Intel’s next Itanium2 processor likely will run at 1.5GHz, a 50% increase from its predecessor and an indication the company is getting better at meeting development goals for its high-end chip family. Read the report at ZDNews.
Next Intel Itanium to Run at 1.5GHz
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
so Opteron will still be better, then?
Intel could have made something great, but from what I hear their R&D went more towards featuritis rather than making the best possible design. Itaniums are heavily reliant on the compiler to be efficient, and then Intel went ahead and made it next to impossible to make a decent compiler for it. AMD’s 64bit processors might come too late to make much of a difference, as P4s with HyperThreading will be commonplace by that time and they will (in all likelihood) kick a**.
What about the IBM Power4 “lite” (ppc 970, i think) chips supposedly shipping in Macintosh computers (late) next year. They are supposedly supposed to be running @ 1.8 GHz. Of course they are projected to come out at the end of next year.
oh bother …
Just wondering how the two relate to each other in terms of projected prformance and if either will gain acceptiance in widespread use.
all this 64-bit hootenanny makes my head hurt.
The Itanium 2’s right now are more comparable to the Power4 chip than the 970. Intel isn’t selling a low cache model on a small die. OTOH the only thing stopping this is Intel’s marketing theory. If you subtract off the cost of the cache the Itanium 2 actually costs less than a P4!
My guess is that using a good compiler at similar clock speeds the Itanium 2 will outperform the 970. OTOH IBM has shown a prototype of a 970 running at 6Ghz (buggy but definitely proof of concept) so its possible that Apple might actually have the much faster chips sometime in 2004. Also the Power5 is about a year away and this doubles performance per clock cycle for the Power4 line (and thus for the 970 line).
All told I’d say its close but IBM has the upper hand.
>P4s with HyperThreading will be commonplace by that time >and they will (in all likelihood) kick a**
Hyperthreading can improve the performance of some applications and decrease others. Intel talks about 30% best case, 10% in general.
Hardly what I would call kick a**.
Cheers
David
Hyperthreading can improve the performance of some applications and decrease others. Intel talks about 30% best case, 10% in general.
As more and more applications become multithreaded to get the benefit of HT, the average performance increase will be between 20% and 30%. You can’t forget, at 3.06ghz , 30% improvement is equivalent to 3.97 ghz, nothing to sneeze at.
HT is free (speaking in terms of hardware, and what’s physically in your computer, not price unfortunately) performance. How can you not see this as a good thing?
My guess is that using a good compiler at similar clock speeds the Itanium 2 will outperform the 970. OTOH IBM has shown a prototype of a 970 running at 6Ghz (buggy but definitely proof of concept) so its possible that Apple might actually have the much faster chips sometime in 2004. Also the Power5 is about a year away and this doubles performance per clock cycle for the Power4 line (and thus for the 970 line).
All told I’d say its close but IBM has the upper hand.
I’d have to agree with that. The Itanium 2 may well out-perform the 970, but look at the price:
Intel declined to comment on pricing, but Insight64’s Brookwood expects the cost to stay about the same as the $4,226 price of the current Itanium 2 and the $4,227 first-generation Itanium. “What they’ve done with the Xeon and Itanium line is that each new generation of chips is using exactly the same price,” he said.
I have no idea what the pricing on the 970 will be, but if Apple is going to be using it for volume desktops, you can be sure the price isn’t going to be anything close to that! If the 970 is even remotely competitive in performance, you can be pretty comfortable the 970 will eat the Itanium 2’s lunch in the marketplace.
Of course, the 970 is designed specificly for desktops, and the Itanium 2 is designed primarily for servers. It remains to be seen whether the 970 will prove to be a suitable foundation for commodity servers.
Yes the Itanium is achieving good things, but at a cost. That’s a very sizeable die size, which means costs are up all over the board. Given that quote for price (and it doesn’t look like it can come down any) the Itanium is simply _not_ even close to viable for desktop consumption.
So what is? Well the P4 is here now… and it’s only getting better. The 970 does look promising, but we’ll see when it arrives. If Apple gets their hands on this technology putting them back on top for performance, I think we can see Apple machines (and the chip) eating some serious market share — especially in Pro.
Ah actually in the above I was addressing performance. If you look at Intel’s price list(http://www.intel.com/intel/finance/pricelist/) and subtract off the cost of the cache the Itanium 2 is less than the P4. Intel choose to sell it only in a pricey configuration.
Where I think IBM has the upper hand is that the Itanium 2:
1) Isn’t backwards compatable
2) Doesn’t support instruction reordering (which was the major accomplishment of the P2 line and when taken out of the early P4s killed performance).
3) Intel is having trouble boosting the speed
4) IBM is on the verge of another generation
>HT is free (speaking in terms of hardware, and what’s >physically in your computer, not price unfortunately) >performance. How can you not see this as a good thing?
It is a good thing, I was commenting that it is not like it doubles performance or anything.
If I really wanted to rant I would be asking why Intel doesn’t fix their bus architecture so that multi-processor boards can perform. A dual processor machine with a proper bus CAN double performance but Intel isn’t interested in making one except on their Xeon chips.
So instead we get HT which may or may not give you a not quite insignificant increase in some applications.
HT is interesting but it doesn’t get that exciting for me at least.
Cheers
David
it’s another comany who makes them, bu I forgot the name.
It’s free?
The only reason it works is that the instruction set is so hard to optimize that they have to insert a lot of NOPs into the instruction stream to handle instruction dependances. With HT they insert an instruction from another thread instead of a NOP.
If you can get 30% increase, then your program is not compiled for the P4. Once all the programs are recompiled to run on a P2 or higher instead of a 386, expect to see an increase of less then 10%.
You can’t do out of order on the Itaniums. The design is for the compiler to determine which instructions can run together and to reorder the code to for speed. Itaniums instructions are created in a block format of arournd 3 instructions in size. The entire block is submitted for execution at the same time.
For the linux people, the GCC will need a major rewrite to handle the Itanium well. One of these days, I want to find the time to a rewrite of it. I’m thinking that the front end and back ends should be hot-swapable thus making it a cross compiler. This would also remove the target instruction set from the intermedate form allowing more high level optimiztions to take place with the intermedate. The back-end would become a small compiler for compiling the intermedate language into the target processor.
I would say otherwise. Power5 definately has the upperhand over Itanium, but on the other hand, 970 doesn’t have much of a upper hand against Pentium 4 (and Xeons). The only thing for them is 64-bit, which for majority of Apple’s customers, it would only give them bragging rights (and no more).
But just say in 2004, IBM releases that 6GHz 970 (minus the bugs, of course). How long would IBM have the upper hand? If IBM and Apple (as well as AmigaOne makers…) are the only one using it for workstations, the future for it is bleak. Why spend more money on something not all that profitable?
Costing two times more than a Athlon XP 2600+? Free? I think it is much cheaper having a dual processor Athlon XP machine than a HT Pentium 4. Personally, I think HT is way overblown. It is nice, but really how many of us can actually use it? By time it gets down to a price where we can afford, getting dual-processing workstations seems like a better idea.
But just say in 2004, IBM releases that 6GHz 970 (minus the bugs, of course). How long would IBM have the upper hand? If IBM and Apple (as well as AmigaOne makers…) are the only one using it for workstations, the future for it is bleak. Why spend more money on something not all that profitable?
The power processor line is key to IBM’s hardware platform. The pSeries (used to be RS6000) and RS/6000 SP are very profitable. Further there is talk that the iSeries (used to be AS/400) will move to the power chip. So most of the work that goes into the 970 will be covered by other parts of IBM’s business. And these parts are very profitable
The only things the 970 has to pay for is:
a) The cost of manufacture
b) The R&D cost of taking the low yield high cost Powerchips to high yield low cost 970s.
Apple’s sales in and of themselves help cover the costs but even without Apple IBM would want something like the 970. More important to IBM is the 970 makes them able to put low end pSeries boxes in (that only cost a few thousand) in place of Xseries (x86 PCs) boxes for customers. This means they automatically sell DB2 not SQL server or Oracle, Lotus Notes not Exchange and Websphere and not a .Net solution. That can drive huge profits for many years.
Finally if they can get Apple to standardize on the 970 it might open the door to an even larger strategic alliance. Apple becomes the desktop of choice for IBM and in exchange IBM becomes the business server / consulting solution of choice for Apple.
For example in print technology there is really no reason that Apple couldn’t eventually change directions, lead their customers away from Adobe technology like postscript and pdf and move them towards afp. With afp large print shops would be buying IBM printers and not Xerox, using IBM consultants and not Xerox consultants, etc…
In other words I’m not sure the 970 ever has to be profitable in and of itself for it to be highly profitable for IBM. From Apple’s perspective having a processor in their iMac which scales up to something like a 64 way iSeries machine (with Powerchips) running OS/400 in Darwin emulation mode….
Epic makes a lot more sense than current out of order processors. Imagine a scientific computing application. Imagine a program that could take years of computing time to run, but because of a good compiler, it takes you an extra hour to compile. I would most certainly wait an hour to wait for the compiler to try to find the best way of doing things to cut months off of run time. And as long as you dont recompile, you get the speed up at every run. A large amount of transistors can be dropped if everything is explicitly parallel. Efficient assembly programming becomes much harder, but thats a whole other point. Considering the amount of code written in high level languages, it almost makes sense for intel to rely on compiler writers to know whats going on under the hood. I think Itanium could be an extremely competitive processor. I think 64-bit is mostly marketing hype though. Who needs (I think it’s 42 bit addressable physical) 4 terabytes addressible to run gaim and mozilla? I dare a desktop user to prove they need more than 4 gigs addressable for any one program at a time.
http://www.theinquirer.net/?article=6482
The power processor line is key to IBM’s hardware platform. The pSeries (used to be RS6000) and RS/6000 SP are very profitable.
Never said it wasn’t. POWER is a important part of IBM’s enterprise server line-up.
The only things the 970 has to pay for is:
a) The cost of manufacture
b) The R&D cost of taking the low yield high cost Powerchips to high yield low cost 970s.
This itself isn’t cheap. POWER is a processor that traditionally needs a lot of power, cache, and emmits a lot of heat. Transforming POWER into a PowerPC-compatible processor require a lot of R&D money.
Cost of manufacturing doesn’t really cost. Traditionally, PowerPC (and other RISC-based) processors cost less to manufacture than x86. Plus with IBM new fabrication plant, you obviously see now this isn’t close to my point. My point is that Intel have every reason to release a 6GHz processor 18 months from now, while IBM have less reason to do so.
Apple’s sales in and of themselves help cover the costs but even without Apple IBM would want something like the 970.
Obviously. Most of IBM’s G3 sales actually aren’t with Apple, but with networking routers…. Besides, I could see how 970 could fit in IBM’s eServer line up.
More important to IBM is the 970 makes them able to put low end pSeries boxes in (that only cost a few thousand) in place of Xseries (x86 PCs) boxes for customers.
You mean iSeries? Anyway, while this is possible, it is also unlikely. Many iSeries customers depend on x86-only apps, especially those who decide to take the Windows NT route.
This means they automatically sell DB2 not SQL server or Oracle, Lotus Notes not Exchange and Websphere and not a .Net solution. That can drive huge profits for many years.
But unless IBM makes 970 a extremely closed platform, Microsoft SQL (which doesn’t run on IBM Linux-based servers anyway) and Oracle could possibly run on IBM machines. Same with Exchange, not available on Linux. (Besides, how is WebSphere a direct competitor to .NET?). IBM servers, on the low end, don’t have a huge market share. If they make it all 970, it doesn’t mean the market is following them. The biggest player now is HP.
Finally if they can get Apple to standardize on the 970 it might open the door to an even larger strategic alliance.
Apple once standardize on the G3, that didn’t open doors to strategic alliances like you mentioned.
Apple becomes the desktop of choice for IBM and in exchange IBM becomes the business server / consulting solution of choice for Apple.
Unlikely. First, IBM would need to keep Windows as the prefered desktop of choice mainly on two reasons
– Windows have a much larger marketshare
– Apple have little to no stake in the enterprise, which IBM is targeting.
Plus, with Apple’s XServe, trying to hit on low end iSeries, a alliance is as likely as a alliance between Sun and Microsoft.
For example in print technology there is really no reason that Apple couldn’t eventually change directions, lead their customers away from Adobe technology like postscript and pdf and move them towards afp.
Why would they do they? They already have large investments in PDF technology. Heck, even OS X depends on PDF.
With afp large print shops would be buying IBM printers and not Xerox, using IBM consultants and not Xerox consultants, etc…
Possible, but I don’t really understand why this can’t happen with PDF and Postscipt. (Plus, most businesses pick Xerox because of price. I don’t really know all that much about this business, but I notice maintainence cost with Xerox is very low).
From Apple’s perspective having a processor in their iMac which scales up to something like a 64 way iSeries machine (with Powerchips) running OS/400 in Darwin emulation mode….
Having something as fast as a POWER processor is unlikely. Besides, having just moved from G3 to G4, it is unlikely that iMacs/eMacs would change processor in such a short time again.
This itself isn’t cheap. POWER is a processor that traditionally needs a lot of power, cache, and emmits a lot of heat. Transforming POWER into a PowerPC-compatible processor require a lot of R&D money.
Cut the cache, fix the manufacturing process… Yes its expensive but Intel for example has to do this every year to take their x86 lines and make them good for laptops…
Cost of manufacturing doesn’t really cost. Traditionally, PowerPC (and other RISC-based) processors cost less to manufacture than x86. Plus with IBM new fabrication plant, you obviously see now this isn’t close to my point. My point is that Intel have every reason to release a 6GHz processor 18 months from now, while IBM have less reason to do so.
If you agree they have every reason to do it for the Power chips after you agree to the below regarding the role of the 970 you can see how this follows.
Jeff: More important to IBM is the 970 makes them able to put low end pSeries boxes in (that only cost a few thousand) in place of Xseries (x86 PCs) boxes for customers.
Rajan: You mean iSeries? Anyway, while this is possible, it is also unlikely. Many iSeries customers depend on x86-only apps, especially those who decide to take the Windows NT route.
iSeries = AS/400s
xSeries = Intel based
So I didn’t mean xSeries. I assume you did too.
Jeff:
This means they automatically sell DB2 not SQL server or Oracle, Lotus Notes not Exchange and Websphere and not a .Net solution. That can drive huge profits for many years.
Rajan: But unless IBM makes 970 a extremely closed platform, Microsoft SQL (which doesn’t run on IBM Linux-based servers anyway) and Oracle could possibly run on IBM machines. Same with Exchange, not available on Linux. (Besides, how is WebSphere a direct competitor to .NET?). IBM servers, on the low end, don’t have a huge market share. If they make it all 970, it doesn’t mean the market is following them. The biggest player now is HP.
The 970 would be open, but at the same time Microsoft doesn’t make most products for powerPC lines. Sure HP wouldn’t follow but that doesn’t stop IBM from pushing towards this sort of solution.
Jeff:
Finally if they can get Apple to standardize on the 970 it might open the door to an even larger strategic alliance.
Rajan:
Apple once standardize on the G3, that didn’t open doors to strategic alliances like you mentioned.
Sure but at that point App, Motorolla and IBM had different goals. Apple now would be interested in the alliance.
Jeff: Apple becomes the desktop of choice for IBM and in exchange IBM becomes the business server / consulting solution of choice for Apple.
Rajan: Unlikely. First, IBM would need to keep Windows as the prefered desktop of choice mainly on two reasons
– Windows have a much larger marketshare
– Apple have little to no stake in the enterprise, which IBM is targeting.
Plus, with Apple’s XServe, trying to hit on low end iSeries, a alliance is as likely as a alliance between Sun and Microsoft.
As for larger market share there is little added value for IBM in the Wintel market place for this very reason. IBM could push for “better not more common”. As for Apple no stake in the enterprise if IBM were pushing Apple machines I think their attitude would change. Apple has no credibility with the enterprise marketplace but I don’t think they would object to selling to it.
As for the Apple Xserve, IMHO this really is a niche product for Apple and they would be willing to dump it. Further it is more aimed at small-mid business (large business doesn’t need super easy to administer servers) which isn’t IBM’s core market. Further, IBM might not even mind selling the XServe to customers for locations where hands on administration is needed but skills are not highly present (for example international).
For example in print technology there is really no reason that Apple couldn’t eventually change directions, lead their customers away from Adobe technology like postscript and pdf and move them towards afp.
Why would they do they? They already have large investments in PDF technology. Heck, even OS X depends on PDF.
I understand that Apple is strong in PDF/Postscript and has been for years. The problem is that unlike the 80’s Apple really doesn’t have huge advantages in the PDF/Postscript world over PCs. They are better but it is close. There are definite advantages of AFP over postscript, and vice versa.
But the major reason they would do it is that IBM would pay them to do it. Apple owns the print shop world. A switch to AFP would push IBM technology hard into print shops where IBM could sell:
printers
enterprise print managers
rip engines
consulting services
all stuff that Apple doesn’t make.
Conversely Apple gets (for example):
a) Direct input into the print language (i.e. they get control over AFP)
b) All sorts of powerful rendering technology that IBM has developed over the years
c) A huge library of very good quality fonts that they could give away free. Free good quality fonts is a big issue.
d) High power systems for document composition which IBM currently sells bundled
e) High end print management
In other words a strategic partnership. IBM gives technology and assistance to Apple which lets them pull ahead of PCs.
Jeff: With afp large print shops would be buying IBM printers and not Xerox, using IBM consultants and not Xerox consultants, etc…
Possible, but I don’t really understand why this can’t happen with PDF and Postscipt. (Plus, most businesses pick Xerox because of price. I don’t really know all that much about this business, but I notice maintainence cost with Xerox is very low).
IBM doesn’t really make production postscript printers (they do/did make desktop / workgroup postscript but that is much easier). IBM’s investment has been entirely around AFP. Xerox is years ahead of everyone else in terms of Postscript.
Xerox otoh doesn’t own the rights to pdf/postscript so an alliance there is more difficult for Apple.
Jeff: From Apple’s perspective having a processor in their iMac which scales up to something like a 64 way iSeries machine (with Powerchips) running OS/400 in Darwin emulation mode….
Having something as fast as a POWER processor is unlikely. Besides, having just moved from G3 to G4, it is unlikely that iMacs/eMacs would change processor in such a short time again.
So it takes another year or to for the iMac/eMac. I’m talking more long term strategy here in terms of the 970.