current series of compile to asm compilers.

Actually... Java/.NET and JIT compilers are exactly why "Merced" or "Itanic" isn't well suited for the very things it was supposed to be good at. You see, for a VLIW machine like those, the degree of compiler optimization required to achieve good performance is much greater than for a traditional RISC-ish machine (in which I'm including x86, for reasons I'm not going into). Essentially, to get maximum performance requires a great deal of compilation, profiling, and compiling again. This is all front-end overhead on your process. The whole idea behind JIT is that it's supposed to be fast, and occure when you download new code... But now the opposite is true. At this point, you're just as well off using a traditional-style compiler/profiler that produce traditional binaries.

Sorry. No VM utopia here.

Re:Not to rain on your parade...

[Chris Burke]

Which are all good features... for some architecture that isn't IA-64. Plus, optimizations based on number of processors or amount of RAM are not going to be something the -compiler- does, but something the -programmer- does.

Re:Not to rain on your parade...

[Chris Burke]

Well this is exactly what Crusoe does.

We're talking about IA-64 here, which isn't exactly what Crusoe does. IA-32 -- and Crusoe's underlying instruction set -- are much less dependent on compiler optimizations for performance.

Emulation, code translation...

[Nyarly]

I've seen a couple of threads devolve into "this proc emulates that one...", so I wanted to drop a question in this forum.

I've long wondered what the insurmountable hurdles were to convert the machine language of one processor were to any other. Should it not be possible to abstract the results of instructions and convert between instruction sets?

I can see there being a huge time complexity argument, since the algorithm I have in mind runs something like: translate instructions from source to a base language (say, a modified version of MMIX, or whatnot), idealy designed as a RISC, where there are no convenience instructions - every instruction is atomic (by which I mean not the usual "is completed all at once" but "can't be split into other instructions) - and then regather translation language instructions into destination language instructions. That "gathering" process is, I'm sure, analogous to some solved problem, and might even be NP-complete.

I can see a few difficulties, such as differences in registers (both in number and roles), byte-widths, word lengths, bus function and memory size, but I see those primarily as features (non multiple byte width conversion, allowed?) or run-time checks (oops, that program needs more memory than the target machine can address). But the biggest issue would be the time complexity of the problem.

But if the only workable algorithms are NP, so what? I mean, the concept here is that I write a set of rules for converting to and from any proc, and you can write your code for anything (even the JVM...) and have it be converted into nice code for everything. It might take a while, but at one run per app, there aren't as many actual instances of the problem, if it is large and complex, you know?

Better still, I wouldn't need the source for the original code. So I could conceivably run Windoze through this code, and get an app that'll run on whatever proc I want.

So, the question is: why isn't this feasible, or if it is, why hasn't it been done?

Re:Emulation, code translation...

[Nyarly]

For stuff like NP-completeness, and Turing machines, etc, the dicipline you're looking for is call "Theory of Computation" by most universities. The Lewis & Papdimitriou textbook is decent and comprehensible. It's one of the few courses, though, that I'd say is difficult to screw up, so I'd almost say any inexpensive text on the subject would suit your needs.

Hell, this might be a good use for the Slashdot journal. I mean, what use is a CS degree if I can't inflict my misunderstanding of mathematics on the uneducated?

Make you a deal, AC. Get a Slashdot login, and I'll start adding journal entries regarding Computation.

Re:Sparc dead..? One word....

[rtaylor]

Funny.. Hardware is the only reason we buy sparc. Not thousands mind you, but several hundred.

Anyway, any and all software we write is developed under another unix (bsd, linux, osx) then moved to Solaris for testing and final implementation.

So.. The only reason those Suns are around is the ability to handle a load of 1000 which still being reasonably responsive. Most of that reason is the hardware.

Development hopes someday that research will switch the OS to something a little more reasonable. I want tar with gzip built in it damnit!

Re:Sorry...you're wrong..

[rtaylor]

If you look at how Boeing got into the market in the first place it's not out of line to consider a company like the above mentioned knocking them out.

Boeing dropped everything on a huge risk called the Jet. Make or break. If Bombardier were to drop trains, planes, snowmobiles, boats, etc. and make a 'kick ass' shuttle thats cheap to buy they could potentially take over most of the airplane market. Leaving Boeing in it's wake.

If Boeing remembers what they did, and can make that jump in technology (again -- like twinprop to jet) they'll corner the market for quite some time.

Bought them a 15 year lead last time, and took 'em from second (third?) in the market to first with a huge margin.

byte order

[Paul Jakma]

javierco is right, and you're full of the proverbial bull's faeces.

little/big endian makes no difference to performance. if anything, for networking little endian are at a disadvantage to big-endian machines as the likes of htons(),ntohs() are not no-ops.

similarly, big-endian machines might have to byte-swap little-endian data formats. (eg ext2/3, etc.) and PCI is little-endian, so data to/from PCI cpu has to be byte-swapped, but that's arbitrary. there is no inherent performance advantage to either byte order.

also, some CPUs can use either byte order, eg MIPS.

that said, nearly everything, is little-endian, simply because of the prevalence of i386.

UltraSparc RoadMap

[spiro_killglance]

AcesHardware as this info on the UltraSparc
RoadMap which looks pretty strong to me.

http://www.aceshardware.com/#55000446

With Sun's presentation here: http://www.sun.com/analyst2002/presentations/Shoem aker_WWAC_020702.pdf

An UltraSparc V running at 2.0-3.0GHz should
be very competative with Madision (McKinley II),
and it should have on chip multiprocessing, SMT or SMP (or both).

Re:Actually, 4gb is not enough

[roca]

The latest Everquest recommends 512MB of RAM. I think we're at most 3 years away from desktop applications (games) feeling constrained by having only 2 GB of addressable memory.

Re:Sorry...you're wrong..

[Paul Jakma]

the de Havilland Comet was the "lets take a huge risk and build a jet" trailblazer.

boeing et al learned from the Comets mistakes (rapid development of metal fatigue in fully pressurised aircraft being the major one).

where boeing did well was its 247 twin-engined airliner, which paved the way towards development of large four-engined aircraft, most famously the B-17.

Sounds like a trollish or clueless post.

[guacamole]

My world view is that Itanium based systems will become commodity products very quickly after good silicon is available in reasonable volume. At that point, why should one spend $8-10k for that hardware from the likes of HP, Compaq, Dell and others when one can build it for $2k (or even less)?

When peolpe start buying Itanium systems in volume, then the prices will drop on the Itanium systems. The reasons, they're expensive is not because the chips are hard to come by but because no one wants to buy them right now.

However, this comment alone makes me wonder about he posters cluelessness. He obviously hasn't worked in any real production environment. You people should realize that you simply can't build the kind of systems that Dell, HP, etc sell -today- out of commodity components. Take a look at a typical high-end SMP Dell server: propietary OEM motherboard, propietary case, hot-swap hard drives, hot-swap redundant power supplies and cooling, LOM support, etc. All components have been carefully designed to work together to produce a reliable, and scalable server system. You will never ever build the same kind of system on your own and if you do it's not going to be cheaper than buying one. Plus you don't get the vendor support.

The comment about SPARC being death is completely astonishing at the time when Sun is -THE- unix market leader. SPARC CPUs were never faster than the competition but that didn't worry Sun users as long as they were up to par with the competitors. The reason people buy Sun hardware is not the CPUs (CPU is alone is useless) but Solaris which is THE enterprise class OS and its applications, Sun's excellent support, massive multiprocessor scalability of Sun systems, massive I/O bandwidth, etc.

Current Sun chip is not bad at all (UltraSPARC III) and Sun is working on UltraSPARC V.

Re:Sounds like a trollish or clueless post.

[enkidu]

When peolpe start buying Itanium systems in volume, then the prices will drop on the Itanium systems. The reasons, they're expensive is not because the chips are hard to come by but because no one wants to buy them right now.

Uhmmm. Have you seen the die size specs for the Itanium? It's a freaking monster size chip (McKinley's >400mm2 I believe). Unless Intel is able to pull off an order of magnitude decrease in defects, the prices in Itaniums ain't ever going to drop compared to other CPU's (Sparc, MIPS, PowerPC etc.)

He obviously hasn't worked in any real production environment. You people should realize that you simply can't build the kind of systems that Dell, HP, etc sell -today- out of commodity components... All components have been carefully designed to work together to produce a reliable, and scalable server system. You will never ever build the same kind of system on your own and if you do it's not going to be cheaper than buying one. Plus you don't get the vendor support.

Really? Have you worked in a real production environment? The motherboard may be proprietary, but the chipset (90% of the development cost if not more) certainly isn't. AFAIK you can build your own with almost all of the features you need from off the shelf components. With vendor support also if you want. Checkout Penguin Computing if you want to see some solid server boxes made out of quality commodity parts and with vendor support.

The comment about SPARC being death is completely astonishing at the time when Sun is -THE- unix market leader. SPARC CPUs were never faster than the competition but that didn't worry Sun users as long as they were up to par with the competitors.

I don't know what you mean by "-THE- unix market leader"; but without objective criteria it's hard to say where you're coming from (unless it's Sun marketing literature). BTW, SPARC CPU's are not up to par with the competitors, especially with regards to floating point processing. In FP, PowerPC, Athlon, Pentium, all wipe the floor with the SPARC. And the Alpha still wipes the floor with all of *them*.

Re:Sun processors - buggy and late

[Multics]

Expect the next chip to BE LATE - as usual.

... and benchmark poorly on real jobs - as usual.

The reality of this is Sun's lost any claim on 'high performance' computing. E10,000 has just not got the bang. It especially doesn't have the bang compared to high end Intel or AMD CPUs. The Sun supporters need to run real jobs on E10k then run the same job on an Intel/AMD cluster that cost 10% as much. The differences are profound and not in Sun's direction.

Sun's last 5 years of CPU development have it way behind the performance curve. Can they catch up? Perhaps. I doubt it.

-- Multics

SPeed of ...

[leuk_he]

I think you are talking about the speed of electricty, which is much slower than the speed of light.

By the way the speed of light in matter (glass) is slower that the speed of light in vacuum.

And to answer your question: Yes.

think bigger

[NovaX]

Sorry this is so late, /. must have gone on their backup server, cuz my account stopped working and I couldn't post... oh well.

Your missing a huge number of issues. For one, these high end chips aren't usually put together by hobbiest and used in enterprise situations, like a PC is for a desktop. Here the branding is half of the importance - support contracts. The other half is the technology of the company - the O.S. and tools, the hardware solutions, and finally relative performance. HP made a killing for years at government labs through support contracts on UNIX workstations, and while the systems became ancient they stayed around due to the support until they had to be killed off.

HP, IBM, Compaq, etc. bring solutions to companies, which is what matters. The chip itself isn't to important, so the R&D gets to be a hassle making few players. The chip must simply have relative performance to the competition. The factors involved are then the scalability of the system, which involves the chipset. The chipset sets the system bus, memory type, multi-processing ability, etc. This is the 'solution,' which is what the companies are offering. Intel offers the core technology, but the system manufacturers bring in their own 'secret sauce' (as Paul DeMone called it) to the table. Above that, they bring the support, Operating System, and platform.

Therefore, while sure, you could build a workstation Itanium (but definately not for $2k), the big money are the clusters and the like. But the workstations will be built, and for a high end, high quality machine who would you trust, a brand like Dell with support contracts, or your neighbor? Its not commodoty yet, so few will turn to the guy down the street until prices drop significantly.

There are definately players left. The $1B mark R&D mark I assume is from how much Intel spent on the architecture, but that's on building an entire platform + obtaining corperate support. Sparc already exists and upgrading it is like doing the same to x86, it simply requires talent. Sun has mentioned both SMT and multi-cores in the future, and has spent work on their compiler in recent months (notice the big SPEC jump?). Don't count them out, their getting hit at the low end by PCs and the high by IBM. Its not Intel killing them just yet.

Lastly, some people on here seem to still believe Itanium is huge due to its core. In actuallity, the core is 25 million trasistors and rest is all cache. For the exact numbers, read the RWT forum, since I forget them offhand.

IA-64 wont rule the world due to expense, compilers, software platform, and where the market positions it. And lets not forget performance, different jobs need different skills and Itanium shines at FPU unneeded for desktops, fails at integer which they love. Vice-versa for many scientific apps. Intel wont distroy everyone else, but they'll certainly make a dent. Its already been felt.

Re:When will we finally have CLOCKLESS CPUs?

[Paul Komarek]

If I understand "clockless" cpus, wouldn't the analogy be a ballet with lots of bands waiting on dancers, and dancers waiting on each other?

-Paul Komarek

High tech: Commodity=Debranding, = Open?

[Paul Komarek]

After checking 3/5 pages of >=+2 comments, I got tired of checking if this opinion has already been stated. At risk of repeating someone else's post, here's my opinion.

If and when mainstream processors become true commodities (there's a good definition in here somewhere), the value of Intel's brand will vanish. This will reduce Intel's revenues, and hence R&D and capex budgets. Companies like AMD, who have always run their business on slim margins, will have a structural advantage over "fat" companies like Intel; this will allow them to survive.

The above scenario occurs when nobody has any cost-efficient means of stimulating the mainstream market for cpus. This will leave customization and services as the profitable parts of cpu design and manufacturing. It is easy to imagine that focus will turn toward other bottlenecks in system architecture. For example, suppose an effort is made to push parallel processing into mainstream computers, with multiple vendors create a multitude of approaches requiring quick and cheap cpu modifications. I expect that such a market will favor increasing openness in cpu design, just as the software markets are discovering the serviceability benefits of open source software.

Therefore, my prediction (woohoo, let's guarantee that I'll be completely wrong by predicting the future) is that cpu commoditization will lead to the prominence of open cpu designs. Because of the increasing success in open source software, this shift will occur somewhat quickly and less painfully than it has for operating systems. Of course there are major differences between open cpu cores and open source software, there will be plenty of exciting socioeconomic developments leading to profitable open cpu cores.

In the end, this could be a major win for established semi-open cores like SPARC. However, community-developed open cores like the FCPU and various efforts with open StrongARM implementations, will have a jump on building the community infrastructure that allows open markets to be efficient and useful.

That's my contribution on the subject. It won't be a technical change that we'll see next in the cpu market; rather it will be a socioeconomic change.

-Paul Komarek

Re:What's Next?

[mikera]

Or.... you can subdivide the 128-bit pointer into two 64-bit chinks. Use one as the primary pointer, like you say we don't need 128-bit addressing for a while yet. And then use the *second* 64-bit pointer to implement some memory management technique. Rather like the car and cdr parts of a cons cell in LISP, or something like that.

If you could implement garbage collection in the hardware, the performance of complex systems would rocket. Could also be a tremendous boon to things like functional programming languages.

128-bit chips might just be able to facilitate this. Don't knock it 'till you've tried it. I certainly look forward to having a play with one in 10-15 years time.

How was this a troll?

[Marx_Mrvelous]

Being modded "troll" means that they're jealous of your comment ;)

Re:SPARC Dead? I don't think so.

[MidKnight]

. . . 64 max CPUs together ( E15k will have more ), or 8 CPUs.

The E15k already has more: up to 106 processors. And you're right, the supposition that SPARC is dead simply because of the current financial state of Sun is pretty silly.

Even more silly to me is the fact that people are still equating Mhz to the overall value of a chip. And before people start whining, I'm not talking about the "Megahertz Myth", either. Currently, per-chip computing power is dirt cheap. It seems to me that factors like scalability, power consumption, and (as you mentioned) error checking are quickly becoming more important in general-task CPU's.

--Mid

Re:Remeber what about the Pentium Pros?

[Talonius]

Obvious you can read.

I didn't say the PPro was a technical failure. I said it was a MARKETING failure.

Before you spout off at the mouth with random insults - READ.

I have a Pentium Pro 200 x 2 running in my office right now and I do like it. IT WAS NOT SUCCESSFULLY RECEIVED AT THE RETAIL LEVEL.

End of story.

Re:Sun processors - buggy and late

[BadlandZ]

"Let's see... the last chip was only 2 years behind schedule."

I'll wait and see. I sort of know more than I should, and more than I should say... But, let's just wait and see ;-)

Multiprocessors!

[morcheeba]

There is a large barrier to entry for designing huge processors, but I think processor startups will always exist. With resources a small fraction of the big guys, they can't design and manage a device this complex. Instead of designing a 25 million transistor beast, people are designing much smaller processors, making sure that each can network well, and then populating chips with multiple copies of these processors.

Startup Pact (with a staff of only 30!) has designed a 30 million transistor chip by iterating a 200k transistor processor 128 times, yielding a theoretical maximum 12.8 billion MACs/sec @ 100 MHz. Of course, that's the theoretical max... it will take the correct problem and/or some good programming to get anywhere near that maximum.

Lexra has put 16 MIPS32 processors on a single die. Again, with only 30 employees.

How many of all those fabs do .13 micron?

[Mike Greaves]

Answer that and you'll just have shot down your entire first paragraph. There are almost no CPU fabs at .13 micron, outside of the PC market. Are there *any* CPU's at .13, other than Power4 and x86?

Your second and third paragraphs are essentially agreeing with me - there is on-going, and will be more, fab consolidation...

Re:You work for a research firm?

[Wakko Warner]

Like you could with the Pentium Pro?

Stupid fucking retard.

- A.P.

Java bytecode native to CPU

[leonbrooks]

here's a link to Jazelle, the (90% of) Java bytecode native to CPU trick

Excellent! WeirdX might run in realtime... (-: