Sign in with
Sign up | Sign in
Your question

HP to adopt Opteron

Last response: in CPUs
Share
Anonymous
a b à CPUs
January 26, 2004 9:21:57 PM

Stick a fork in Itanium, its done. Even HP is embracing Opteron, and while certainly it will be restricted it to 1-4 way initially (hence, not competing with its higher end >4 way Itanium enterprise servers), this is going to kill any momentum Itanium may have enjoyed. Can't wait for HP to announce its AMD64 port of HP-UX :) 

<A HREF="http://zdnet.com.com/2100-1103_2-5147619.html?tag=zdfd...." target="_new">HP to adopt AMD's Opteron in servers</A>

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

More about : adopt opteron

January 26, 2004 10:14:28 PM

What I said in that another post...

AMD, step by step, little by little, with Opteron is gaining high-end market, and the more important servers builders, first IMB, later Sun, now HP... this is unstoppable ;-)

<P ID="edit"><FONT SIZE=-1><EM>Edited by eugeneMC on 01/27/04 01:16 AM.</EM></FONT></P>
Anonymous
a b à CPUs
January 26, 2004 10:20:15 PM

I'm not sure about the "unstopable".. if intel announces a 64 bit capable, glueless SMP Xeon chip tomorrow, Opteron will quickly become just another xeon clone, and OEM's may ditch opteron in favour of "Xeon64" and intel's sweet co-marketing dollars. The only small downside of this strategy for intel would be that it would probably kill Itaniums chances to become anything but a loss generating, high end niche product for good.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
Related resources
January 26, 2004 10:35:08 PM

Looks like Itanium will be buried.
If Opteron will be used in more server from different computer companies then I think Itanium will not be able to gain ground again since <b>PRICE speaks louder than PERFORMANCE</b> in these days.

<b><i>"The Past Is Over;
A New Beginning"</i></b>
January 26, 2004 10:35:09 PM

Looks like Itanium will be buried.
If Opteron will be used in more server from different computer companies then I think Itanium will not be able to gain ground again since <b>PRICE speaks louder than PERFORMANCE</b> in these days.

<b><i>"The Past Is Over;
A New Beginning"</i></b>
Anonymous
a b à CPUs
January 26, 2004 10:44:42 PM

Hmmm.. there is little inherent to Itanium that makes it has to be (much more) expensive than Opteron; I doubt either chip costs more than $100 to produce (not counting r&d obviously). In fact, Opteron 8xx are priced in the same category as Itaniums. Either way, in the 4+ way market, cpu cost is only a small fraction of total system cost. In 16 way servers, its probably single digit percentages at most.

The real cost of itanium for end customers is not the chip, nor the system; its porting the software.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
January 27, 2004 1:29:33 AM

Once opteron has moved the market to 64 bit, Itanium will shine. Either intel will wait, or they will devolop a platform that supports xeon (with 64 bit) and itanium. They may be able to pull off a new platform quickly, but not too quickly.
January 27, 2004 5:01:13 PM

That a good 1 4 way opteron will destroy us all.

I dont like french test
January 27, 2004 5:13:29 PM

That allready in progress single chipset for XEON and itanium single socket also mothersboards will able to take Itanium and Xeon.There allready some OEM that sold systeme that support Iatnium and Xeon in the same box

I dont like french test
Anonymous
a b à CPUs
January 27, 2004 7:28:11 PM

A single 4 way opteron is not likely to destroy anything, but the significance of the fact that HP of all companies, decides AMD64 is a credible solution, can not be overstated. After all, HP dropped its own PA Risc and Alpha architectures in favor of Itanium as their single 64 bit server solution.. and now.. it also adds another 64 bit server platform: Opteron. What message do you think that will send to its customers ? Alpha wasnt good enough, but opteron is ? Sure, they will position initial opteron products in the "lower end" 4 way market (if you can call 4 way server market low end, that is), but that is where the bulk of Itanium sales go, not in those precious few 16-128 way systems. Now, what do you think is going to happen when these opteron boxes sell ? People will port their software to AMD64, and as their requirements and confidence in the platform grow, will start asking compatible higher end solutions. By releasing 4 way opteron servers, HP will actively help in creating a credible AMD64 market, and might well force intel into releasing an opteron clone, instantly killing just about any possibility of Itanium growing out of its current loss generating, high end, low volume niche markets (and even then, mostly only HPC).

= The views stated herein are my personal views, and not necessarily the views of my wife. =
January 27, 2004 10:18:10 PM

I'll quote the inquirer on this one, "How long before the Dell Domino falls?"

I to think this is pretty significant in showning the naysayers that AMD64 is a realy force to be reckoned with. But Dell....if they're xeon server sales start going to IBM, HP, and SUN operton setups you can bet they will have a ready to go product on the market tomorrow.
Anonymous
a b à CPUs
January 27, 2004 10:33:03 PM

who cares ? having the same socket may help lower the burden to IA64 adoption a little bit, like by allowing cheaper chipsets and mainboards, or an upgrade from xeon to itanium, but the main problem has nothing to do with sockets, or even cost of cpu's/chipsets/mainboards but everything with the cost of porting software. Heck, for many companies a $5.000 opteron system might be MUCH more interesting and FAR cheaper than a free Itanium server.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
January 27, 2004 11:01:26 PM

The drop in long term strategy.PA8800 dual core was release 2 week ago.EV7A will be release in few month and Open VMS still only support Alpha and itanium soon.

Ibm have give support for only 1 to 4 way same thing for HP.SUN are losing so much money solaris market share are in decline and 8 way are not avaible and i can guess they are not in developement.

I dont like french test
January 27, 2004 11:03:42 PM

With your logic you can guess why anyone would better like itanium

I dont like french test
Anonymous
a b à CPUs
January 27, 2004 11:08:41 PM

"only" 4 way is a gross misrepresentation. 1,2 and 4 way servers constitute the bulk of the market, both in revenue, and by a HUGE margin in unit sales.

Also, currently Opteron realistically only scales to 4 way in single image systems, 8 way if you want to sacrifice inter cpu latency. Cray and other companies are working on chipsets and HT bridges to allow >4 way opterons, and adding cache coherency to the HT link (which is really needed if you want a decent performing >4-8 way system), but today, even if HP/Sun or IBM wanted to bring a 16 way opteron server to market, they just can't -yet.

BTW, how long did it take before you could actually buy a (single image) Itanium with more than 4 cpu's ? At least two years I think...

= The views stated herein are my personal views, and not necessarily the views of my wife. =
Anonymous
a b à CPUs
January 27, 2004 11:14:10 PM

>With your logic you can guess why anyone would better like
>itanium

Its always easier to understand something with logic, than without it :)  Anyway, plain simple: not exactly every corporation has x86 legacy code it wants to port to 64 bit. How about any company running its mission critical stuff on HP-UX, what choice do they have but move to IA64 sooner or later ? And most of the HPC market loves Itanium's FP power, and doesnt care much about porting issues, as they (re)compile all the time anyway. Those markets are Itaniums natural habitats, but they do not represent a terribly big percentage of the entire market..

= The views stated herein are my personal views, and not necessarily the views of my wife. =
January 28, 2004 3:00:10 AM

Wow its nice that you read the Aceshardware review on how to kill x86. Seems as of late your limitless insight in how the market works and what theses high price point users think is quite remarkable.

-taitertot

If this post has attitude, seems to be overly aggressive, rude, distasteful to 99% of the users here, and shows a zealous defense of Intel... It’s probably Spud.
Anonymous
a b à CPUs
January 28, 2004 3:16:55 AM

Yeah I read it, what is your point ? Is there anything in that article that wasnt known already for ages ? Any radical new insights I might have parrotted ? copy pasted ? Or are you implying I am clueless, but read the article, and answered with just the knowledge contained in that article or something.. ? Well.. if that where the case, at least I apparently did a good job in making a short abstract.. although I don't even seem to recall Aces' mentioned the HP-UX factor, so just maybe I read a SECOND article on "expensive servers" somewhere ?

Oh well, why bother answering. You're spud right ? you're the guy that copy pastes entire intel white papers while changing a few words left and right (even making the information incorrect at times, cause you havent got a clue whats in those papers), and then pretends its your work ?

= The views stated herein are my personal views, and not necessarily the views of my wife. =
Anonymous
a b à CPUs
January 28, 2004 3:36:49 AM

I quickly reread aces' article.. sorry, i can't find anything in my posts here that would even seem inspired by that article. the only resmeblence is that both mention Itanium Opteron and HPC. maybe HPC was a new word for you that caught your attention ?

= The views stated herein are my personal views, and not necessarily the views of my wife. =
January 28, 2004 3:43:41 AM

Its quite ammuseing that you constantly post opinions and theories and speculations as fact. You are missleading many folks here by saying anything about HPC markets. Especially with no links of credable support.

But I do suppose thats what forums are about is opinions and speculations. I just find it unusual that the HPC market has anything to do with Opteron usage by HP to begin with. Since the news is just that HP is adopting the Opteron line not that they are going to be tossing them in their superdomes.

-taitertot

If this post has attitude, seems to be overly aggressive, rude, distasteful to 99% of the users here, and shows a zealous defense of Intel... It’s probably Spud.
Anonymous
a b à CPUs
January 28, 2004 3:59:43 AM

>Its quite ammuseing that you constantly post opinions and
>theories and speculations as fact.

anything that is in the future tense is pretty much by definition a theory/speculation, and if you think they are not realistic for whatever reason, feel free to counter my opinions. Anything I've written in the current tense, is to the best of my knowledge, fact.

>You are missleading many folks here by saying anything
>about HPC markets. Especially with no links of credable
>support.

To support what exactly ?

> I just find it unusual that the HPC market has anything to
>do with Opteron usage by HP to begin with. Since the news
>is just that HP is adopting the Opteron line not that they
>are going to be tossing them in their superdomes.

I mentioned HPC as a market where Itanium is currently rather successfull, mostly because of its FP performance, and the fact porting software if far less an issue than in eg a transactional environment. What so hard to understand about that ?

= The views stated herein are my personal views, and not necessarily the views of my wife. =
January 28, 2004 1:02:57 PM

Quote:
feel free to counter my opinions

.
With what, I don’t pretend to know anything about HPC users and their needs. When asked do I feel that the Opteron has an opportunity to make a good splash there based on what I do know of HPC users. I can safely say no since these guys do write their own code and regardless of the compiler dependencies the Itanium has it still has the highest IPC of any silicon currently available to the market. But again that’s my opinion based on information I've had a chance to read and what I have seen.

Quote:
I mentioned HPC as a market where Itanium is currently rather successful, mostly because of its FP performance, and the fact porting software if far less an issue than in eg a transactional environment. What so hard to understand about that?

Your talking enterprise servers last I checked and enterprise 4-8 way systems are classified as HPC. HPC is still currently owned by the Itanium, since it doesn't need to be in a superdome to be called enterprise.

Truth be known though buddy at Aces was right huge discounted Itanium servers are floating around. It’s tough to move over to a whole new environment so they need to entice users to move over. The Opteron will need the same procedure on the level of recoding\ compiling\ debugging

-taitertot

If this post has attitude, seems to be overly aggressive, rude, distasteful to 99% of the users here, and shows a zealous defense of Intel... It’s probably Spud.
January 28, 2004 2:51:32 PM

Quote:
Truth be known though buddy at Aces was right huge discounted Itanium servers are floating around. It's tough to move over to a whole new environment so they need to entice users to move over. The Opteron will need the same procedure on the level of recoding\ compiling\ debugging

No, Opteron won't need the same recoding/compiling/debugging procedure. Opteron doesn't need everything recompiled at once. A lot of existing apps don't need 64-bit and actually run faster when left as 32-bit apps.

That's the part where IA64 falls down. It requires everything to be ported at once for optimal performance, even stuff that otherwise would run faster when left as 32-bit code.

Sure, there's a "decent" IA32 emulation layer, but how "decent" is it? Intel says it's on par with a 1.6GHz part. 1.6GHz what? 1.6GHz Banias? Cool. 1.6GHz Willamette? Yechh. :eek: 

<i><Lionel Hutz> I'll be defending...The SCO Group!!!??? Even if I lose, I'll be famous!</i><P ID="edit"><FONT SIZE=-1><EM>Edited by kelledin on 01/28/04 10:52 AM.</EM></FONT></P>
Anonymous
a b à CPUs
January 28, 2004 3:00:26 PM

> I can safely say no since these guys do write their own
>code and regardless of the compiler dependencies the
>Itanium has it still has the highest IPC of any silicon
>currently available to the market.

it wouldnt be safe to say this, because on certain types of code, especially brach sensitive code, Opteron has a healthy performance lead -regardless of IPC which matters to no one, performance/$$$ is what counts in HPC, and to a lesser extent, performance/watt since watt==$$$. opteron is (more than) competitve in both, unless pure raw FP is needed, then Itanium is king.

>ast I checked and enterprise 4-8 way systems are classified
>as HPC.

Hu ????? Check again please.. HPC is "high performance computing", simplified, you could say this is the supercomputer Linux cluster market for calculating things like weather predicitions, oil drilling, mathematical & physical simulations, etc, etc, Those apps rarely (or never) require 8 way single image servers, and the systems to run this kind of software are generally built with cheap 1 or 2 CPU linux systems. Redundancy is king, as RAS doesnt matter a whole lot (if a single node goes down, so be it, no one cares).

OTOH, 4-8 and up way enterprise servers is generally for transactional processing, think ERP (SAP, BAAN, Siebel, ...) or datamining/businesess intelligence (Oracle, SQL server, Sybase, Syntel, ...). Requires ONE machine with enough I/O and CPU power to run a single imag of this business app. This is a *completely* differnt market. CPU (fpu/int) performance is generally far less important than RAS (Reliability, Availability, Serviceability), I/O power, and things like compatibility with existing software base, support, etc... Those are the markets Itanium would love to enter, but so far hasnt really. Its Power, PA Risc, SPARC and Alpha territory. To a lesser (but increasingly important) extent, x86 territory.

Frankly, I think you may need a 101 initiation crash course on the server market before trying to argue with my P'sOV

= The views stated herein are my personal views, and not necessarily the views of my wife. =
January 28, 2004 3:21:05 PM

Quote:
No, Opteron won't need the same recoding/compiling/debugging procedure.

When the software moves to the 64bit aspect it sure will though and thats what I am trying to say. Sorry for not being more clear for the initial get go.
Quote:
it wouldnt be safe to say this, because on certain types of code, especially brach sensitive code, Opteron has a healthy performance lead

Could you please show a link that shows the Itanium loseing heavey branched programs please?
Quote:
regardless of IPC which matters to no one

Sorry I am a bit confused by that statement as IPC has been you biggest weapon in the Intel Vs. AMD debate as of since the K7's arrival.
Quote:
Hu ????? Check again please.. HPC is "high performance computing", simplified, you could say this is the supercomputer Linux cluster market for calculating things like weather predicitions, oil drilling, mathematical & physical simulations, etc, etc

Um I think thats what I was trying to get at, is that it doesnt need to be 4-8+ to be considered HPC. Which is my fault since I most likely didnt correctly convey my thoughts well last night before I headed to work.
Quote:
Stick a fork in Itanium, its done. Even HP is embracing Opteron, and while certainly it will be restricted it to 1-4 way initially (hence, not competing with its higher end >4 way Itanium enterprise servers), this is going to kill any momentum Itanium may have enjoyed. Can't wait for HP to announce its AMD64 port of HP-UX :) 

But the initial first post gave me the impression that you were hinting that the HP adoption of the Opteron wasnt HPC till it hit the 4+ area noted to be in control by the Itanium. But thats how I interpreted it which may or may not have been correct.
Quote:
if a single node goes down, so be it, no one cares

Gah have you ever witnessed the event of a node going down and the reprocessing to rebalance the load that particular node was working on...
Quote:
Frankly, I think you may need a 101 initiation crash course on the server market before trying to argue with my P'sOV

Your not giveing me enough credit :tongue: .

-taitertot

If this post has attitude, seems to be overly aggressive, rude, distasteful to 99% of the users here, and shows a zealous defense of Intel... It’s probably Spud.
Anonymous
a b à CPUs
January 28, 2004 4:48:18 PM

>When the software moves to the 64bit aspect it sure will
>though and thats what I am trying to say

Don't be too sure if you don't have a clue of what you are talking about. If you believe porting x86 code to AMD64 is anywhere near as complicated/expensive/time consuming as porting from ANY isa to IA64, then I don't think we have anything left to discuss.

>Could you please show a link that shows the Itanium loseing
>heavey branched programs please?

Do your own research.. i don't feel obliged to provide 'proof' of any statement I make, especially not when it concerns well know facts. If you don't take my word for it, just mail Vincent Diepenveen (author of Diep chess program). Chess is probabl the most branch sensitive code you could imagine, so just ask him how happy he is with the performance of his code on his Altix Itanium supercomputer. Ask him how it compares to athlon or opteron. If you can't use google, try locating him on www.realworldtech.com 's forums.

Alternatively, try SPECint2k (www.spec.org) and compare 197.parser or 254.gap performance between IA64 and K8.

>Sorry I am a bit confused by that statement as IPC has been
>you biggest weapon in the Intel Vs. AMD debate as of since
>the K7's arrival.

If you are really *that* clueless, don't expect me to enlighten you, I have better things to do.. and if you arent, then why ask ?

>Um I think thats what I was trying to get at, is that it
>doesnt need to be 4-8+ to be considered HPC.

Reread my posts, I think you still didnt get it.

>But the initial first post gave me the impression that you
>were hinting that the HP adoption of the Opteron wasnt HPC
>till it hit the 4+ area noted to be in control by the
>Itanium. But thats how I interpreted it which may or may
>not have been correct.

Entirely incorrect. Reread please. I just gave HPC as an example of a market where Itanium could over time be viable in spite of the AMD64 and/or Xeon64 threat. One of the very few examples.

>Gah have you ever witnessed the event of a node going down

Yes. The admin got an email the node was down.. he would typically not read it until the next day, then have a look, order spares if required, and decide to resolve the issue early next week, along with 3 other nodes that had been down since.

Ever witnessed a mission critical server eg doing production planning not even going down, but generating a warning ? It would typically page and wake the tech support on duty and get him there within 20 minutes, in the unlikely case no one would be on site in the first place. If such a machine would ever go *down*, lights turn red, sirens howl and Defcon 5 is put into action. Not to mention the IT manager is very likely to lose his job if he doesnt come up with a VERY good explanation before the board of directors.

Just a way of speaking, both are really, utterly incomparable.

>Your not giveing me enough credit

Im probably giving you both too much credit and too much of my time.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
January 28, 2004 5:06:40 PM

Quote:
Don't be too sure if you don't have a clue of what you are talking about. If you believe porting x86 code to AMD64 is anywhere near as complicated/expensive/time consuming as porting from ANY isa to IA64, then I don't think we have anything left to discuss.

Not saying that either all I am saying is there will be a need to recompile for native 64bit mode. I also didn’t say anything regarding the price or time issues inherit to porting for IA64.
Quote:
Do your own research.. i don't feel obliged to provide 'proof' of any statement I make

Bugger!!!
Quote:
Vincent Diepenveen (author of Diep chess program). Chess is probabl the most branch sensitive code you could imagine, so just ask him how happy he is with the performance of his code on his Altix Itanium supercomputer. Ask him how it compares to athlon or opteron. If you can't use google, try locating him on www.realworldtech.com 's forums.

Last I checked Vincent chess program was very heavy Integer as well as overly based on complex branches based and he was complaining about existing compilers for IA64. But ill be sure to get a hold of him since I remember reading this in Aces hardware forum and be damned if I can find the page... But ill be sure to check because I don’t want to go thinking something that isn’t correct.
Quote:
If you are really *that* clueless, don't expect me to enlighten you, I have better things to do.. and if you arent, then why ask ?

Still haven’t clarified for me, and calling me clueless won’t help the debate.
Quote:
I just gave HPC as an example of a market where Itanium could over time be viable in spite of the AMD64 and/or Xeon64 threat.

Finally we cleared that one up :smile: .
Quote:
Ever witnessed a mission critical server eg doing production planning not even going down, but generating a warning ? It would typically page and wake the tech support on duty and get him there within 20 minutes, in the unlikely case no one would be on site in the first place. If such a machine would ever go *down*, lights turn red, sirens howl and Defcon 5 is put into action. Not to mention the IT manager is very likely to lose his job if he doesnt come up with a VERY good explanation before the board of directors.

Very suave that made me smile, and people tell me you’re a smacktard.
Quote:
Im probably giving you both too much credit and too much of my time.

Just when I thought we were getting close, you have to push me away :wink: .

-taitertot

If this post has attitude, seems to be overly aggressive, rude, distasteful to 99% of the users here, and shows a zealous defense of Intel... It’s probably Spud.
Anonymous
a b à CPUs
January 28, 2004 5:20:55 PM

>Not saying that either all I am saying is there will be a
>need to recompile for native 64bit mode.

In many cases, just a very small portion of the x86 code may need (or rather, benefit from) recompilation to access > ~3GB. All the rest can remain 32 bit pretty much forever,and run full speed on Opteron style cpu's. Porting just one module, or an entire implementation is quite some difference regardless of any architectural resemblance between x86 and AMD64 (and unlike IA64)

>Last I checked Vincent chess program was very heavy Integer
>as well as overly based on complex branches

Exactly.

>based and he was complaining about existing compilers for
>IA64

.. and their performance. Since you seem to know him, and/or his work, it shouldnt be difficult to find more info. Do a search on RWT forums (aces still doesnt have a search :(  )

>Still haven’t clarified for me, and calling me clueless
>won’t help the debate.

Are you serious ?? *sigh*. Okay, performance = IPC x clock. What more is there to be said ? Who cares Itanium achieves its performance performnce levels at 500 Mhz or 5 GHz ? Hence, who cares about IPC ?

>Very suave that made me smile, and people tell me you’re a
>smacktard.

I'm not easily insulted, especially not by 'people on THG'. That being said, I have no idea what a smacktard is, but I doubt its a compliment :) 

= The views stated herein are my personal views, and not necessarily the views of my wife. =
January 28, 2004 5:28:11 PM

Quote:
Exactly.

Holy crap we have made first contact :smile: .
Quote:
Since you seem to know him, and/or his work, it shouldnt be difficult to find more info.

*spud picks nose* yuppers I know things.
Quote:
Hence, who cares about IPC ?

Oh I knew the performance equation and whatnot, but I see what you mean now *pats beany on the back* did good son did good.
Quote:
I'm not easily insulted, especially not by 'people on THG'. That being said, I have no idea what a smacktard is, but I doubt its a compliment :) 

Oh I hope you didnt take that as a insult because it wasnt meant to be taken that way. It was a small jest to kinda break the "I hate you, I want to spoon yer heart out from yer ass" deal you and me got going on. But I must admit you arent as bad as I initially thought you were... smacktard :smile: ...

-taitertot

If this post has attitude, seems to be overly aggressive, rude, distasteful to 99% of the users here, and shows a zealous defense of Intel... It’s probably Spud.
January 28, 2004 7:24:47 PM

Those high profil software will need recompiling and porting for Itanium and opteron.The only advantage opteron got is many software cost less for IA32 version SLES cost much less for a Xeon that a Itanium even windows 2003.

I dont like french test
January 28, 2004 7:34:18 PM

IA64 ELL is fast work and ready for download.And as you say is not everything that need 64 bit.

I dont like french test
January 28, 2004 7:43:16 PM

Do your own research.. i don't feel obliged to provide 'proof' of any statement I make, especially not when it concerns well know facts. If you don't take my word for it, just mail Vincent Diepenveen (author of Diep chess program). Chess is probabl the most branch sensitive code you could imagine, so just ask him how happy he is with the performance of his code on his Altix Itanium supercomputer. Ask him how it compares to athlon or opteron. If you can't use google, try locating him on www.realworldtech.com 's forums.

By the way is software use pure ALU and he dont have try many option and he have disbale 1\2 of the cluster.
Many have disagree with him also

I dont like french test
January 28, 2004 9:09:47 PM

Quote:
Are you serious ?? *sigh*. Okay, performance = IPC x clock. What more is there to be said ? Who cares Itanium achieves its performance performnce levels at 500 Mhz or 5 GHz ? Hence, who cares about IPC ?


Well then, who cares about clock?

:tongue:

We both know that high clockspeeds don't automatically mean high performance: AMD has proved that. Now you want to throw away IPC as if it means nothing... huh? That's what Taiter was getting at. If Itanium has an extremely high IPC, then the only thing holding it back would be clockspeed... kind of the same situation AMD has with their chips. Athlons have a much better balance between the two than Pentium 4s, so despite AMDs rather large clock defecit they still do as well as the P4s (or better in some cases).

So first you say clockspeed means nothing, now you say IPC means nothing. So if they're both meaningless, why aren't we still using 286s? :tongue:

<font color=red> If you design software that is fool-proof, only a fool will want to use it. </font color=red>
Anonymous
a b à CPUs
January 28, 2004 9:28:36 PM

So first you say clockspeed means nothing, now you say IPC means nothing. So if they're both meaningless, why aren't we still using 286s?

Because while clock doesnt matter (why else would anyone want a ALpha ?), and IPC doesnt matter (why buy a P4 ?), performance <i>does</i> matter. Anyway, I'm sure any halfway intelligent being understands whats being meant, especially in its context.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
Anonymous
a b à CPUs
January 28, 2004 11:33:19 PM

>By the way is software use pure ALU

Exacly.. I was just giving an example of types of code Itanium typically does not perform very well on, that would definately be integer code, and (hard-to-predict) branch sensitive code. Chess being an obvious example.

>And he dont have try many option

Oh, if you have some suggestions for him, I'm sure he'll gladly listen. If I'm not mistaken, his program ended 5th or 6th in the chess worldchampionship, so if you have a miracle compiler option that will increase his throughput, I'm sure he is all ears.

>and he have disbale 1\2 of the cluster.

Which was not really a testimonial to the (IA64) setup he was using. BTW, its NOT a cluster, its a SGI Altix server unless i'm terribly mistaken. One of the largest single image Itanium servers on the market. He uses it because its the machine he has access to through his university, but I'm sure you'll agree he is anything but happy with it.

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