Sign in with
Sign up | Sign in
Your question

Intels 64 bit mode sucks

Last response: in CPUs
Share
April 4, 2004 10:24:03 AM

<A HREF="http://www.heise.de/ct/04/08/020/" target="_new">Andreas Stiller's Prozessorgeflüster </A>

<A HREF="http://babelfish.altavista.com/babelfish/urltrurl?url=h..." target="_new"> babelfished </A>


" "It sucks" - so the abschaetzige comment of a beta tester - in particular if larger address ranges are addressed. Perhaps the processor must still emulate 64 bits here and there, who knows. Intels 64-Bit-Partner HEWLETT-PACKARD will know it anyhow - and that decided only once for Opteron. "

= The views stated herein are my personal views, and not necessarily the views of my wife. =

More about : intels bit mode sucks

April 4, 2004 2:50:05 PM

Ouch... If true, this is no good news for Intel.

But I'm still a little skeptical that they can't iron it out by the actual time Nocona gets launched. So far, this is all Internet rumors. Not to say it's not true, but I'd rather wait and see.

A company over four times bigger than AMD, unable to get this right? Maybe in the earlier versions, but they will certainly get it. Or maybe it's just my humanist side speaking.

<i><font color=red>You never change the existing reality by fighting it. Instead, create a new model that makes the old one obsolete</font color=red> - Buckminster Fuller </i>
April 4, 2004 2:56:58 PM

Babelfish suckes too. I just read it twice and still don't know what it was trying to say. :frown: "...smaller American small towns godfather confessed for the next year to the planned successors."
Related resources
April 4, 2004 6:52:27 PM

The article is interesting...

I just realised after reading it that all the Intel/AMD cores are named after towns and cities.

Why?




A long long time ago, but I can still remember, how that music used to make me smile... <A HREF="http://www.nexus.hu/zonix/DIGGER.MID" target="_new"><b><font color=blue>Digger rulz</font color=blue></b></A>
April 4, 2004 7:57:51 PM

And rivers

Dichromatic for your viewing plesure...
April 4, 2004 9:44:23 PM

HP is still gonna use Itaniums and 64-bit nocona Xeons. it's true that the ALU runs at only half speed while executing 64-bit registers on the Nocona, and it doesn't have that NX feature non-execute bit hacking protection overflow buffer thingy. But, Intel is gonna bank on the real 64-bit processor,'Itanium'. Nocona is made available for servers and high end only, not desktop.

-------
(2x512mb) 1GB DDR333 Dual Channel
INTEL Pentium 4 2.8B
ATI RADEON 9800 PRO
Asus P4P800 Bios 1016
38,476 Aquamarks
April 4, 2004 9:47:35 PM

too much german to read. It takes me too long. Maybe i should of payed more attention in german class.
April 4, 2004 10:24:03 PM

Quote:
But, Intel is gonna bank on the real 64-bit processor,'Itanium'.

"Real" meaning...?

Whatever it means, it may have some value to Intel, but the rest of the world isn't realy going to care.

For all IA64's claims of being a "real" (non-evolved?) 64-bit architecture, AMD64 does just as well or better in most common cases. So far AMD64 even outperforms IA64 for the 64-bit features (i.e. integer math and addressing in 64 bits). IA64 excels at floating-point math, but that has nothing to do with its so-called "real" 64-bitness.

Quote:
Nocona is made available for servers and high end only, not desktop.

That doesn't really count for much. It just means one market (server/workstation customers) is going to be annoyed with Intel about crummy 64-bit performance, while the other market (consumer desktop customers) is going to be annoyed with Intel about no 64-bit extensions. The half-assed effort known as Clackamas Tech isn't going to woo any properly informed customers or developers away from AMD64.

<i>Look! Up in the sky! It's a bird! It's a plane! It's...

...an asthmatic werehamster?

<LHGPooBaa> Well, @#!& on me.</i>
April 4, 2004 10:33:00 PM

i doubt it. Alpha died because they did not have the resources to keep the chip alive. Intel is worth 50 billion dollars. They just announced 2 new 1.40ghz, 1.50ghz Itanium2 with 3-6mb cache and new software emulator(ia-32 execution layer) which will make Itanium run as fast as a 1.5ghz Xeon in 32-bit mode.

it doesn't look like Intel is ditching the itanium, in fact it looks like they are stenghtening it. Now that AMD standing on Cisc x86-64 architecture , Intel is going with Risc architecuture, which is exclusive and has no competition.

<font color=blue><b>Intel(R)</font color=blue></b>
(2x512mb) 1GB DDR333 Dual Channel
INTEL Pentium 4 2.8B
ATI RADEON 9800 PRO
Asus P4P800 Bios 1016
39,470 Aquamarks
April 4, 2004 10:54:12 PM

I have some difficulties labeling Alpha as a failure. DEC failed selling it perhaps, but Alpha was by far the best cpu for most of the last decade, and arguably, even though it has been robbed of any serious ongoing R&D funds, its still one of the fastest cpu's available. Alpha was (and is) technically an increbible succes, commercially, a minor failure :) 

= The views stated herein are my personal views, and not necessarily the views of my wife. =
April 4, 2004 10:58:18 PM

>But I'm still a little skeptical that they can't iron it
>out by the actual time Nocona gets launched.

Launch is imminent, its not likely they will change whatsoever now. Wether or not they can sort it out B4 desktop chips hit the shelves is another thing, but my guess is Intel is not going to spend much effort on redesigning Prescott now, so it will most likely take until Tejas b4 any issues with AMD64 are ironed out. Coincidentally, this is pretty much what I predicted earlier this year :) 

= The views stated herein are my personal views, and not necessarily the views of my wife. =
April 4, 2004 11:02:08 PM

LOL.. don't sweat it, most of the article is about upcoming chips and its code names (derived from city's or rivers, for both intel and AMD), but doesnt contain anything we didnt know yet. Apart for the claims Nocona's 64 bit performance would suck obviously.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
April 4, 2004 11:17:53 PM

>i doubt it. Alpha died because they did not have the
>resources to keep the chip alive.

It died mostly because it didn't sell, in spite of performance that wasnt even in the same league as its competitors most of the time. Blame DEC for the poor marketing; Alpha outperformed competing architectures by a factor 2 or more. If there is one lesson to be learned from Alpha, its that performance alone doesnt sell.

>t doesn't look like Intel is ditching the itanium, in fact
>it looks like they are stenghtening it.

No, of course intel isnt ditching it yet. they still have a multi billion dollar investment to recover. Dollar adjusted, the costs so far would be roughly around 5-7 billion dollar.

The real question is: will they ever recover this investment? I personally doubt it. Simple math learns you that in the niche Itanium is currently competing, there just isnt the required volume to finance even the ongoing R&D, let alone the initial development costs if you are only selling the cpu's and not entire systems, along with software and services like IBM, HP, SGI, Sun, etc.

Intel sold 100.000 Itanium chips last year, and they where quite obviously extremely happy with that result. The harsh reality however is that 100k chips brought them some $150M in revenue, probably around $120M or less gross margin, while intel spends $4-$500M per year just in developping those chips, chipsets, compilers, etc. They can keep that up for a few years, intels pockets are deep enough, but if IA64 doesnt go mainstream/high volume over the next few years, it doesnt make any sense for intel to keep it alive.

I've done quite a bit of financial simulations, and frankly, no matter how you look at it, if you do not foresee IA64 to ever replace x86 when and where 64 bit becomes an issue, IPF doesnt make sense financially. With the announcement of "EM64T" aka iAMD64, it has become obvious IA64 will not likely ever replace x86, and therefore, IPF is doomed to be an ongoing major financial loss. Its not for no reason HP wanted to outsource CPU development, its just much too expensive in a market where you sell a couple of 100k's units a year. And that is even in spite of the fact HP, unlike Intel, reaped the benefits of selling entire systems along with software and services in highly lucrative markets. Intel doesnt share that advantage. I firmly expect Intel to pull the plug on Itanium in ~5 years, and/or hand the project back to HP. HP might use it as a replacement for PA Risc, and swallow the losses in the CPU development by selling S&S, intel can not.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
April 4, 2004 11:21:53 PM

Quote:
i doubt it. Alpha died because they did not have the resources to keep the chip alive. Intel is worth 50 billion dollars.

Alpha died because it didn't pay for itself. DEC/Compaq had a hard time reaching break-even point in the Alpha's niche market.

In case you haven't noticed, Intel's having the same problem. Having billions and billions in reserve is, commercially, good for no more than a temporary band-aid fix, and even then it ticks off shareholders to see tons of money get poured into something that can't seem to generate profits.

Quote:
it doesn't look like Intel is ditching the itanium, in fact it looks like they are stenghtening it.

DEC spent years strengthening Alpha as well. That didn't keep it from getting kicked to the curb though. The kicking began when Compaq inherited the platform and discovered just how much R&D cash it could burn through.

Quote:
Now that AMD standing on Cisc x86-64 architecture , Intel is going with Risc architecuture, which is exclusive and has no competition.

IA64 is <i>not</i> RISC. IA64 is VLIW, which if anything is dramatically more complex than CISC.

Besides, being non-CISC never protected any other architecture from competition by the x86 world. It's dismally idiotic to think it might protect IA64.

<i>Look! Up in the sky! It's a bird! It's a plane! It's...

...an asthmatic werehamster?

<LHGPooBaa> Well, @#!& on me.</i>
April 4, 2004 11:28:40 PM

> it's true that the ALU runs at only half speed

I sincerely doubt that. Got a link ?

>while executing 64-bit registers on the Nocona,

You havent got a clue what you're talking about, but what else is new.

>Intel is gonna bank on the real 64-bit processor,'Itanium'.

The banked on it, and lost their bet. The anouncement of AM64T/Clackamas/IA32e/AMD64/CT/iAMD64/x86-64/whatever name they come up with next, just about killed any hopes anyone could have had of Itanium becoming anything else than a low vomume high end niche product competing with Sparc & power.

>Nocona is made available for servers and high end only, not
>desktop.

Good thinking Einstein, Nocona is a Xeon. the irony however, is that iAMD64 will only be enabled for Nocona, which is limited to 2 way operation. >2 way Xeons (the MP's) will remain 32 bit until somewhere next year. What do you think requires 64 bit addressing more, an entry level 2 way server, or a 4/8/16/32+ way machine ? No wonder HP, IBM and mostly Sun are jumping on the opteron bandwagon for 2/4/8 and even 8+ way servers (in casu Sun).

= The views stated herein are my personal views, and not necessarily the views of my wife. =
April 5, 2004 3:00:52 AM

The rapid execution engine in netburst runs at 2x of core frequency. P4 3.2ghz has 6.4ghz rapid execution engine. rapid execution is Intel's marketing term for ALU. i've read that under 64-bit extended mode, this ALU runs at half the speed. meaning, it will run at 1x of core frequency in 64-bit mode or at 3.2ghz only.

"In any event, the Opteron will be faster doing 64-bit processing because the P4/Xeon loses their double-pumped ALUs when switching modes. " Ardrid, AMDforums




<font color=blue><b>Intel(R)</font color=blue></b>
(2x512mb) 1GB DDR333 Dual Channel
INTEL Pentium 4 2.8B
ATI RADEON 9800 PRO
Asus P4P800 Bios 1016
39,559 Aquamarks
Anonymous
a b à CPUs
a b α HP
April 5, 2004 6:50:16 AM

For all IA64's claims of being a "real" (non-evolved?) 64-bit architecture, AMD64 does just as well or better in most common cases. So far AMD64 even outperforms IA64 for the 64-bit features (i.e. integer math and addressing in 64 bits). IA64 excels at floating-point math, but that has nothing to do with its so-called "real" 64-bitness.


You over many never been to prove that.Get the fact itanium is faster and scalle better and got OS that go with it unlike opteron.

Fell refreshing
April 5, 2004 7:40:21 AM

hmmm, you say he cant prove that, well you prove what your saying as well, its only fair lol.

btw, i probably think itaniums beat out opterons in more then a few repects, it is built to be used a segment opteron isnt aimed at anyway. opterons compete with xeons, not itaniums.
Anonymous
a b à CPUs
a b α HP
April 5, 2004 7:52:14 AM

So you have a link to a complete commercial or HPC 4 way opteron or 8 way that come with 64 bit linux and apps systemes manager 64 bit compiler

Fell refreshing
April 5, 2004 8:14:32 AM

I know a few French Canadian hardware guys. They all write english (Tech) better than I do. They couldn't get through the texts otherwise. Are you actually spud having fun with us, or are you pretending you dont write english well because you are just full of crap.
April 5, 2004 9:14:05 AM

Yeah, same philosphy. But "same type of processor" only if you think the ARM or Xscale powering your PDA or MP3 player to be the "same type of processor" as an Alpha EV7 in a supercomputer (both are RISC).

= The views stated herein are my personal views, and not necessarily the views of my wife. =
April 5, 2004 9:25:35 AM

<A HREF="http://h18004.www1.hp.com/products/servers/proliantdl58..." target="_new">http://h18004.www1.hp.com/products/servers/proliantdl58...;/A>
<A HREF="http://www.appro.com/product/server_4144h.asp" target="_new">http://www.appro.com/product/server_4144h.asp&lt;/A>

Just 2 examples of 4 way opterons, that come with 64 bit linux, tons of 64 bit apps and 64 bit compilers (your choice gcc, portand, absoft, pathscale ..)

= The views stated herein are my personal views, and not necessarily the views of my wife. =
Anonymous
a b à CPUs
a b α HP
April 5, 2004 10:20:42 AM

also elbrus is VLIW

Fell refreshing
Anonymous
a b à CPUs
a b α HP
April 5, 2004 10:32:11 AM

HP 4 way is due to 2Q still not avaible.Appro 4Way or rebrand amd quarlet seen be to avaible.

You get kinda limited if your out of C++ i found limited support for fortran and nothing for cobalt.No Windows 64 bit.ZERO software have certification.

Fell refreshing
April 5, 2004 10:41:13 AM

>HP 4 way is due to 2Q

so ? Where can I buy Nocona chips ?

> 4Way or rebrand amd quarlet seen be to avaible.

Its shipping. Again, where is Nocona ?

>You get kinda limited if your out of C++

yeah C++ is a rare language.

> i found limited support for fortran

Hu ?

> nothing for cobalt

Cobalt ?? You mean Cobol ? Oh whatever you mean, i don't care.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
April 5, 2004 2:18:56 PM

Quote:
You over many never been to prove that.Get the fact itanium is faster and scalle better and got OS that go with it unlike opteron.

It's common knowledge. Just check out SPEC scores. AMD64 edges out IA64 in integer ops, and IA64 edges out AMD64 in FP ops.

Besides which, both IA64 and AMD64 are in roughly the same position WRT operating system, except that AMD64 can perform well in 32-bit mode.

There's a version of Windows for AMD64, but it's in beta.

There's a version of Windows for IA64, but it's only available to a limited audience, has relatively little software available, and is probably going into bitrot.

Linux (and most Linux apps) is available for both processors. Probably the same goes for the BSDs.

IA64 only scales "better" if you slap it into a NUMAflex-based Altix box or something similar. In the far more common 8-way or less arena, IA64 is typically limited by a shared-bus topology. In that case AMD64 scales much better via ccNUMA topology.

Quote:
You get kinda limited if your out of C++ i found limited support for fortran and nothing for cobalt.No Windows 64 bit.ZERO software have certification.

Cobalt? You mean COBOL.

FORTRAN/COBOL support is limited because hardly anyone uses those languages anymore, especially for building 64-bit apps. Hardly anyone cares. People are more interested in C/C++/Objective-C/Java/Python/Perl/Ruby/PHP, all of which have been 64-bit ready for ages.

<i>Look! Up in the sky! It's a bird! It's a plane! It's...

...an asthmatic werehamster?

<LHGPooBaa> Well, @#!& on me.</i>
April 5, 2004 4:05:16 PM

Quote:
I know a few French Canadian hardware guys. They all write english (Tech) better than I do.

I am one of these French Canadian (or "Frogs"). :smile:

--
Would you buy a potato powered chipset?
Anonymous
a b à CPUs
a b α HP
April 6, 2004 11:07:43 PM

http://www.aceshardware.com/SPECmine/index.jsp?b=0&s=1&...

http://www.aceshardware.com/SPECmine/index.jsp?b=1&s=1&...

I will admit that that this score for Itanium 2 is reach with HP compiler.On FP code there more that 50% advantage.On scaling as the number of cpu scale itanium leave opteron futher and futher.I cannot compare many thing as there ZERO submission for anything bigger that 4 way.

On a average 8 way itanium will still have access to a minimun of 12 to 24 GB/S of ram with either numaflex superdome rxXXXX ZX1 or E8870.So all chipset avaible.
\\\\\formula use 4 or 2 cpu per memory hub 4 or 8 channel per hub PC1600 or PC2100 use on the channel/////////

Most seem to have forget that overall latency increase by 25 ns (50 clock cyle for a X46) by moving from 2 way to 4 way.

OS support windows XP is coming you will not plan to ship that with a 4 way or even a 2 way anyway there be license issue for 4 way.Windows 2003 is set to be in 2005.Windows XP and 2003 is ready for IA-64.No major unix is ready for opteron EX: HP-UX tru-unix64 solaris AIX.Only 1 will be port to opteron, solaris if the budget was not cut in the last layoff of sun.Linux support for Itanium and opteron is there except suse and redhat have certification for year and apps with certification for IA-64 and there relatif linux distro.Freebsd i never see or hear a server with this maybe you got a link.


On the very old fortran/cobol there still some place that use it as at my work there still some cobol on the mainframe those software been rework and work flawless.Many corp have in-house software that run for the last year.Some may think mainframe is dead but ibm/HP revenue for those have increase.Personnaly i dont have see any ASM compiler for X86-64


Have a good day mr Kelledin
Waiting for your reply
Fell refreshing<P ID="edit"><FONT SIZE=-1><EM>Edited by rahxephon on 04/06/04 07:18 PM.</EM></FONT></P>
!