x86-64 vs. Itanium

[CTho9305]

<< You can load Windows 95 on an Itanium and it would work. >>


and as I'm told, run it about the speed of the CPUs of the time, too
will future IA64 processors have faster IA32 performance, or slower?

 

[Nothinman]

will future IA64 processors have faster IA32 performance, or slower?

I doubt it will get a whole lot better, I mean why buy an Itanic and then complain that it runs NT 4 poorly?

 

[Locutus4657]

You're wrong on several counts... However I suppose that you are write that C/C++ code written specifficaly for spark *with* in line assembly *specifically for* sparq chips won't run well on a Hammer. As a matter of fact it wouldn't compile on a hammer because the hammer will be missing those sparq instructions... But then again the same thing *IS TRUE* for any IA-64 chip which will ever be developed becaue the instruction set is different... However if we are talking about *standard* C/C++ code then your statement is not provable until benchmarks can be done. It is expected that AMD's hammer will be *faster* than Intels Itanic currently. Also, AMD with its integrated memory controller, 128bit DDR interface and hypertransport tunnels is designed to scale *MUCH* further than Itanic with its classic intel shared memory bus. As a matter of fact it would be much easier to take Hammer to the 32CPU level than it would be Itanic. The only reason why AMD isn't being that ambitiouse is because they know they have a bit more work to do on the Motherboard side before a company like Unisys will pick up their chip. As far as Non-validated motherboards... Do you think anyone in the market we are talking about cares about Via or any motherboard based on their chipset? Really, we're talking about the server market here... You really think AMD is going to try and penetrate this high end market with Via? Why do you think the Hammer has an integrated memory controller!



<< Besides marketing propaganda what makes you say this?

<< I don't think comparing IA64 to Hammer is a correct comparison at all. They are aimed at completely different markets and run entirely different software. If you run software that's normally written for the x86 market, yes it won't run well on IA64. If you run high-end server software made to run on RISC CPU's such as Spark and Power, it won't run well on Hammer. >>


Highly optimized Sun software won't run well on anything else than a Sun system. Highly optimized HP software wont run well on anything but an HP system. From what i've seen modularity and portability are a joke for very high end systems. High end server markets have software that are designed specifically for that system. IA64 is designed for scalability for up to hundreds of CPUs and Hammer officially only goes up to 8. Their markets will overlap at the "low end" but certainly not anything above that.

And do you seriously think that the initial Hammer will be cheap? Its in my opinion that they will go for prices in the few thousand for CPU alone, because AMD repeated stated that they were solely targetted for servers.

Another reason why AMD hasnt had market penetration in the workstation/server market is because they dont have the validation of the likes of the others. The only AMD board that has the decent validation is the 760MPX. Go ask any major company whether they perfer an Intel or AMD system, most of them wont even poke an AMD system with a stick mainly because of validation and questionable stability (especially with VIA chipsets). Most poeple acknowledge that AMD is a better price/performance ratio, but most wouldnt care to lose even one bit of stability for a decent gain in performance (hence i845's domination in the oem market despite SIS solutions beating it performance and price wise). End users like most of us wouldn't care less because I for one dont care if my computer randomly crashes like twice a year. >>

 

[Locutus4657]

Actually AMD's 7th geration Athlon was targeted at the workstation market, AMD's hammer is targeted at the low to midrange server market which is dominated by Xeons. AMD want's a big Iron CPU, and they have a real contender with the Hammer.




<<

<< Besides marketing propaganda what makes you say this?



<< I don't think comparing IA64 to Hammer is a correct comparison at all. They are aimed at completely different markets and run entirely different software. If you run software that's normally written for the x86 market, yes it won't run well on IA64. If you run high-end server software made to run on RISC CPU's such as Spark and Power, it won't run well on Hammer. >>

>>



The difference between Sun's Spark and IA64 isn't as drastic as x86-64 and Spark. x86-64 will be aimed at the workstation market, which runs entirely different software than the high-end server market. Currently, that workstation market uses Xeons (mainly) and hence, x86 makes sense. x86-64 may even make sense. But the high-end server market has said goodbye to x86 long ago. I highly doubt they'd want to go back. >>

 

[Nothinman]

This is bothering me, it's Sparc, not Spark or Sparq.

 

[Locutus4657]

Sorry... But anyway, code written specifically for it still wont run on any IA-64 or x86-64 chip ever implemented by any company anyway....



<< This is bothering me, it's Sparc, not Spark or Sparq. >>

 

[Nothinman]

Sorry... But anyway, code written specifically for it still wont run on any IA-64 or x86-64 chip ever implemented by any company anyway....

Maybe Coruso (or whatever Transmeta calls it) that has the translation layer they talked up so much.

 

[Locutus4657]

I'm sure that transmitta could make such a chip, but it would be worthless. Seeing as you can buy a Sun Blade for less than $1,000 there's no point for Trasmetta to come up with some "low cost" low preformance version. Sun offers systems low priced enough for even the average consumer to afford if they wanted a sun for whatever reason. And the preformance is good enough for basic development, something that you wouldn't be able to say for any Sparc Cruso.



<< Sorry... But anyway, code written specifically for it still wont run on any IA-64 or x86-64 chip ever implemented by any company anyway....

Maybe Coruso (or whatever Transmeta calls it) that has the translation layer they talked up so much. >>

 

[dexvx]

Standard C/C++ on high end servers? You're joking. First and foremost, if a customer is plopping millions of dollars to buy such high end equipment it might as well that the software for those systems be specifically hardcoded for that system. Like I said, the reusability of such code is near negligible. There is NO reason to make code on high end servers portable enough so competing platforms can run them. Tahts why upgrading them is such a pain.

As for Itanium's scalability, it has already been proven. IBM and HP have already built 1000+ processor Itanium servers. You *will* need to write specific code for them to ensure they work optimally, but if you're plopping down the money for a 1000+ cpu server, you'd better well make sure it runs optimally.

As for Sun Blade systems under $1000, I havent been able to mess with Sun systems recently, but from the likes of previous use, they're NOT a uber high performing system. A $1000 Blade system consists of an Ultra Sparc IIe @ 500Mhz, 128MB SDRam, 20GB HDD, and Solaris 8. Unless you get some specific sun optimized software, I fail to see the point in buying Blade systems when you can configure better performing AMD MP/Northwood systems for the same price range.

 

[Nothinman]

As for Sun Blade systems under $1000, I havent been able to mess with Sun systems recently, but from the likes of previous use, they're NOT a uber high performing system. A $1000 Blade system consists of an Ultra Sparc IIe @ 500Mhz, 128MB SDRam, 20GB HDD, and Solaris 8. Unless you get some specific sun optimized software, I fail to see the point in buying Blade systems when you can configure better performing AMD MP/Northwood systems for the same price range.

For real Sun/Sparc experience? Not all of us have the good fortune of being able to play on work's machines and for $1K it's not a bad deal. Especially considering you can add more memory and a bigger hard disk pretty cheaply. For business use it's debatable.

 

[Locutus4657]

Tell that to IBM, they're all about Java technology on their high end servers. Yes, C/C++ have a place in the high end server world, it's sane to write your applications in C/C++, the programs produced by C/C++ compilers are fast and it is not sane to do this in assembly. That's why all of the major parallel libraries are written for C/C++, as well as fortran for those stubborn engineers who refuse to learn a sane language. I haven't heard of 1000+ Itanium systems, a link to a web page describing them would be helpful. I don't doubt it, probably distrubuted memory systems (like most 1000+ CPU systems). My point is scaling hammer is easier than scaling Itanium, due to the bus architecture of both CPU's, something which is hard to argue. $1000 Sun Blades deffonetly have a place in the world of bussiness, if you want to be able to develope *real* Sparc programs with out putting the extra load on your server, or an extra big dent in your wallet. Just think, programmers can work on in expensice Sun Blades, then port their work to the server once everything is debuged. The Sun Blade isn't a server platform, it's a development platform, and one of Suns better ideas!



<< Standard C/C++ on high end servers? You're joking. First and foremost, if a customer is plopping millions of dollars to buy such high end equipment it might as well that the software for those systems be specifically hardcoded for that system. Like I said, the reusability of such code is near negligible. There is NO reason to make code on high end servers portable enough so competing platforms can run them. Tahts why upgrading them is such a pain.

As for Itanium's scalability, it has already been proven. IBM and HP have already built 1000+ processor Itanium servers. You *will* need to write specific code for them to ensure they work optimally, but if you're plopping down the money for a 1000+ cpu server, you'd better well make sure it runs optimally.

As for Sun Blade systems under $1000, I havent been able to mess with Sun systems recently, but from the likes of previous use, they're NOT a uber high performing system. A $1000 Blade system consists of an Ultra Sparc IIe @ 500Mhz, 128MB SDRam, 20GB HDD, and Solaris 8. Unless you get some specific sun optimized software, I fail to see the point in buying Blade systems when you can configure better performing AMD MP/Northwood systems for the same price range. >>

 

[pm]

<< I haven't heard of 1000+ Itanium systems, a link to a web page describing them would be helpful. >>


The latest announcement is here for a 1400 Itanium system for use by the US Department of Energy. This was announced back about two weeks ago. This is, if I recall correctly, the third announcement of planned deployment of a large scale Itanium system. The other two were some time ago and finding links to them would be harder, but I can look them up if you are suffuciently interested. I believe one of the other machines was being built by IBM.

 

[Nothinman]

The problem I have with that announcement is this : "I'm pleased that the U.S. Department of Energy will now have the world's most powerful Linux-based computer for our pivotal work in biological and environmental research,"

Seeing as how Linux can not use 1400 CPUs, I believe 32 is the hard limit right now as a int bitmask is used to identify CPUs. So that means it must be a cluster of many dual, quad or 8-way boxes.

 

[Locutus4657]

This is exactly what I was saying... 1000P systems don't really show off the scalability of a processor because frankly all such systems (that I'm aware of) are distributed memory systems. This means you arne't showing how your processor can scale in a shared memory system, it shows how your networking technology can scale (we all know how well networks scale). What I'm interested in is what is the largest *shared memory* system being developed...



<< The problem I have with that announcement is this : "I'm pleased that the U.S. Department of Energy will now have the world's most powerful Linux-based computer for our pivotal work in biological and environmental research,"

Seeing as how Linux can not use 1400 CPUs, I believe 32 is the hard limit right now as a int bitmask is used to identify CPUs. So that means it must be a cluster of many dual, quad or 8-way boxes. >>

 

[Degenerate]

<< Seeing as how Linux can not use 1400 CPUs, I believe 32 is the hard limit right now as a int bitmask is used to identify CPUs. So that means it must be a cluster of many dual, quad or 8-way boxes >>



So does that mean AMP Cpu's can go to that many as well? Right now, they have dual configs. So they can cluster that to like 2X 700?

 

[Nothinman]

AMD should have an easier time going a lot higher than dual, because of the hypertransport which I believe they licensed from Compaq/Digital before the Alpha IP was sold to Intel.

 

[Locutus4657]

No, actually AMD invented Hyper Transport and made it an open platform. AMD did licence EV6 from Compaq/Digital however.



<< AMD should have an easier time going a lot higher than dual, because of the hypertransport which I believe they licensed from Compaq/Digital before the Alpha IP was sold to Intel. >>