HP Terminates Itanium Workstations 472
vincecate writes "The largest Itanium system maker,
HP, has terminated its Itanium workstations.
It seems their workstation customers have spoken in favor of x64.
In related news, Intel expects to ship
over 100,000 Itaniums in all of 2004
while AMD is estimating
1.5 to 2 million AMD64 chips in Q4."
AMD64 include consumer processors (Score:5, Informative)
Of course, the ASP of Itanium is a lot higher, so Intel need to sell a lot fewer Itaniums to get the same money back as AMD. On the other hand, AMD haven't sunk $billions into K8!
What about servers? (Score:4, Informative)
Re:Low power CPUs? (Score:2, Informative)
AMD sell a 35W Opteron, 1.8GHz I believe, I'm not sure. Maybe it is 55W @ 2GHz.
OTOH AMD's consumer processors include Cool'n'Quiet which downclocks the processor when you don't need lots of processing power, and hence cuts the power consumption a lot. With a decent fan the fan will also slow down.
Or get an iMac G5.
Re:Low power CPUs? (Score:2, Informative)
I only say 1.8 because IIRC, that's as slow as you can get on a Socket 478 mobo, and you probably don't want a 423 based board, because it's likely to only support SDRAM or RDRAM.
Get one of those big Zalman passive heatsinks if you don't want a fan. Just be careful moving the machine or find a way to brace it properly. I bolted mine through the mobo and directly to the steel backplate of the case, all stress on the case and not the mobo. I just dont trust a plastic clip to secure a half kilo of copper just inches above my Radeon 9800.
The newer prescott based Celeron D's perform like a champ (as in, really close to their P4 counterparts) from the reviews I've seen. I think they start at 2.4 gigs.
Stay away from bleeding edge stuff. It's just a waste of money, and won't improve your computing "experience", unless you consider bragging about artificial benchmark scores "computing".
Re:Low power CPUs? (Score:3, Informative)
90nm A64s seem to draw much less power than 130nm A64s.
There is also Transmeta [www.transmeta] which produces the Efficeon CPU and VIA [via.com.tw] which makes EPIA.
You may also get an AMD Geode [amd.com]
Where are OpenVMS user going to go? (Score:2, Informative)
http://www.theinquirer.net/?article=132
Now that HP will stop making Itanium servers, is HP going to have a migration plan for OpenVMS users to go onto Operton?
Hopefully this won't kill off OpenVMS, the Operating System that won't die.
Re:Yeah, Itanium tanked... So what? (Score:3, Informative)
IA64 is proprietary and closed, AMD64 is not. That's why Intel are the bad guys as far as this Itanic thing is concerned. Also, if they are selling something you will never be able to buy to your home, it's natural to root for the solution that you might very well be able to afford yourself.
Re:Yeah, Itanium tanked... So what? (Score:3, Informative)
They shifted too many things off of the CPU and into software when that didn't preform well they started trying to optimise it. It's a situation that reminds me of NT and microkernels.
The result is something that needs a huge die size just to preform on par with the Xeon and thanks to the huge die size it will always be priced much higher. I keep hearing that smaller transistors will fix this problem but everything they benchamark against will get those too.
Intel really missed the boat on this one.. a more intellegent route would have been emt-64 without the 16 and 32 bit backward compatability.
you're not the only one mocking the Itanium... (Score:5, Informative)
Re:Itanium will crush all... hardly (Score:1, Informative)
Is HP a monopoly in CPU business. Remember IA64 is the successor to PA-RISC. HP did most of the design on IA64 and to some extent did the micro-architecture for Itanium 2.
Re:Low power CPUs? (Score:2, Informative)
The P4 core loves bandwidth. The low cache on the Celeron kills performance. Getting a slower Celeron is okay I suppose, it won't be affected as much. Faster Celerons showed extremely bad scaling, in lots of REAL WORLD benchmarks, not just the artificial ones.
I suppose that you can't see the difference, you have nothing that needs it. To be fair, doubling the memory in a system is a better investment than another 20% clockspeed though.
I suppose my point was that for $x, you could get okay performance with Celeron, or good performance with a different processor.
Re:There goes VMS on the desktop again... (Score:3, Informative)
Re:Yeah, Itanium tanked... So what? (Score:1, Informative)
Proprietary and closed how? You can download all the necessary information to program for IA-64, as you can for any CPU. No CPU maker would ever not release this information; they wouldn't sell a single processor.
Re:What about servers? (Score:5, Informative)
In the longer run, IMHO it sounds somewhat problematic, considering that all the engineers developing software will be running on x86-64. I.e. the software will first be available on x86-64, more tested etc.
So why should the customer shell out money for an Itanium server instead of an x86-64 server which has better bang-per-buck and runs the software more reliably? In the short run HP can probably contain x86-64 in low end servers, keeping high end stuff reserved for Itanic. But in the long run, they'll have to start providing higher end x86-64 gear too, or their customers will move to a competitor that will.
Re:Should have stuck with Alpha (Score:5, Informative)
Exactly what I was thinking.
HP and Intel deserve this for killing off the two most powerful processor lines in history.
Back when PA-RISC and Alpha were in production, the gap between them and the next fastest CPU lines were staggering. I used to check the CPU Info Center at Berkeley every time a new one was released, just to see how badly it humiliated the competition (sadly, the CPU Info Center is no longer maintained).
The Athlon (before it was named such) uses the Alpha's bus... and the original slot-A design was compatible with both the Alpha and the Athlon, all you would need to sell a motherboard for the other one is a different BIOS. This was the selling point that convinced many motherboard manufacturers to actually make these boards. Unfortunately, only a tiny handful of companies actually marketed the resulting systems using the Alpha CPUs (mostly in Linux Journal & Linux Magazine as rackmount servers).
They could have done so much more... oh well.
My current favorites are UltraSPARC and PowerPC (with POWER close behind).
Re:Yeah, Itanium tanked... So what? (Score:3, Informative)
> Itanium for Windows and Linux are pretty godlike
> in my experience.
I'm sure they're good, but they're not good enough.
> Look at AMD benchmarks and usually they are done
> with the Intel compiler.
That'd be the x86 compiler, not the Itanium compiler.
> Your definition of sucked differs from mine.
Stringing vast arrays of processors together to build supercomputers tells you almost nothing about the performance of the individual processors.
You'd have done better to quote SpecFP numbers, but see my comment above about relative transistor count. IA-64 just doesn't give much bang for the buck (or transistor). If you strap a jet engine to a pig, sure it'll fly.
> Coke does the same thing over RC cola. Windows
> does this over OS X.
Yeah, and we don't have to like it.
IDC have to revise their 96% mistake (Score:3, Informative)
Being that wrong takes talent. Pulling something out of your ass qualifies as precision work compared to this.
Comment removed (Score:3, Informative)
Re:Could it be? (Score:5, Informative)
Bzzzt... wrong answer.
In AMD64 chips, the chipset doesn't have the memory controller - it's in the CPU.
AMD's CPU-to-DDR latency is much lower than Intel's.
Re:A victory for 32 bit backwards compatibility (Score:3, Informative)
64bit software runs faster on the amd64 than 32bit software - any slowdown due to larger instruction size (negligable since the 64bit data path can read more data in parallel anyway) is more than offset by the vastly increased number of registers.
Re:Could it be? (Score:5, Informative)
A64 FX's and Opterons support dual channel ddr and have much lower latencies then intel at the same mhz (400mhz ddr X 2). Usually the FX's and Opteron's win the memory bandwidth benchmarks.
As low power AMD has a line of mobile barton core processors that use as little power as 45 and even 35 watts. They can also be placed in a destop motherboard. The xp2400 35 watt is also under $100. But there is a good chance the pentium-m uses less power but they are only found on certain laptops (can't be bought).
Re:Should have stuck with Alpha (Score:2, Informative)
It's my job to evaluate, amongst other things, processor performance for use in a grid-enabled batch transaction applications. The only CPU slower than UltraSPARC III Cu @ 1.2 GHz is the HP PA-RISC 8700 and 8700+, and that's so old the industry was playing ring around the dinosaur when it was introduced.
The fastest CPU for this is kind of work is, believe it or not, the Xeon MP chips at 3.0 GHz or greater with 1 MB L3 cache or greater. It's even faster than the dual core POWER4+ @ 1.7 GHz. Too bad you have to live with a 32-bit memory model.
We've been evaluating AMD Opteron against our application - *very* promising.
Re:How Ironic (Score:3, Informative)
Heh. I've heard that before:
"The Pentium Pro is a server CPU; it is not suitable for desktop use.." - Intel, 1996
"The Pentium II, based on the Pentium Pro core..." - Intel, 1999
"The 486 is intended as a CPU for high-end computing needs." - Intel, 1991
Re:A victory for 32 bit backwards compatibility (Score:2, Informative)
Re:How Ironic (Score:3, Informative)
From my perspective, IA64 is already dead.
Re:Yeah, Itanium tanked... So what? (Score:5, Informative)
(* Itanium2 doesn't even do next-line prefetching!)
Explicit speculative loads was a major mistake because in many kinds of code the compiler cannot place speculative loads far enough ahead of the actual use for it to pay off. Often the address to be loaded from is simply not known far in advance of the load (consider executing the C code "x = a->b->c"). So Itaniums spend a lot of time stalled waiting on memory accesses. That's why Intel spends so many transistors on gigantic on-chip caches, to try to reduce that pain. The architecture's pretty good for workloads with very regular and compiler-analyzable access patterns (regular number crunching, SpecFP) but it's bad for everything else (servers, user applications, irregular numeric codes).
Yes, IA-64 is a aggressive, radical, clean and somewhat novel design, so it's understandable that some geeks love it. However, it is not a good design.
If it was a good design, then with Intel engineering, 5x the transistor count, and no backward compatibility requirements, it would be absolutely crushing Opteron performance. Instead it is merely competitive.
BTW it is quite odd to consider IA-64 a small tweak over RISC chips. IA-64 is the most dissimilar of all viable architectures today.
Re:That's actually quite sad (Score:3, Informative)
Itanium 6 integer units shortpipeline, 3 branchunits, multiple FMAC, smaller CORE size than netburst on equal process. Yes, on same process itanium2 core is smaller than netburst. Biggest block after bus interface and L3 cache on itanium2 are block called IA32 and FP-units. Itanium brings huge execution resources with low die area costs, and then puts huge caches that have redundancy so that MFG costs wouldn't be much worse than celerons mfg costs, while giving HPC performance.
I had some doubts on itanium until I read the die photo and compared to my knowledge on whats taking die area on x86. When itanium will ditch IA32 hardware compability since these days dynamic translation software is faster than what hardware does and it still gets better. With carefull powermanagement (By no doubt FP takes substantial portion of Itanium2:s powerbudget and ditching IA32 should also free reasources for other use.
X86 is not just a instruction set its whole set of laws how instructions interact exceptions memory models etc... many tricks for x86 to run fasts are not needed, so IA-64 can be smaller. Now don't put that it some how needs bigger cache. No the software that is run needs certain sized DATA caches no matter the architecture, for a given performance level. Now the itanium looks like a great TARGET for dynamic translation software as SGI has MIPS simulations running much faster than native mips, and Itanium2 x86 compability SOFTWARE layer runs fast too... And in next year itanium will get 512KB L2 Icache so what ever you think for code density problems intel fixes by trowing more cache as a solution...
Now HP dumping workstations its nothing big. Nearly all Itanium sales where in server side, because the price for workstation level chipsets, especially because of HP pricing policy, of 2X the risc worstation price and 5X the x86 price. 2007 is the year when Itanium either sinks or swims, its year where it will get price parity with xeon, by sharing motherboards and drivers with it...
Re:So, question for the crowd... (Score:3, Informative)
Attempts to get away from x86:
iAPX432
i860
i960 (still viable in the embedded market)
IA-64.
Re:How Ironic (Score:4, Informative)
HP was originally in the scientific instrument business. That makes Agilent the true successor, not the current computer company HP. I'ld say what remains at HP are mostly the ruins of DEC, and Compaq. The best of those companies seems to have left for other opportunities.
Re:How Ironic (Score:4, Informative)
So far as I can tell, the HPC shops are largely shunning the Itanium.
I have access to about 10 supercomputers at various locations: not one of them is based on the Itanium. We have clusters based on Xeons, clusters based on Opterons, machines based on Alpha 21264, IBM computers based on Power4 processors, and Cray X1s, based on their own proprietary chips.
But *not one* machine based on Itaniums.
On the Top500 Supercomputing sites list, only 13/500 are using Itaniums. 14/500 are using AMD processors.....
The Itanium may be an HPC processor, but it's one that the HPC community mostly doesn't want.
--PM
Re:How Ironic (Score:3, Informative)
Intel is not in the business of making special purpose CPUs. That's something Cray and Sequent (IIRC) wer famous for, and look where it got them?
Intel makes general purpose CPUs and is very good at it. i960 (you may argue this
It wasn't obvious? (Score:1, Informative)
As someone who spent an number of years (1992-2000) working for Digital ("DEC") in the group making binary translators/optimizers/StrongARM tools/... I actually know a bit about the facts.
Don't blame HP for the cancellation of Alpha. It was smashed by Crap-paq via Mr. Capellas' golden parachute (ok, so the parachute part is speculation). But "Mikey" was the one that cancelled it (destroying API Networks, a joint Samsung/Digital venture doing $300M/year in sales as a side effect).
The Itanic was a compiler-writers dream - permanent (life + 99 years) employment because it exposed so much of the machine to the compiler, requiring it to perform heroic (if not impossible) feats in order to get code to perform at full theoretical capability. An interesting RESEARCH architecture.
However, with Intel's marketing muscle, it was exceptionally successful - it killed Alpha and MIPS utterly, in spite of it probably being a multi-thousand dollar loss on a per-itanic-cpu-sold basis (based upon the amount of truly exceptional engineering talent being wasted on the Itanic - a large number of people from DEC's legendary AMT and VSSAD groups)
But in "defense" of the Itanic, it is a wonderful engine for FORTRAN code which has very large and long straight-line code sequences to perform lots of floating point ops, when thermal conditions (and subsequent cooling costs) are irrelevant.
That is, if you can afford to ignore the refrigeration and power consumption costs, and want to model large, memory-predictable, floating point intense systems in FORTRAN, Itanic might have been a good choice. Otherwise, it wasn't, and that fact was blatently obvious to a lot of Alpha developers from day zero.
rcg
Not all K8s are AMD64 (Score:1, Informative)
That estimate of 1.5 to 2 million is for all K8 cores. However, the new Semprons are K8 cores but have 64-bit disabled. So the real number of AMD64 chips is lower.