Sign in with
Sign up | Sign in
Your question

Opteron tie or even small lead vs Itanium2 1GHZ

Last response: in CPUs
Share
January 16, 2003 9:58:11 PM

http://biz.yahoo.com/djus/030116/0259000197_1.html

change in the roadmap for IA-64 2 version of madison and a big delay on Montecino but move to a full dual core GO transistor count 1 billion?? SPEC FP 5000 or more


http://www.heise.de/newsticker/data/as-16.01.03-002/
On SAP benchmark Quad opteron as good score beating
Include Itanium2 RX 5700 it just found weird they dont have take RX6000 and not oracle wich is much more optimaze for IA-64 also under oracle they score 600 wich beat Opteron by a small margin even so a itanium have loss compare to Opteron a this battle let see what will happen on release<P ID="edit"><FONT SIZE=-1><EM>Edited by juin on 01/16/03 07:07 PM.</EM></FONT></P>
Related resources
Can't find your answer ? Ask !
January 17, 2003 1:14:28 AM

So perhaps AMD's 1.4GHZ Opterons aren't SUCH a bad idea...?

--
This post is brought to you by Eden, on a Via Eden, in the garden of Eden. :smile:
January 17, 2003 1:42:46 AM

yes i hae also to agreed myself but may be that just 1 good benchmark for AMD like Itanium 1 in chemical benchmark as etremely fast itanium 2 leave everyone in the dust.I guess intel will have to increase INT for future cpu.It still wait to see a opeteron 1.6 1.4 under heavy FPU benchmark

Now what to do??
Anonymous
a b à CPUs
January 17, 2003 6:05:50 AM

I think its quite amazing AMD can outperform the new bright and shiny IA64 Itanium2 using a patched 20 year old "extint" instruction set, with a cpu 1/3 the size, with 1/3 the ammount of L2 cache, probably requiring +-1/2 the powerconsumption, using infinitaly cheaper chipsets and motherboards.. and remain fully backwards compatible with all existing software.

Imagine Opteron would be produced by Intel, and it would get 3 or 9 Mb L2 cache and the intel compiler support... No one would miss IA64...

= The views stated herein are my personal views, and not necessarily the views of my wife. =
Anonymous
a b à CPUs
January 17, 2003 6:11:56 AM

>yes i hae also to agreed myself but may be that just 1
>good benchmark for AMD like Itanium 1 in chemical
>benchmark as etremely fast itanium 2 leave everyone in
>the dust

Itaniums FPU lead is pretty much undisputed. However, FPU performance isnt as important in these markets. Webserving, databases, ERP, CRM.. all these are purely integer applications where it appears the Opteron really shines. When you need a system to compute all sorts of simulations, IA64 might make more sense..

= The views stated herein are my personal views, and not necessarily the views of my wife. =
January 17, 2003 7:22:43 AM

<A HREF="http://www.intel.com/products/server/processors/server/..." target="_new">Itanium2</A> has 256k L2 cache. The huge 3mb cache is an L3. Stats I could fine on the Itainum2 cache are...cache lines are 64byte L1 128byte L2/L3. Cache latency is about 6 L2 and 12 L2. Cache associations are 8 L2 12 L3.

Best info I could find on the <A HREF="http://www.intel.com/design/itanium2/download/isscc_200..." target="_new">3mb L3 Cache</A>

Dichromatic for your viewing plesure...
January 17, 2003 8:33:42 AM

It's weird, x86 processors have been able to score ridiculously better than all of the big-bin RISC based processors. A P4 3.06 is able to out match the top of the line Power4 in SpecInt. I wonder why this is anyway.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.
January 17, 2003 10:51:08 AM

A paradoxial case hmmm?

--
This post is brought to you by Eden, on a Via Eden, in the garden of Eden. :smile:
January 17, 2003 11:54:14 AM

Quote:
I think its quite amazing AMD can outperform the new bright and shiny IA64 Itanium2 using a patched 20 year old "extint" instruction set, with a cpu 1/3 the size, with 1/3 the ammount of L2 cache, probably requiring +-1/2 the powerconsumption, using infinitaly cheaper chipsets and motherboards.. and remain fully backwards compatible with all existing software.

Yeah, and it only took clocking their processors 600MHz higher to do it.

Not that I'm backing up Intel, since I've never been an Itanium fan in the first place. It just strikes me rather funny that all of the AMD fanboys yell out how an Athlon has more IPC than P4, but you sure don't hear a peep from them about Itanium having more IPC than SledgeHammer. Heh heh.

Anywho, the true point of this is that these benchmarks are completely pointless in the first place. Four-way Opteron systems aren't even for sale yet, and by the time they are the Itaniums will have probably made it to their next revision, Madison, with it's 1.5GHz clock and 6MB of L3. And the Opterons will probably be at a 2GHz clock instead of 1.6GHz. So then we'll have to do the benchmarking all over again. None of these benchmarks matter. It's just a paper war.


PC Repair-Vol 1:Getting To Know Your PC.
PC Repair-Vol 2:Troubleshooting Your PC.
PC Repair-Vol 3:Having Trouble Troubleshooting Your PC?
PC Repair-Vol 4:Having Trouble Shooting Your PC?
Anonymous
a b à CPUs
January 17, 2003 1:27:20 PM

>Yeah, and it only took clocking their processors 600MHz
>higher to do it.

So ? You could also say it beat the [-peep-] out of Xeons running 400 Mhz higher frequency, who cares ?

>It just strikes me rather funny that all of the AMD
>fanboys yell out how an Athlon has more IPC than P4, but
>you sure don't hear a peep from them about Itanium having
>more IPC than SledgeHammer. Heh heh.

"He heh..".. big deal, who cares about IPC only ? IF you have a cpu 3x as big, 2x as power consuming, apparently performing on par on INT code, it could run at 200 Mhz or 5 GHz.. what do I care ?

>Anywho, the true point of this is that these benchmarks
>are completely pointless in the first place. Four-way
>Opteron systems aren't even for sale yet, and by the time
>they are the Itaniums will have probably made it to their
>next revision, Madison, with it's 1.5GHz clock and 6MB of
>L3.

That doesnt make them pointless IMHO. These are the first Opteron benches I've seen, and they look *very* promising. Even the sheer fact that there are 4 way systems up and running is quite an achievement. How long did it take the P4 to go 4-way ? How long for the Itanium 1 ? How long for Athlon to go 2 way ? And here you have a system of a non released cpu, with a pretty revolutionary design (integrated mem controller, Hypertransport topology,..) apparently hapily running in a quad config. Not bad if you ask me.

As for the performance, Opteron is slated to debut end of Q1, Madison is still scheduled for Q3 AFAIK. So when Opteron comes out, it will still be competing with McKinley.. although we all know Opterons natural competitor is not Itanium, but Xeon. And it looks like a 4 way 1.6 Opteron system outperforms a 8 way big cache Xeon 2 Ghz system on a real world, industry leading benchmark. Not bad at all.. factor in a possible 2 Ghz at launch, and performance wise, Opteron looks like a Xeon killer. A near 100% performance advantage, I dont see how Intel is going to close that gap. Latest roadmaps I've seen suggest a 2.5 Ghz Xeon by the end of the year.. thats 25% from where they are now. You may find this nothing but speculation and useless, paper wars or whatever, but I find this very promising for AMD ...

= The views stated herein are my personal views, and not necessarily the views of my wife. =
January 17, 2003 3:01:19 PM

Quote:
So ? You could also say it beat the [-peep-] out of Xeons running 400 Mhz higher frequency, who cares ?

Anyone trying to run 64-bit code on a Xeon would, I'd imagine.

Quote:
"He heh..".. big deal, who cares about IPC only ? IF you have a cpu 3x as big, 2x as power consuming, apparently performing on par on INT code, it could run at 200 Mhz or 5 GHz.. what do I care ?

I was just saying that in light of AMD's recent 'AMD Me' marketting scheme which is so concerned with IPC, and considering how many AMD lemmings print out pages from it to stick in their Bible, it's just funny to see that the only way that AMD can beat an Itanium is with a faster clock speed on a chip with a lower IPC. If you can't see the humorous irony in that, then you must be blind.

Sue me for having a sense of humor.

Quote:
That doesnt make them pointless IMHO. These are the first Opteron benches I've seen, and they look *very* promising.

They're just as pointless as all of the other Claw and Sledge benches that've been out there. Until final product meets final product, it's all just spotty benchmarking to fuel fanboy catfights. If you haven't seen any other benchmarks before this one, well, you must not have been looking very hard.

Further, as I understand it, this benchmark is driven almost solely by the ALU and the memory bandwidth. These are hardly indicative of complete performance. So even just as a benchmark itself, it is sorely lacking for a complete comparison.

Hence, this benchmark is pointless. I'm not saying that the Itanium is better and these benchmarks are faked. I'm saying that the benchmark is horribly insubstantial and inconclusive of released products and to draw any real-world conclusions from it is pointless. All that can be said from it is that yes, a 4-way Opteron platform looks pretty good for running a database. We already knew (or at the very least assumed) that much.

Quote:
Even the sheer fact that there are 4 way systems up and running is quite an achievement. How long did it take the P4 to go 4-way ? How long for the Itanium 1 ? How long for Athlon to go 2 way ? And here you have a system of a non released cpu, with a pretty revolutionary design (integrated mem controller, Hypertransport topology,..) apparently hapily running in a quad config. Not bad if you ask me.

To have it do any less would have been disappointing if you ask me, since that was the entire <i>point</i> of the Hypertransport technology in the first place.

If AMD <i>couldn't</i> have had a 4-way Opteron running right from the beginning, it would have shown that Hypertransport was a failure or that AMD couldn't actually design what they set out to design. It's cool that it's working, but all that it really shows to me is that AMD did exactly what they said they were trying to do, no more, no less. Which is cool, but certainly nothing to sing and dance about since we'd have all been disappointed had they done anything less.

Had they snuck it in there or originally intended 1 and 2 way systems only, it would have been something to be impressed by that they jumped the R&D gun to 4-way. But since it was something that they had been aiming at from the very beginning of SledgeHammer, it makes the fact that they finally achieved it nothing more than what we would have expected from them.

Now, of all of those other chips, how many of them started out with the intention of going into 4-way systems? (Or 2-way in the case of the Athlon.) How long did it take any of them to achieve the original goals that were set out for them? That's all that 4-way Opteron did. Granted, AMD had some impressive original goals, but that's still all that they did was meet them.

And actually, there still hasn't even been proof that AMD has met their original speed goals yet. This close to their deadline not having finalized yet is pretty bad, considering that they still will take about another 3 months after that to produce the quantity needed for the initial launch. (That is, unless AMD does yet another paper launch.)

So I'd wager that comparing when AMD will <i>actually</i> be selling these buggers on the retail market to when Intel will be selling Madisons on the retail market puts them within 3 months of each other easily.


PC Repair-Vol 1:Getting To Know Your PC.
PC Repair-Vol 2:Troubleshooting Your PC.
PC Repair-Vol 3:Having Trouble Troubleshooting Your PC?
PC Repair-Vol 4:Having Trouble Shooting Your PC?
January 17, 2003 4:35:40 PM

itanium 2 L2 cache lantency is 5

Now what to do??
January 17, 2003 4:36:43 PM

i think madison will be the fastest on int for a short time

Now what to do??
January 17, 2003 4:39:33 PM

There score is real and will be the same at release opteron will be release at 1.4 1.6 not 2.0 the 2.0 ghz mark is for the A64 AMD like to mislead newbie with K8 opteron hammer A64 they tend to mix them up.

Now what to do??
January 17, 2003 4:48:09 PM

Cache latency is determined by the speed of the processor. A 1.0ghz Itanium2 would have a cache latency of 5, while a 1.2ghz Itanium2 would have a cache latency of 6. Assuming internal busses run at 200mhz. If you notice I said about 6.

Dichromatic for your viewing plesure...
January 17, 2003 4:48:49 PM

Madison it claim to be release on Q2 on E 8870 so dell and others will be able to built systemes around IA-64 all itanium are backward compatible socket.

Xeon 2.5 will be release in Q2 maybe a a faster FSB 533 or even 667/800

Once again opteron will not run at 2.0 before a very long shot and be release on Q2 have a mass production only at the end of Q2

Now what to do??
January 17, 2003 4:53:06 PM

Then again there are no 1.2ghz Itanium2. So you are correct sir.

Dichromatic for your viewing plesure...
Anonymous
a b à CPUs
January 17, 2003 8:24:23 PM

> was just saying that in light of AMD's recent 'AMD Me'
> marketting scheme which is so concerned with IPC

I dont think AMD ever even mention IPC in their campaign. They mention "performance". And performance <> MHz.

>Sue me for having a sense of humor

I'd rather be blind, than have your "sense of humor".

>Further, as I understand it, this benchmark is driven
>almost solely by the ALU and the memory bandwidth. These
>are hardly indicative of complete performance. So even
>just as a benchmark itself, it is sorely lacking for a
>complete comparison.

It is perfectly indicative of SAP performance. Its not some obscure handtuned synthetic benchmark, this benches the performance of an ERP suite used by THOUSANDS of large corporations worldwide. It is *THE* ERP application, and as such at the heart of more businesses as you could imagine. And its exactly the sort of mission critical, power hungry application that warrants a serious investment in >4 way systems. Honestly, I dont think you could come up with a more telling, important bench than this one. Give me one.. ? So go ahead, and ridicule it, wether it is ALU or memory bandwith restrained doesnt interest companies one bit. How well it runs SAP however, DOES matter. Companies tend to rely more on their ERP systems for their productivity, than on SPEC you know. But whatever, next time we see an Oracle, DB2, or Siebel benchmark, Im sure you'll say the same.. I dont care. Most CTOs have more sense than you..


= The views stated herein are my personal views, and not necessarily the views of my wife. =
January 17, 2003 10:23:59 PM

slvr_phoenix, you have a point there with your IPC story.

I guess an argument is only a good argument as long as it doesn't work against yourself... but then you can always try to explain that your argument was wrong in the first place. Ah, I like marketing :) 
January 18, 2003 1:59:49 AM

You are totaly right about SAP benchmark that the biggest win for opteron up to now not there little fake benchmark again a P4 2.0 with a brookdale chipset saying they win.SAP it a huge software i am now waiting for TPC-P benchmark.Seen that IBM will back Opteron good thing so it leave to a conclution that some compétitor of IBM may be push on IA-64 or SUN for platfrom the last becoming really really slow not faster that a K7 thunderbird 1.4

Now what to do??
January 18, 2003 2:05:56 AM

I dont think AMD ever even mention IPC in their campaign. They mention "performance". And performance <> MHz

That relate to desktop base marketing AMD do ZERO marketing for opteron they help OEM to advertise the whole systemes.HP/NEC/SGI dont say itanium 2 is better, they say RX6000 or ALtix 3000 or nec 32 way is the fastest the box it mean the whole package OS chipset CPU RAM SOftware compiler use driver tech support service managing and the TOTAL cost is important not they given CPU or chipset only the Box is important.HP use Linux HP unix oracle
Nec/.net SGI/linux all have in house chipset boards PCI-X controleur.

Now what to do??
January 18, 2003 8:14:34 PM

I getting better now that my girlfriend is english and thank you

Now what to do??
January 18, 2003 9:04:50 PM

Lol I see...

I thought that if you had been here for over year you'd have improved, but I dunno if your daily life in whatever region in Québec has this much english people...

Here in Hull it's quite filled, our french school is slowly being populated my people who speak english lol.

--
This post is brought to you by Eden, on a Via Eden, in the garden of Eden. :smile:
!