Dell Dumping Itanium

njcoder writes "In a PC World article it is disclosed and confirmed by Intel that Dell is dropping support for Itanium processors. 'After Advanced Micro Devices demonstrated that 64-bit extensions to the x86 instruction set offered a smoother transition to 64-bit computing, Intel released a version of Xeon with similar technology, and Dell now offers 64-bit Xeon processors across its product line.'" More from the article: "The chip maker has since backed off its original statements about Itanium and is now promoting the chip as a high-performance replacement for reduced instruction set computing (RISC) processors in Unix servers from companies such as Sun Microsystems and IBM. Hewlett-Packard, a co-designer of the processor, has embraced Itanium as the processor of choice for its high-end servers. Fujitsu. and NEC are also among the system vendors that sell servers with the processor." The story is also being reported at Ars Technica.

Comments

I Blame Sun Microsystems

[AKAImBatman]

Guess McNeally got under Michael's skin. :-P

Re:I Blame Sun Microsystems

[Markus_UW]

Good job, Scotty!

I must say, that was a most entertaining ad campaign, and Sun's new servers really do seem to be really, really good. It'd be nice to see Dell go down in the entry-level server space.

Re:I Blame Sun Microsystems

[Intron]

I suspect they are just following IBM's lead six months later on this one. Box makers have to go with what their customers want. Its too easy to switch vendors if someone else gives you better price/performance/features.

Exactly what *is* the Dell aversion to AMD?

[popo]

One has to wonder, outside the obvious explanation of Intel's anti-competitive trade practices, what is Dell's aversion to AMD 64-bit / dual-core processors?

Clearly there is significant (and growing) demand for Opterons.

Dell's outright refusal to offer AMD chips seems almost like proof of itself that Intel is acting in an anti-competitive manner.

Has Dell ever put forth a better explanation?

Re:Exactly what *is* the Dell aversion to AMD?

[js3]

or maybe intel offers a better deal for them and they are greedy bstards?

Re:Exactly what *is* the Dell aversion to AMD?

[mihalis]

One has to wonder, outside the obvious explanation of Intel's anti-competitive trade practices, what is Dell's aversion to AMD 64-bit / dual-core processors?

I think that Intel gives a slightly better volume discount to Dell than anybody else. Partly this is because Dell's volume is bigger than most anyone else (I forget if they have exceeded HP yet), but the obvious suspicion is that there is also an "exclusivity bonus" - yet lower prices for a vendor who does not sell any of the competition's products. If Dell actually sold AMD Opteron based products, I suspect they would do very well on those products, but if they drove up their costs on every other system they sell, all still containing Intel cpus, then it might be a net loss, at least initially.

Re:Exactly what *is* the Dell aversion to AMD?

[nrgy]

If you read some of their press releases they often make the hilarious comment "Are customers just aren't asking for them". But then browseing their store you can find tucked away mostly hidden links to Opteron systems for sale. So Dell which is it?

Re:Exactly what *is* the Dell aversion to AMD?

[popo]

IMHO that's not an acceptable explanation for offering zero AMD servers.

Its not as if Dell sells AMD servers at a higher price. Clearly there is an enormous amount of demand for Opterons. All the market metrics show Opterons taking a larger and larger piece of the server market. Dell's server business is hurting as a result, and still they offer no AMD machines.

Furthermore, if as you say "Intel offers a better deal" -- and that deal was based upon exclusivity. (In other words: "You get a 15% discount if you sell only Intel chips"), It seems to me that that would be illegal and anti-competitive.

Re:Exactly what *is* the Dell aversion to AMD?

[squison]

Just because one company doesn't buy from the competition doesn't mean Intel is anti-competitive. It just means they won over AMD in this case. It's like saying Ford is anti-competitive because you don't own a Chevy.

Re:Exactly what *is* the Dell aversion to AMD?

[truesaer]

One has to wonder, outside the obvious explanation of Intel's anti-competitive trade practices, what is Dell's aversion to AMD 64-bit / dual-core processors?

I think that Intel gives a slightly better volume discount to Dell than anybody else. Partly this is because Dell's volume is bigger than most anyone else (I forget if they have exceeded HP yet), but the obvious suspicion is that there is also an "exclusivity bonus" - yet lower prices for a vendor who does not sell any of the competition's products. If Dell actually sold AMD Opteron based products, I suspect they would do very well on those products, but if they drove up their costs on every other system they sell, all still containing Intel cpus, then it might be a net loss, at least initially.

That is exactly the allegation AMD seems to be making in their lawsuit. From a monopolist, an exclusivity bonus is illegal.

Re:Exactly what *is* the Dell aversion to AMD?

[popo]

No its not like saying that at all.

If Coke said to your local supermarket: "We'll give you a 10% discount if you don't stock Pepsi -- even though Pepsi represents 36% of the market", that would be anti-competitive behaviour.

The case already went before the Japanese Trade Comission and AMD won. I have a hard time believing that the scenario is different in the US.

Re:Exactly what *is* the Dell aversion to AMD?

[supra]

After Advanced Micro Devices demonstrated that 64-bit extensions to the x86 instruction set offered a smoother transition to 64-bit computing, Intel released a version of Xeon with similar technology, and Dell now offers 64-bit Xeon processors across its product line.

You'd think that having AMD "demonstrate that 64-bit extensions to the x86 instruction set offered a smoother transition to 64-bit computing", the likes of Dell would realize where the real innovation is coming from and begin diversifying. It's almost as if there's some other agenda behind Dell's behavior.

Sniff-sniff...smell that? ... smells like a long-overdue anti-trust suit in the works.

Re:Exactly what *is* the Dell aversion to AMD?

[popo]

Because ultimately anti-competitive behavior *always* hurts the consumer.

Just when you think you're winning on pricing, it usually turns out you're losing on innovation.

Re:Exactly what *is* the Dell aversion to AMD?

[$RANDOMLUSER]

Close.
Coke is saying "We'll give you a better discount than any of your competitors if you don't sell Pepsi".
If I can sell Coke cheaper than any one else, people will go out of thier way to buy from me.

Re:Exactly what *is* the Dell aversion to AMD?

[popo]

... right, but either way its the "if you don't sell Pepsi" part that's the illegal bit.

Re:Exactly what *is* the Dell aversion to AMD?

[NimNar]

A recent story on slashdot mentioned that Intel's production cost per chip is $40. Retail price can be as high as 10 times that. The answer to your question is clear.

AMD is offering Dell tremendous discounts on their chips--probably selling to Dell at prices other PC assemblers can only dream about. This allows Dell to sell at rock-bottom prices. For Dell and Intel this creates a "virtuous circle" where Dell builds increasing market share using only Intel chips.

I am sure Intel will offer Dell "whatever" it takes to keep the exclusive arrangement. And Dell will always be able to undercut the competition and increase market share.

Re:Exactly what *is* the Dell aversion to AMD?

[js3]

well it is up to dell to take the deal or not, that's why I say they are greedy. If someone offered me an exclusive deal (the key word here being offered not forced) then my greediness would decide whether to take it or not.

Re:Exactly what *is* the Dell aversion to AMD?

[Dun Malg]

That is exactly the allegation AMD seems to be making in their lawsuit. From a monopolist, an exclusivity bonus is illegal.

Indeed, as an "exclusivity bonus" is just a nice way of saying there's a penalty for giving their competitor any business.

Re:Exactly what *is* the Dell aversion to AMD?

[ZuggZugg]

While I'm a fan of the Opteron and AMD in general, Dell's decision makes sense to me.

Not knowing the explicit details of the Intel / Dell relationship there must be financial incentives for Dell to remain loyal to their partner Intel. And they don't necessarily have to be illegal or anti-competitive. Fact of the matter is that Dell hands off a lot of R&D to partners like Intel. It's doubtful that they would get similar levels of R&D support from AMD. Secondly there is a cost to doing business with more than one supplier, vendor and contract negotiations are not trivial matters, especially when dealing in the volume that Dell has.

I suspect one of the main deterrants for Dell has been availability of supply from AMD...AMD to this day struggles to supply it's customers with chips...something that Dell is not used to as numero uno with Intel.

Re:Exactly what *is* the Dell aversion to AMD?

[popo]

> "I am sure Intel will offer Dell "whatever" it takes to keep the exclusive arrangement"

What you're describing is illegal anti-competitive behaviour.

Re:Exactly what *is* the Dell aversion to AMD?

[vanango]

I don't think it's an aversion towards AMD, however they benefit more from Intel.

Intel uses a very effective pull marketing strategy. Most consumers don't care about what processor they have. They don't understand it and they don't care to learn. However after their commercials, the name is known which makes the name sound reputable. Because of their "Intel inside" advertising, people aren't going out and buying processors, but demanding them from retailers.

In addition, humans get confused easily. If you have 3 options: basic, standard, advanced let's say, the decision is easy. You know what youw ant and what you need and how much you can spend. When you throw in too many alternatives, people hestiate. Every moment a customer is hesitating and not purchasing, companies lose money. Now obviously there are issues with customibility, but giving customers too many alternatives can be a bad thing. Take Apple for example, why did they replace the mini with the nano? Was colored iPods a good idea? Is there a reason the nano is offered (as of yet) in 2 colors?

Now imagine Dell's advantages to exlusivity. Ever additional product they offer, they lose money. It's costly in advertising, in inventory, in maintenence, in troubleshooting, the list goes on. Also as mentioned, they can buy more intel chips creating economies of scale for both companies.

And of COURSE they are greedy bastards, it's a corporation. An American one out for themselves, hell, have you tried talking to Dell on the phone?

Re:Exactly what *is* the Dell aversion to AMD?

[ring-eldest]

Furthermore, if as you say "Intel offers a better deal" -- and that deal was based upon exclusivity. (In other words: "You get a 15% discount if you sell only Intel chips"), It seems to me that that would be illegal and anti-competitive.

I don't think that word means what you think it means. You seem to be viewing "competition" as something that would somehow benefit you, while sticking it to the man. Kind of like a chicken debating the ethics of being fried or baked; you're still cooked at the end of the day, fella.

Exclusivity clauses suck. They're bad for pretty much everyone involved except anyone getting paid off to make it happen. The customers get screwed, along with the vendor (who now has a forced audience of people who hate their sucky service [no incentive to provide good service = suckage] along with the vendor's host who has to put up with the same complaints and sucky service or risk getting sued for violating their contract.

They suck. But they're not illegal. Just ask the cunts at Aramark who own exclusive rights to sell food on many a college campus. Or banks, who fight over exclusive right to provide banking services. Or the classic coke / pepsi vending wars going on just about everywhere. Try buying a coke at Taco Bell or a Pepsi at McDonald's sometime and let us know how that goes.

Re:Exactly what *is* the Dell aversion to AMD?

[Richard_at_work]

Explain the fact that I cant buy a Pepsi at any McDonalds anywhere, but I can get Coke. Exclusivity is not illegal, predatory pricing that leads to exclusivity is.

Re:Exactly what *is* the Dell aversion to AMD?

[Doctor Memory]

IMHO that's not an acceptable explanation for offering zero AMD servers.

Well, then, you're a retard. Why would someone give up a N-where-N-is-large% discount from the supplier that will provide 95% of your processors just so you can sell 5% of your volume with processors from another vendor? How are you going to explain to your shareholders that you're going to raise production costs by millions of dollars just so you can do a couple million dollars more in sales? Do you honestly think that AMD is able to offer Dell better prices than Intel?

And let's not forget that Opterons, despite "taking a larger and larger piece of the server market" still have a less than 12% share. If Dell's going to tweak Intel's nose, it would make more sense for them to go with IBM's Power line of processors for the high-end. IBM's doing well with that line; unlike HP, whose Itanium sales haven't managed to outpace PA-RISC declines.

I like Opterons (I have a Sun Ultra 20 at home), but suggesting that Dell destroy its margins just so they can offer a processor that's more appealing to you (I'm guessing you're not a Fortune 100 CIO planning to outfit a dozen or so new data centers) is just stupid.

Re:Exactly what *is* the Dell aversion to AMD?

[lawpoop]

Coke and pepsi do this all the time. Go into McDonald's and you will find only Coke products. Go into Taco Bell and you will find only Pepsi. At the Ohio State University, you will find only Coke machines, and the hot dog vendors sell only coke products. Same for the Ohio State Fair.

Re:Exactly what *is* the Dell aversion to AMD?

[numark]

While Intel doesn't specifically say to companies, "Use only our chips and we'll give you a discount," they seem to hide it poorly. AMD even has lawsuit out for anticompetitive practices against Intel, saying that Intel repeatedly interfered with AMD's ability to compete.

If a company made any effort to use AMD's products, some companies have reported that Intel would "suddenly" run out of important server-class chips to ship out to them, and marketing incentive payments would dry up and not be paid. Intel allegedly even did direct payments to certain companies (including, allegedly, Dell) in order to shelf AMD-based products that were under development. So, as you can see, there are some serious questions about Intel's conduct in dealing with other companies.

Re:Exactly what *is* the Dell aversion to AMD?

[Anonymous Struct]

We buy a lot of Dell hardware, and we've asked this question every single year for the last 3 years. Every time, the answer has been that AMD can't provide chips fast enough to make them viable for distribution. Basically, they don't feel like they'd be able to meet demand, and they'd have customers waiting on hardware that may or may not be coming some time soon. That's what they tell us, at least.

Re:Exactly what *is* the Dell aversion to AMD?

[Austerity Empowers]

It's really easy, contrary to popular opinion Dell is not only a system integrator, but a HW mfg, and works pretty closely with Intel on building the latest and greatest boxes. Usually when Intel announces a new chip, Dell announces a new machine using it. That same relationship does not exist with AMD and would cost money to develop. Why would Dell develop it? Only if there is a considerable demand for AMD devices such that Dell lost enough sales to warrant it.

So far, that is not happening. I for one would like to see Dell sell AMD, but John Q. public doesn't care much (hence www.dell.com PCs won't have em), and enterprise level purchasers don't seem to care. Maybe that will change as word gets out but for Dell to change I think it almost has to already be happening.

Re:Exactly what *is* the Dell aversion to AMD?

[Doctor Memory]

Ah, but it's only an "opportunity cost" sort of penalty. Intel isn't threatening to raise Dell's costs above those it charges anyone else, or cut off their supply, or only sell them low-end chips. They're just offering them better terms than they can get elsewhere, which Dell is free to decline. And I'm sure if it ever makes economic sense for them to do so, they will.

I once spent a long afternoon eating ice cream, because the proprietor of the shop (my uncle) told me I could have as much as I wanted as long as I ate it while standing directly in front of the shop. All things being equal, I would rather have had a French vanilla cone from the shop down the road, but things weren't equal and free won big time.

Re:Exactly what *is* the Dell aversion to AMD?

[bhny]

find tucked away mostly hidden links to Opteron systems for sale

no you can't
where is this link?

Re:Exactly what *is* the Dell aversion to AMD?

[alc6379]

Please link to these claims; otherwise they are unsubstantiated. Thanks.

Re:Exactly what *is* the Dell aversion to AMD?

[MickoZ]

If you get the same amount of client, and sell the same price and get the same satisfaction (from your target audience), then it is not wrong at all to be greedy. That is how you make more money sometime. Except if in the long run those choice turn out to give you a bad reputation.

Re:Exactly what *is* the Dell aversion to AMD?

[corngrower]

Are you planning on purchasing any Opteron systems? I'm sure those folks at Dell would be interested in knowing. Their other story is that customers aren't demanding AMD processors in products.

Re:Exactly what *is* the Dell aversion to AMD?

[rainman_bc]

The hole in the Coke v. Pepsi argument is that McDonalds would be required to have twice the amount of fountain dispensers, placing burden upon them. They just simply don't have the room for both dispensers.

Notice that 7-Eleven and other convenience stores frequently have both brands in their coolers because of the low overhead costs? You'll be hard pressed to find any company with only type of bottled beverage. They all carry both.

Re:Exactly what *is* the Dell aversion to AMD?

[Glasswire]

If a company made any effort to use AMD's products, some companies have reported that Intel would "suddenly" run out of important server-class chips to ship out to them, and marketing incentive payments would dry up and not be paid.

Funny, HP has a very broad line of Opteron systems, but, as far as I can see, has no shortage of Intel processors and is probaly getting getting their incentives pretty much as usual because HP sells a lot more Xeon-systems than Opteron-systems and neither HP nor Intel would want to damage that mutually beneficial realationship.
The real reason Dell uses Intel, is probably that AMD can't deliver the quantitiy of product Dell needs to introduce a server model as a sucessful product. Other vendors can have niche products which they don't expect to sell as much of, but that doesn't work for Dell. Itanium is actually growing in market share year on year in the big system space - BUT it still wasn't generating the volume of business that works with Dell's model.

Re:Exactly what *is* the Dell aversion to AMD?

[furrywithwings]

It's all over the net if you look. The main reason factor is supply capacity. Dell buys a stupid huge number of things at once, and the last article i read said something like some crazy % of amd's entire production would have to go to dell. ;)

Re:Exactly what *is* the Dell aversion to AMD?

[evilviper]

Every time, the answer has been that AMD can't provide chips fast enough to make them viable for distribution. Basically, they don't feel like they'd be able to meet demand,

Dell seems to be quite good convincing people that this is true, but it's certainly not.

http://slashdot.org/comments.pl?sid=158117&cid=132 47676

Re:Fair play

[grub]

Think back many years to when AMD was known as "Advanced Micro Devices" and made Intel compatible chips. That's why we had the "Pentium(tm)" and not "586" being marketted.

Don't forget SGI

[Anonymous Coward]

SGI uses Itanium for their Altix line of products that run Linux. They need Itanuim for its ability to handle hundreds of processors in one system with cc:NUMA, and its huge physical address space for their customers who need several terabytes of RAM in one system.

Re:Don't forget SGI

[Anonymous Coward]

Somehow this reminds me of when George Bush said - "Don't forget Poland!" when listing the allies in the Iraq war.

Re:Don't forget SGI

[merreborn]

Don't forget SGI

I did. Is that bad?

Seriously, I remember them being the biggest name in graphics back in '96. I thought they were dead and gone.

Re:Don't forget SGI

[timeOday]

Oh yeah, I do remeber SGI... they lost their shirts trying to sell very expensive, exotic hardware even after commodity hardware caught up and then surpassed their niche offerings.

But that was graphics hardware and this time it's CPUs, so I'm sure things will turn out just fine for them.

Re:Don't forget SGI

[turgid]

SGI uses Itanium for their Altix line of products that run Linux. They need Itanuim for its ability to handle hundreds of processors in one system with cc:NUMA, and its huge physical address space for their customers who need several terabytes of RAM in one system.

Strange, that.

SGI once had its own line of very sophisticated and top-performing 64-RISC processors called MIPS that did exactly that until they drank the intel itanic Kool Aide. (And Windows NT as well, but that's a whole nother rant).

Look at poor SGI now.

I blame Intel

[vlad_petric]

For not making Itanium competitive enough ...

Re:Hmmm.

[markass530]

I kinda like it. Something products that I am loyal to, I am loyal to the point of it being like a sport team. As a niner fan, I love it when the raiders loose, or for that matter when anyone makes fun of raiders, or they look stupid for any reason. As a long time fan AMD Products (As a broke 18 year old with an aging 166MMX, the k6-2 380 Was a godsend) I enjoy the adds. Same way as a pepsi fan, I like the commercials that poke fun at coke.

Ummmm ....

[gstoddart]

Isn't everyone dumping Itanium? Why is Dell any different?

Re:Hmmm.

[bradleycarpenter]

Yikes....look like someone brainwashed you real good! Ever try to innovate something on your own? Didn't think so.

Re:to the management

[Anita Coney]

It took me three seconds to find this one. "Front and rear IEEE 1394 (firewire) ports and rear S/PDIF port."

Writing has been on the wall

[akuma(x86)]

Itanium is all but dead... relegated to the supercomputer niche - and we all know what happens to supercomputer companies :)

Intel has spent billions on Itanium and seen an effective return of 0%. Investors won't tolerate this for much longer. AMD's x86-64, and Intel's subsequent introduction of EMT64 (same thing), have finally pushed this ill conceived idea into its well deserved death spiral.

It has no technical merit. But technical merit sometimes is a secondary matter in the business world. However, the economics don't make any sense - you can't introduce a new ISA into a mature software market and expect it to fly just because you're Intel.

It was a mistake - write it off and move on.

This should free Intel to deploy those valuable Itanium engineers (like the ex-Alpha team) to work on something that actually generates cash (like x86 servers). So while AMD might have a short term lead - the giant resources of Intel are more than enough to catch up and re-assert their leadership position.

Re:Writing has been on the wall

[Zemplar]

"It has no technical merit."

Excuse me? Modded +5 Interesting, but incorrect on this one point.

I hope you just made an honest mistake and don't really believe that nonsense.

Re:Writing has been on the wall

[TheRaven64]

This should free Intel to deploy those valuable Itanium engineers (like the ex-Alpha team) to work on something that actually generates cash (like x86 servers).

Many of the Alpha team went to work on the Opteron. Or did you think that it's resemblance to the Alpha was a coincidence?

And I don't really think Intel need to do much catching up. They are behind in the server market, and ahead in the laptop market. The server market is shrinking, and the laptop market is expanding - it sounds like they are exactly where they wish to be.

Re:Writing has been on the wall

[aaronl]

It is not the first time Intel took a gamble on a new ISA and lost. Hell, the very existence of x86 was as a filler move between non-x86 product lines. The ISA is one of the worst out there, and it *can't* get better.

It's a real loss to everyone that the industry won't just drop the damned architecture. MIPS64, PowerPC, IA64, Alpha, and UltraSPARC are vastly superior to x86. One of the reasons that all of those architectures have lower clock speeds than current x86 is because they still execute more instructions per cycle than Intel and AMD do on x86.

I can't say that I wanted Itanium to prevail, but I would've been happy with not running x86. At least us UNIX types have the option of not being stuck on that archaic architecture. Unfortunately, the majority of people out there run Windows, and they *are* stuck.

Re:Writing has been on the wall

[akuma(x86)]

Many of the Alpha team went to work on the Opteron. Or did you think that it's resemblance to the Alpha was a coincidence?

The fingerprints are there :) Sure, some of the Alpha engineers went to AMD. But there is still a core design team in Mass. that currently works for Intel. You didn't think that every single engineer just uprooted their family to move cross country to the Bay Area did you?

Intel doesn't just have the Alpha team. They also have the ex-PA-RISC team from HP. I hear they're not too shabby either :)

Bottom line is this - Intel has high quality, experienced engineers in quantities that exceed the competition by a wide margin.
This doesn't always guarantee success because management can always screw it up, but it's a pretty significant advantage.

The server market is shrinking, and the laptop market is expanding

Your data is incorrect. The server market is growing very quickly as is the laptop market. Hell, even the desktop market is growing (but at a much much slower pace).

Re:Writing has been on the wall

[akuma(x86)]

x86 processors have a fixed amount of decoder logic overhead vs. RISC. The decoders essentially dynamically translate the x86 instructions into more machine-friendly micro-ops which are very RISC-like.

As transistor budgets increase exponentially (thanks to Moore's law), that fixed overhead gets smaller and smaller each generation - to the point that it's insigificant (less than 5% today and getting smaller tomorrow). So in the early 90s you could make a case for more efficient computing with RISC vs. x86, but today it's just so negligable that you don't care. There are also numerous micro-architectural tricks to get around the limited registers and wacky addressing modes.

Couple this with the fact that 99% of all of the world's software is written for x86 and you have a very large inertia to overcome in order to change the ISA.

Why would any software vendor port their application to a new architecture if that architecture is brand new and nobody is using it initially? This is a very expensive and risky task. Let's say that the incentive is increased performance with a new ISA (highly unlikely given that the ISA doesn't matter anymore given the very large transitors budgets). But let's be generous and give it a 50% performance advantage (again - this is fantasy land). Do you spend the 8 months porting, debugging, testing Photoshop? Or do you just wait 8 months for a 50% faster x86 to come out and instead spend that time improving your product as opposed to keeping it the same on a different architecture?

You'd have to be crazy to take that tradeoff. And so, you see what we have today - x86 everywhere.

Re:Writing has been on the wall

[akuma(x86)]

I really do believe it has no technical merit.

I am paid to design processors and have worked on SPARC, MIPS and x86 designs for a span of over 12 years.
I spend my days thinking about how to improve processors. That's all I do... all day long.

So please... enlighten me on how the Itanium architecture improves computing on any metric.

Any performance advantage that you see today is solely due to their having much larger die size and pin count budgets vs. other processors just to compensate for their having a crappy ISA. If you give the same budget to a comparable x86 or traditional RISC processor, their absolute performance and performance/watt would far exceed any Itanium.

Put a 9MB cache on an Opteron and see how well it does on SPECFP for example.
An Opteron beats the Itanium 2 handily on integer code with just 1MB of cache.

Re:Writing has been on the wall

[rsynnott]

Apple would hardly be impressed...

Re:Writing has been on the wall

[Lemming Mark]

Would a 9Meg cache really help an SpecFP benchmark that much? I'd have guessed an Opteron's FP pipes would be permanently saturated with well less than the full 9MB on many benchmarks.

> So please... enlighten me on how the Itanium
> architecture improves computing on any metric.

Personally I like the predicate registers, conditional execution and rotating register files - a neat way to pipeline loops without unrolling. I like the concept of VLIW-style without huge amounts of nop padding (although it can't eliminate it). It also has pretty neat support for doing explicit cache prefetches in software.

Some of these have obviously been done before to varying degrees (but I've not heard of rotating register files in a general purpose chip). Whether or not the chip as a whole was worthwhile, I think the nice bits have technical merit and I'd expect some of them to pop up in other places.

Intel went through an interesting trend throughout the company of trying to push huge amounts of complexity out of the chip and into the compiler. This wasn't just Itanium: the IXP network processors have a *really* weird programming model (c.f. the much more conventional IBM designs). I occasionally wonder what caused *all* the architects at Intel HQ to go down this road on completely unrelated products (something in the water supply?). I must admit I *really* like the idea of pushing cleverness up the stack but if you're going to do it, you should provide *real* benefits.

Re:Writing has been on the wall

[CaptainPinko]

Some of these have obviously been done before to varying degrees (but I've not heard of rotating register files in a general purpose chip).

Isn't that way a Sparc does when the input registrars become output registrars during a function call?

Re:Writing has been on the wall

[multipart]

Yes, but the difference is that on Itanium you can explicitly tell the CPU to rotate its register file. The result effectively is register renaming, and it can be used for software pipelining. See e.g. slide 7 of http://www.gelato.org/pdf/Workshops/geneva05/compi ling_itanium_muthu.pdf.

Re:Writing has been on the wall

[Lemming Mark]

I think you're referring to Sparc's "register windows" that (I believe) are rather like IA64's "register stack". IA64 additionally has the ability to rename the registers on each iteration of a loop - that's the rotating register files. This helps you do software pipelining (e.g. when iterating in an inner loop, hide load latencies by preloading future data into a register before doing operations on the current data.) without needing to bloat the code by "unrolling" the loop.

Re:Writing has been on the wall

[imroy]

Couple this with the fact that 99% of all of the world's software is written for x86 and you have a very large inertia to overcome in order to change the ISA.

Really? Better tell the Debian guys then. They ship their distro for 10 platforms. Then there's the BSD's: NetBSD takes the cake with 49 platforms listed as stable. OpenBSD has 16 platforms and FreeBSD has 9 platforms.

I think you'll find that most software nowadays is written in a high-level language and not for a specific processor. If you have access to that source code and a compiler for your processor, then it doesn't really matter what it is. The big issue in porting is the operating system and GUI (and other) toolkits, not the CPU architecture.

Re:Writing has been on the wall

[TheLink]

Well, start listing the merits please.

All I see is a 200+ million transistor chip or even a 400+ million transistor chip, that can't do specint faster than x86 chips with 100+ million transistors (or even less). And only do specfp maybe 2 x faster.

And even then, I wonder if the specfp stuff the Itanium is so good at can be split up easily to multiple chips. In which case you might as well have two x86 chips using 200 million transistors (dual core).

Basically you get more performance per transistor for most tasks with x86 than you do with Itanium, unless you are doing tasks that are similar to a few of the specfp tests which the Itanium handles really really well.

I wonder if a DSP might be a good fit for those tho...

Re:Writing has been on the wall

[TheLink]

The advantage in practice is that x86 CISC instructions are often smaller than those of most RISC chips (except maybe POWER- which isn't really RISC ;) ).

Since the x86 CISC stuff is translated to RISC like stuff, you can think of it as on-the-fly decompression.

Compressed code (x86) is stored in slow main memory and caches, and decompressed to RISC ops in the CPU (with the P4 the risc ops are in the tracecache).

So the x86 isn't that bad. Since disk and RAM is slow, smaller programs are a good thing. Caches are expensive too.

Given CPUs were gaining in speed much faster than RAM, what might be faster than x86 ISA would be a better designed CISC ISA- one that worked well with the limited memory bandwidths etc.

But I hear POWER is practically a CISC - microcode and all...

Whatever it is, the x86 chips sure are fast enough for most purposes. And provide better bang for buck in 90% of the cases.

Re:Writing has been on the wall

[douglasd]

"95% of software is written for Microsoft "

That number you've pulled out of your ass may or may not be close to the truth.

However, it does not contradict the parents contention that most software is written for a compiler, not a chip instruction set.

If you want to support the grand parent's absurd contention that 99% of software is written for X86 specifically, then please provide more data, because my experience is that most software is written for a compiler and tuned to many things, least of all the processor.

If you want to contend that most software written with the expectation to run on windows is tuned specifically to X86, presumably via hand coded assembly, please provide same, because I contend the vast majority of software writen does not include assembly written just for that software to run faster on a particular processor.

If you want to contend that 4% of software is coded to a processor, well, that is less outrageuous, but still doesn't pass the smell test. Please provide data for your unfounded assertion.

Yes -- I think you and the grandparent are either living in an assembly centric world and ignoring the majority of software development, or are stuck in an X86 specific world even more deeply distanced from most development.

Re:Fair play

[$RANDOMLUSER]

> That's why we had the "Pentium(tm)" and not "586" being marketted.

No, Intel tried (and failed) to trademark a number and so had to come up with names.

Re:Hmmm.

[Valharick]

Brainwashed hardly. I just prefer to work for ethical companies. But again, that is my preferance.

Once again, crap wins

[DesScorp]

Let's see...you've got a superior technology that suffers from bad company management ... Itanium.

You've got less expensive yet outstanding technology that suffers from poor market share (for the time being)... Opteron

And then you have a bloated, legacy, piece of shit technology that's a crude copy of Opteron, a Pentium 4 with hastily tacked on 64 bit instructions (copied from AMD), a technology that Intel doesn't even believe in, that they themselves think is inferior.... Xeon

Guess which one will dominate the market?

Sometimes IT really does suck.

Re:Once again, crap wins

[LWATCDR]

Yep once again.
It can join
The DEC Alpha
Commodore Amiga
Atari ST
Zilog Z8000
680x0 ISA
Tandy 2000
Zenith 100
And so on...
In evolution cockroaches have an advantage.

Re:Once again, crap wins

[leathered]

Let's see..

McDonalds, the world's largest restaurant chain that wins countless awards every year for culinary excellence and whose head chefs have sold millions of copies of their recipe books.

AOL, the geeks' favourite ISP reknowned for their quality tech support who can effortlessly guide you through tunneling VNC over SSH at 3 o'clock in the morning.

General Motors and Ford, the greatest car companies in the world who consistantly teach their Japanese counterparts lessons in safety and reliability.

Dell, the world's largest computer manufacturer...

I'll stop now, I'm soooooo depressed.

Re:Hmmm.

[FidelCatsro]

"Dell" and "ethical" are only used in the same sentence if the words "are not " happen to be fitting snugly between them

Re:Hmmm.

[Lumpy]

if you are proud in ANY way of your customer support then you have been sucessfully brainwashed by HR and marketing FUD departments.

I used to reccomend Dell's to people, I no longer do because of the nightmare the Tech support is, INCLUDING the platinum support level for the high end servers. I have a 8450 server loaded with a 7 foot tall rack of powervaults connected to it and It was like pulling teth to get the thing fixed. the techs blamed the "cables" of the powervaults several times and took 2 weeks to get us up and running again after a fatal crash because the powervaults were starting rebuilds of spare drives and then offlining them breaking the raid 50.

dell tech support sucks. that is why we are moving back to HP.

I type this from the best laptop that I have ever had, a Lattitude D800, the hardware is sound, but they drop the ball everywhere else.

not anticompetitive!

[conJunk]

if as you say "Intel offers a better deal" -- and that deal was based upon exclusivity. (In other words: "You get a 15% discount if you sell only Intel chips"), It seems to me that that would be illegal and anti-competitive.

What on earth do you mean? That's about as standard as it gets. It's called exclusive licensing, and that's the way it goes. Companies offer price incentives to sign exclusive deals. It's competitive because Dell is free to sign exclusively with anybody.

Here some other examples: Your job. Your company offers you $100,000/year to build widgets *exclusively* for them. If they wanted a clause in your contract that said that you may not build widgets for anyone else, you aren't going to say it's anticompetitive.

How about your car? Toyotas ship with (I'm making this up) Panasonic audio components. If you asked Toyota to make a line with Zenith components, they'll probably say "sorry, but we have an exclusive agreement with panasonic."

I'm not saying it's a good thing, but it assuredly is not illegal.

Re:not anticompetitive!

[atrus]

How about your car? Toyotas ship with (I'm making this up) Panasonic audio components.

Interestingly enough, except for the high end stuff, Toyota does ship Panasonic car radios :)

Re:not anticompetitive!

[popo]

"What on earth do you mean? That's about as standard as it gets. It's called exclusive licensing, and that's the way it goes. Companies offer price incentives to sign exclusive deals. It's competitive because Dell is free to sign exclusively with anybody." No, IANAL but I don't think that's true at all. Were that the case there would be no such thing as "anti-competitive behaviour". Microsoft bundling would be seen as "exclusivity". 'Exclusivity' means that you have the sole right to distribute a product within a certain region or territory. But as far as I know it does not mean that that right comes with the other requirement that you don't stock competing products. As far as I know, that would be the very definition of "anti-competitive". Any lawyers in the house?

Re:not anticompetitive!

[ectospasm]

It is anti-competitive. It's just not necessarily illegal. At least in the US, companies aren't allowed to require exclusive licenses if it can be proven that they have a monopoly. Doing so would be an abuse of their monopoly status. IANAL, and all that.

It is arguable whether Intel has monopoly status or not.

Re:not anticompetitive!

[conJunk]

IANAL either, and I think ectospasm's reply after yours sums it up well: Those kinds of exclusive agreements are only illigal if there is a monopoly situation involved.

Microsoft couldn't get away with it because hardware distributors didn't *really* have a choice with the OS they shipped, so if Microsoft says "take IE or jump in the lake", their options are to ship IE or find a new line of work, because they won't be selling hardware without windows on it.

Fortunately for us, there are other chip manufacturers than Intel.

Re:not anticompetitive!

[AaronCampbell]

it's not whether the deal is based on exclusivity, but rather if the deal went something like this:
"We will sell you our chips at a 15% discount. What? You want to offer AMD chips also? In that case we will NOT sell you our chips (or the chips have a 100% markup, or whatever they do). See how well you business does when you can't offer Intel! Pssh...AMD...we'll show them"

Re:not anticompetitive!

[iluvcapra]

Before this thread goes in farther, I call no "It would be like Coke doing %s with Pepsi" analogies!

Re:not anticompetitive!

[popo]

> Fortunately for us, there are other chip manufacturers than Intel.

As there are other operating systems besides Windows. I don't think the definition of monopoly requires singularity.

Re:not anticompetitive!

[nvrrobx]

Actually, you're not making it up. My 2004 Solara has a Panasonic-built CD player :)

What's in a name?

[moviepig.com]

One has to wonder, outside the obvious explanation of Intel's anti-competitive trade practices, what is Dell's aversion to AMD 64-bit / dual-core processors?

No opinion on the technical merits from this minimally-technical consumer, but, FWIW, I can draw the "Intel inside" logo from memory...

Re:Fair play

[diegocgteleline.es]

for using their x86-64 technology?

Well, then Intel may sue AMD aswell because of the x86 32 ISA, right?

In fact x86-64 is pretty much the same instrucion set except that it has been extended to support 64-bit registers, etc. So you could very well say that x86-64 ISA is a derivative of the x86-32 bit ISA.

Of course intel and AMD have cross-license or some shit so they can use whatever stuff they want without licensing issues, but i think it was worth the post

Re:Hmmm.

[Stanistani]

>"Dell" and "ethical" are only used in the same sentence if the words "are not " happen to be fitting snugly between them

Let's try that out...
Dell are not ethical.
Hmmm... I'm having a little problem here.
Let me add some words...
People who randomly slam Dell are not ethical.

Works better, but is meaner...

Itanium isn't ALL that bad...

[ajiva]

Itanium as an architecture isn't all that bad, and has some great ideas. The only problem is that with Itanium most of the work has to be done by the Compiler writers to get as much performance out of the machine as possible. NOPs are a killer on Itanium because they take up precious space on bundles. X86 and other architectures are not as dependent on compilers for performance (well ok that's not totally true). Either way normal archs have had 30+ years of research into how to optimize code while Itanium realistically has had about 5 or so.

Re:Hmmm.

[bradleycarpenter]

When you really get down to it most companies are not ethical. I don't find Dell to be any more ethical than any other computer company. They are out to make money, and that is about it. Sure they will tell you all these great things they do to help the world, but in reality it is all simply part of the companies marketing. Look at any company in the world and you will find some sort of unethical behavior. Simply part of the world we live in.

Re:Hmmm.

[FidelCatsro]

Randomly slamming Dell is unethical . Though as a fair few people here can attest to , there are many many valid reasons to slam them .

Re:Fair play

[AusG4]

"In fact x86-64 is pretty much the same instrucion set except that it has been extended to support 64-bit registers, etc. So you could very well say that x86-64 ISA is a derivative of the x86-32 bit ISA."

AND that it has twice the registers.

I actually use an Itanium...

[rbinns]

I have 2 machines on my desk for computational stuff. The Itanium2 box is used for my "set up and run overnight" jobs. It seems to run just as fast as the other box, a Dell Xeon box, but can run more jobs at the same time. Both systems have similar spec otherwise (4 gb ram, SCSI RAID, RHEL). The other major issue I have with the Itanium is software support. My processor program's vendor (CFD) has an optimized version for the Itanium, whereas no similar version of the pre-processor exists. So I mesh on the Xeon, run on the Itanium. I wonder if this chip is still a viable solution for heavy computation or if another architecture is superior?

Re:Hmmm.

[XPACT]

Same here. Gold and platinum support are not the things that they suposed to be. Dell hardware is not good too. I am waiting second day(we have 4-hour response plan) for mainboard replacement(back ordered), still nothing. The capacitors of my main bord Optiplex GX270 have popped out. It is fairly new computer. We had problems in the past too. The PowerEdge server line totally sux. The servers mainboards have some chipset named "ServerWorks" never heard about those, it is not Intel, nVIdia, VIA or SYS. I guess the lowest bidder wins, and DELL are putting a lot of crap in their products. I always buy my computers in parts and put them together, but the businesses doesn't like that aproach.

Re:Hmmm.

[Stanistani]

Indeed.

I worked for an organization that basically built an entire wall of Dell servers.

Their shenanigans made us so unhappy, five years later that wall was replaced by a mix of Sun and IBM blades.

Re:Hmmm.

[FidelCatsro]

I had roughly the same problem with my current job .. apparently my predecessor was a budgerigar as the only things that had come out of his lips were "cheap cheap" . Needless to say a lot of stress ensued with Dell , resulting in a nice deal with Sun .

Re:Hmmm.

[Stanistani]

Budgies... my mum was from England, she used to say that. Nice memory.

Here in the States, they're parakeets.

But still make nice metaphors for bird-brained bosses.

I haven't bought a Dell for ten years now, and in that time handled at least three million dollars American in acquisitions.

Even the random white-box servers performed better.

ServerWorks

[Colol]

The servers mainboards have some chipset named "ServerWorks" never heard about those, it is not Intel, nVIdia, VIA or SYS. I guess the lowest bidder wins, and DELL are putting a lot of crap in their products.

ServerWorks chipsets are fairly common, and not bottom-of-the-barrel crap from some random Taiwanese company. ServerWorks is presently a division of Broadcom focused on high-end server applications. They're most common in large processor configurations (quads, 8s, etc.), with Intel having taken over of a lot of the low end (single- and double-processor) with their own stuff. For reference, Sun uses them in their Opterons, HP uses them in their Xeons... they're quality product, quite popular, and not terribly inexpensive.

I have my beefs with Dell, but they're not shafting you on chipset quality by using a ServerWorks set.

Re:Hmmm.

[markass530]

Remind me again, who is in sole possession of 1st place in their division tied for the best record in football? Oh Yea, the Niners. Who is in dead last place in their division, with the worst record in the entire conference? The Raiders. Case closed, end of story.

Re:Hmmm.

[FidelCatsro]

Coming to think of it , When i was a little kid back in Scotland .
My Gran had a couple of budgies .. one used to sit on my shoulder and squawk .. which is a very good metaphor for a dell Sales rep .
Which is precisely why I moved things over to Sun (that and previous experience , which was mostly with Spark servers )

On the main issue of the article :Dell dropping Titanium is not shocking really , I think Intel have probably decided to relegate it and Dell are just following the beckon call

Re:Still good for Intel's business

[williamyf]

As a matter of fact, Itanium is good for every company involved:

Let S be sales:

S(Itanium)S(PA-RISC)
S(Itanium)>S(ALPHA)
S(Itanium)>S(MIPS)

So Intel sends most of the competition in the high end packing. But the guys making PA-RISCs MIPSs and ALPHAs transfer the R&D costs (Sun has not produced a viable spac in years, The latest chips come from Fujitsu, and the Niagara design comes from a startup they bought, check IEEE Spectrum) and Fab costs (how much does a state of the art 90nm fab costs nowadays? 1.5Billion perhaps?) to Intel... From a financial point of wiew, Win-Win situation.

The only thing is that since now those companies do not have ISA lock-in on the customers (and if the customer is runing Linux on top of Itanium the also do not have OS lock-in) some less efficient companies (less eficient technically, or in marketing, or in sales, or in a combination) will fall to the sidelines... Oh, the humanity! Please, someone tell me how taht is a bad thing!

Of course, some people say Itanium is a pile of dong... I do not know (and do not think so, and if it is, intel can solve it with a huge bag of bills ["a realazo limpio"]). Others say that X86-64 is the cure-all-silver-bullet-miracle-medicine. I do not know (and do not think so, all that cruft, you know?).

The point is that for it's intended use (high end servers), and for the time being, Itanium is OK.

Re:to the management

[Anita Coney]

Looking through your short history here it's apparent that your commentary leans more towards the "funny" and flaming versus the "insightful" and "interesting."

Hopefully you'll get bored and leave soon enough. Or maybe we'll get lucky and science will develop a cure.

Re:Hmmm.

[Floody]

I used to reccomend Dell's to people, I no longer do because of the nightmare the Tech support is, INCLUDING the platinum support level for the high end servers. I have a 8450 server loaded with a 7 foot tall rack of powervaults connected to it and It was like pulling teth to get the thing fixed. the techs blamed the "cables" of the powervaults several times and took 2 weeks to get us up and running again after a fatal crash because the powervaults were starting rebuilds of spare drives and then offlining them breaking the raid 50.

Heh.. Try it with over a thousand servers. Dell seems to have little in the way of methods by which large corp customers can bypass tier 1 & 2 support. Actually, Sun (w/ x86 hw, not sparc) is worse in many ways, but that's largely due to a massively bad crop of seagate discs in the sunfire v20zs.

Re:Hmmm.

[graphicartist82]

I know you have to pay for it, (i believe it's $179/yr) but dell has it's warranty parts direct program

Basically, if you know what's wrong, and the product is under warranty, you can go to a web page and order the part you need. No need to wait on hold for 30mins just to re-troubleshoot the problem with somebody reading from a script when you know that your hard drive is dead.

Deja vu

[Guppy06]

"now promoting the chip as a high-performance replacement for reduced instruction set computing (RISC) processors in Unix servers from companies such as Sun Microsystems and IBM."

Weren't they sayin the same damned thing when the first 32-bit Pentium chips came out over a decade ago? They've been catching up to RISC for how long now?

Re:Hmmm.

[Anonymous Coward]

if you are proud in ANY way of your customer support then you have been sucessfully brainwashed by HR and marketing FUD departments.

I used to reccomend Dell's to people, I no longer do because of the nightmare the Tech support is, INCLUDING the platinum support level for the high end servers.

Yeah, we moved from Sun to Dell, and if that wasn't the mother of all fuckups I don't know what is. I'm trying to get us moving back to Sun next year as far as servers go. I don't know what we're doing as far as workstations though as I'm a server guy.. but I do know that a shitload of them are having thermal shutdown issues, and the subcontractor Dell is using in our region for the small stuff seems to be slipping on responsiveness.

Not long now for SGI

[leathered]

They bet the farm on Itanium and will shortly pay the price.

Re:Hmmm.

[rvw14]

Niner Fans wish the season would end today. To bad there is still 16 weeks to play. Go ahead, rub it in for the one week you are in front.

Signed,

Bitter Seattle Fan

x86 is faster than any of those.

[YesIAmAScript]

It's far faster than UltraSPARC (heck, ask SUN), Alpha is slow on current measures. It's measureably faster than PowerPC. And MIPS64 isn't keeping up with the current marketplace much better than UltraSPARC. It might not be faster than IA64, depending on how you measure it.

The reasons all those have lower clock speeds than current x86 (well, P4) is because they are designed to do more per clock, with less clocks. It's pretty simple. Like Pentium M. Either that or because they just don't perform as well (like PPC, SPARC, Alpha).

Note, the 8086 (first x86) was not a filler move between other unrelated product lines. It was a follow on 16-bit extension to the 8080, before the "real" 16-bit processor Intel was working on. Kind of like x86-64 versus Itanium. So it really was a filler move between a related line and a non-related line.

I do agree the ISA sucks. And I hate little endian. But given how little time people (even programmers) spend looking at object code or writing assembly nowadays, the ISA is near immaterial.

But anyway, Intel picked one method of running a lot of code, super-pipelining. It worked for a while, but is running into problems, and they are going to a different approach. AMD shows you can have good x86 ISA performance without 30 stages of pipelining, so don't throw the ISA baby out with the P4 bathwater.

As an aside, I don't get the worship of Alpha on /. Alpha was the first super-pipelined, superscalar processor. In that way, it bears a huge resemblance to P4 (Netburst). And if you saw the chips (and their heatsinks), you'd see a resemblance on performance per watt too! I'm not saying I hate Alpha, but I just don't get how you can hate on superscalar/super-pipelined in one case and worship another.

Re:Hmmm.

[XPACT]

I am not trying to be a troll, but I will always prefer an ASUS mainboard over DELL mainboard. The BIOSes of DELLs are too generic, there are no extra options for tweaks etc.

Re:games

[turgid]

Because on a hot summer's evening, when you're sitting playing halo with a cold beer you don't want a space heater with a leaf blower in your front room.

Re:Hmmm.

[fm6]

But the number one rule we live by is that if we win, it will be done the right way.

So robbing and killing people is acceptable, as long as you don't get caught?