In the Windows 2003 SP1 beta, Intel says it will offer a preview of the 32-bit execution layer that will cause the Itanium 2 processors to run x86 applications at the full clockspeed rating of the processor. For example, the 1.5-GHz Itanium 2 “Madison” processors launched this week would offer that clockspeed to 32-bit applications, as well. Until now, the inability of the Itanium Family processors to run 32-bit applications without a large performance penalty has been a major shortcoming of Intel’s 64-bit Itanium line that has negatively affected industry adoption.
Intel’s 32-bit on Itanium Preview to Come in Windows 2003 SP1 Beta
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
it was only a matter of time before they would follow the opteron (in a sense).
Oh dear, sounds a bit like there could be some competition with Apple’s G5 argument being that it’s 64 bit processor can also do 32 bit processing all the way. I guess we’ll wait and see how this goes…
Isn’t this just software emulation? As such, isn’t it going to perform a bit like bochs? Sure, it’s probably faster than the “native” x86 support in itanium’s hardware, but will it compare favorably with intel’s x86’s and amd64?
apple is almost zero competition to intel, but IBM is. Why is this new ppc chip is almost immediatly apple’s? this is like saying “the new DELL pentium 5 in their fastest pc is gonna kick some ass”
no, the early itaniums had software emulation (i believe there was _some_ hardware), but mostly it was software. this news is saying it will be pure hardware.
the reality distortion crowd had to bring up the g5
no, the early itaniums had software emulation (i believe there was _some_ hardware), but mostly it was software. this news is saying it will be pure hardware. </quote>
actually i thin i’m wrong on the software emulation. I think it was always hardware, but it wasn hardware emulation. not actual x86 silicon. so i guess its anyone’s guess what intel has done. They might have improved the emulation silicon or actually added x86 native silicon. (i’m not a cpu guru, so if anything sounds funny tell me)
This is why nothing gets any better very quickly.
They force us to upgrade. But we force them to maintain backwards compatibility.
How many of us are still running 16-bit applications because the companies that created them never botherd to update them (hello MOTU with your crappy Unisyn!)? How many businesses are still using textmode displays for their HR or database management (hello NCC and Buckeye!)?
By the time 64-bit computing is pure and widespread, they will be announcing the unvieling of the 128-bit processor.
I know I should put my energy elsewhere instead of getting upset about all this but my hatred for the computer industry seems to know no bounds. Besides, I suffer crappy 16-bit software that barely functions on an OS that’s about 8 years later than the one it was designed for (because that’s all they sell, still, today; a 16-bit version that costs about $300). I’m the one that suffers with the crappy CARS system here at this school every time I need to manage student registration or records data.
Are you sure? I thought it was the other way around. As far as I understand the Itanium can run x86 natively, it’s just not very fast. In fact it is so slow that it is faster to use software emulation. The software emulation might be boosted by some hardware perhaps.
no i’m not sure, i guess i should read up on it. Last time i did research on the itanium was years before it even came out
Do you really think that they will release processors capable of handling more than 16 million terabytes of memory before 64-bit computing is widespread? It’s not impossible of course, but I think 64-bit will be enough for a while. For desktop users I think that 8 million TB should be enough for a decade at least…
“Oh dear, sounds a bit like there could be some competition with Apple’s G5 argument being that it’s 64 bit processor can also do 32 bit processing all the way. I guess we’ll wait and see how this goes…”
Apples and oranges, the G5 can do 32bit ops usint the same ISA, the Itanium does 32bit but emulating a completely different ISA.
That’s right. Things don’t move forward very fast because of the requirement to maintain backwards compatibility. Why? Because most all businesses can’t adapt at the rate at which IT is changing.
In any organisation, IT forms a big part of the ‘backbone’. But like everything else, IT is subject to management scrutiny and where purchases are made – or changes are made – they have to be justified with business requirements.
I’ve seen systems of various kinds that don’t work well; unfortunately, they do work, and that fact is enough to scuttle arguments to replace it, involving time and money. “It’s worked for 8 years; why can’t it work for another 8? I mean, before the widespread use of computers people used card files and those could last for decades, surely applications can do the same? They cost enough we should expect them to!” You may have seen such attitudes around…
Wouldn’t it be nice if things were different. Intel could make a break from the 80×86 architecture. But the fact is, without proper 80×86 compatibility, the Itanium would be totally sunk by the Operon. “We’re not upgrading all our server applications because some vendor or other thinks we should” is the irrefutable argument here.
The orginal idea was to do software imulation of the x86 system… guess what, this was too slow.
Now, MSFT purchased Conectrix which is famous in Apple world for their ability to JIT compile x86 into 68k and PPC native. Guess what, the new 32bit layer is a JIT compiled emulator which should be around 2 to 3 times faster then the old emulator.
No contest…
1) The PPC 970 has through put of about 75% of the POWER4
2) The POWER4 is about 2x faster then the I2.
The worst case is that the I2 is the same speed as the PPC 970.
The better compitition will be the AMD Opt vs PPC970.
Intel is killing the x86 so all the current PC people will need to get a new computer come 2008. Microsoft is betting on Virtual PC to keep x86 compatability. AMD is betting on the Hammer chips. And for once, Apple isn’t changing CPUs.
MS distorted people bring up the G5, they have to now in defense.
Um, IIRC, the Itanium2 is *faster* than a Power4. At least according to the latest specfp benchmarks. Also, Eugenia, you should do something like put bad quotes in big red letters.
“For example, the 1.5-GHz Itanium 2 “Madison” processors launched this week would offer that clockspeed to 32-bit applications, as well.”
Does the author have the slightest clue how a processor works? What does he think: the I2 clocks down to 600MHz when running 32-bit code? MHz Myth indeed…
“1) The PPC 970 has through put of about 75% of the POWER4
2) The POWER4 is about 2x faster then the I2.
The worst case is that the I2 is the same speed as the PPC 970.
The better compitition will be the AMD Opt vs PPC970. ”
1) The I2 has through put of about 1769404597% of I1.
2) The I1 is about 2384327x faster than POWER4.
The worst case is that the I2 is 3489348943X the speed of POWER4.
The better competition is I2 versus the Earth Simulator.
Come august and this coming year when the 970 hits 3 gig should be some nice compition.
I have always been a fan of AMD, and now the 970. Just will be nice to see both of the pull a head of Intel.
Hello Don:
<quote>
Isn’t this just software emulation? As such, isn’t it going to perform a bit like bochs? Sure, it’s probably faster than the “native” x86 support in itanium’s hardware, but will it compare favorably with intel’s x86’s and amd64?
</quote>
I believe Intel acquired a good chunk of the Intellectual Property (IP) and team that worked on the DEC Alpha. The Alpha had a neat utility for Windows NT 3.xx called FX32! (or something like that). This utility would convert a Win32/x86 program into a native DEC Alpha program the first time it was run. The converted program would run at 70% to 90% of the speed of the same application on similar speed x86 hardware. I’m not saying this is what the Itanium will be doing but, this and a little hardware (maybe something like op-code interception hooks) would be a good way of avoiding the speed penalties of the “traditional” interpretive software emulation method for running x86 code.
I don’t think Intel wants to (I definitely wouldn’t want to) pollute the EPIC design with emulation hardware (or even emulation micro-code). Especially since they need all the real-estate they can get for cache and EPIC optimizations. They’d be better off making the Itanium faster so that the “emulation mode” doesn’t seem slower than the available x86 CPUs.
Michael.
I may sound cynical, but this is nothing new. Anyone remember the morphing software released by DEC for Windows NT for Alpha so that users could run x86 32bit applications on Alpha?
This is just more technology finally being used by Intel after them aquiring the DEC crown jewels. Hypthreading, yet another DEC innovation, EV8, which apparently will appear in the next itanium. Basically, if you want to see what will happen in the future for Itanium, just have a look at what DEC has created, replace the name with something “hip” and “cool”, voila, you have the Intel revamped version of it.
If so, this has been done with Intel processors for years.
Well atleast now we get to start seeing those amazing technologies that DEC pioneered. Wasnt the xscale cpu started by DEC also? It’s either the xscale or another one of intel’s big embeded CPUs. DEC had an awesome team and did incredible things. Proof: this 1997 DEC box keeps up with all of my friend’s box, which arent too shabby.
Well, all I can say is this. It is ashame that Compaq bought it, raped it and then asset striped it because Compaqs management were to inept at running anything appart from a Microsoft sycophant shop with 0 innovation.
Now HP has bought Compaq, what have we got now? a Microsoft mega-sycophant corporation with 0 innovation. HP/Compaq is no better than Dell.
Aye, the EV7 is an absolutely kickass processor, and what is HP doing? Basically hiding it from the public, so that they can push the itanium instead. They can do what they want i suppose, i’ll just keep using my old alpha until i find something that i deem as good comes along.
Worse still, they’re killing off their PA-RISC processor as well. I know quite a few people who having nothing but praise for its design, heck, even a group of SUN Sparc engineers can’t work out why HP are killing off such a great chip.
As for Alpha, Intel should have killed off Itanium long ago, embraced Alpha and put money into its architecture. Had they done it, they wouldn’t be such a joke in the enterprise market.
I don’t care about the advantages of one ISA over another. PowerPC vs. Alpha vs. Itanium vs. x86. I think that Itanium will succeed solely because Intel is backing it. Even if you have a horrible ISA, if you put 3 mb of cache on chip, it is going to be a really really fast processor. Do I believe that Intel has the best design out there? No. But I do believe that Intel is a production monster, and economics of scale and mass market will eventually make Itanium a success. (I don’t believe it will be AS successful as Intel claims, but Itanium isn’t going anywhere)
Hello,
I could go into pages upon pages of how em86 works, Key facts.
1. Itanium is using a version of the FX32! code. I talk
to MS people about a year ago.
2. MS owns the code to the internal jit engine.
3. Em86 kicks booch ass, It’s also a lot more complicated.
4. With em86 code you could run any OSX software on an intel
box. (native speeds)
5. Think JIT not emulator
6. The library calls are all rethrown to native library
calls making it fast.
Factoid. All i mean ALL library calls eventually end up at libc and libm.
Donaldson, last em86 maintainer.