Where are the 2GHZ opteron

Toggle sidebar Toggle sidebar
J

juin

Distinguished
May 19, 2001
3,323
0
20,780
http://www.theinquirer.net/?article=10254

Allready issue to bring higher clock speed.I dont want to troll but rumor of low clock speed on Opteron and Clawhammer are getting true.As today 2.8GHz Xeon MP is release and major price cut on most lower model.Opteron for 800 series will cost a wopping 2500$ us buck that in itanium range that cost the same price that a Itanium 1.3GHZ 3MB there were allready ((the opteron))slower that Itanium 1GHZ even with 200 extra mghz that wont make a difference.vs mr Xeon Dp thing were pretty close and shadow of Prescott are showing.At best this will end up like Athlon MP good start that finish quick

[-peep-] french
 
<A HREF="http://news.com.com/2100-1001-962312.html" target="_new">Right here</A>

<b>update</b>: Plus where do you get $2500? The Opteron 844 is merely $2149 and likely to be less on the street.

Intel giveth and Microsoft taketh away.<P ID="edit"><FONT SIZE=-1><EM>Edited by dwellman on 06/30/03 01:29 PM.</EM></FONT></P>
 
So the Otperon is priced at around $2150? I highly doubt that a 1.8Ghz Opteron can keep pace with a 1.4Ghz Itanium (which is in the same price range, around $2200).

Besides, those numbers for the Opteron 2.0Ghz (1200/1150) are not nearly enough to compete with Itanium. The state-of-the-art 1.5Ghz can churn out (1300/2100) int/fp score. Assuming that the 1.4Ghz is indeed only 1/15 slower, then it would score around 1200 and 1950 or something...

Plus, Itanium actually doesn`t have limitations as to number of processors used. (this might be a disadvantage, considering that there is no cheaper, 1CPU-only Itanium). Today's launch has put Itanium that much closer to competing adequately with Opteron...
 
The industry buzz circa February through April of this year and benchmarks suggest that Opteron matches Itanium 2 better than you suggest. Additionally, I hear Itamnium has some trouble with 32 bit code

References:
<A HREF="http://news.com.com/2100-1006-997788.html" target="_new">http://news.com.com/2100-1006-997788.html</A>
<A HREF="http://www.extremetech.com/article2/0,3973,635775,00.asp" target="_new">http://www.extremetech.com/article2/0,3973,635775,00.asp</A>
<A HREF="http://www.anandtech.com/cpu/showdoc.html?i=1815" target="_new">http://www.anandtech.com/cpu/showdoc.html?i=1815</A>
<A HREF="http://www.anandtech.com/cpu/showdoc.html?i=1816" target="_new">http://www.anandtech.com/cpu/showdoc.html?i=1816</A>
<A HREF="http://www.anandtech.com/it/showdoc.html?i=1817" target="_new">http://www.anandtech.com/it/showdoc.html?i=1817</A>
<A HREF="http://www.anandtech.com/cpu/showdoc.html?i=1818" target="_new">http://www.anandtech.com/cpu/showdoc.html?i=1818</A>
<A HREF="http://www.alineos.com/Benchs/bench3_eng.html" target="_new">http://www.alineos.com/Benchs/bench3_eng.html</A>

Intel giveth and Microsoft taketh away.
 
First of all, Opteron is meant for Integer workstations mainly. It is A MUCH better suited server CPU than Itanium.
Opteron has HyperTransport, it is likely to be stronger if linked to 7 other Opterons, while Itanium may not scale as much. In the end, the Opteron's value goes up, Integer as well.
I highly doubt Itanium will win in the Int sector and get bought more.

For FPU, I can only encourage Intel.

--
If I could see the Matrix, I'd tell you I am only seeing 0s inside your head! :tongue:
 
The so call myth of Opteron super bandwith is a real joke most itanium 2 have access to a quad channel DDR 2100 opteron only 2 channel only 2 can be use by CPU on itanium but I/O can suck you a nice 1 to 3 GB/S consider a big bunch of netwook CPU and a buch of Seagate 15 RPM or solid state disk.Itanium lantency dont rise as cpu is add exception to cluster interconnecter.AMD can also this disaventage but also more cpu give even more remote access in the systemes put a lot of pressure on a single link leaving only a mere 800 mghz bit link.

TPC result all top 5 are Itanium from HP superdome or Nec.On 32 way or 64 way and the best price performace of the Top 20.Strangely all on Windows 2003 for IA-64.

[-peep-] french
 
I do not understand what are you trying to say with that.

I merely tried to prove that with HyperTransport's flexibility and benchmark facts, the Opteron scaled very well compared to past MP systems, in x-way platforms.

--
If I could see the Matrix, I'd tell you I am only seeing 0s inside your head! :tongue:
 
The myth that Opteron have bandwith to spare so false there memory controleur is not that fast Lantency figure form Aceshardware benchmark show that Opteron Ram latency is about 10% less that Placer chipset.What we need is a point to point chipset to CPU connecteur personalie Redwood is the solution 6.4 GHZ on a 16 bit bi-direction setting 4 CPU per chipset and a 64 bit 3Gio bi-direction at 2.5 GHZ bettween chipset and a I/O chip that controle entrance.

Low pin count high power cost maybe high.Direct connection to ram by the I/O limited remote access high bandwith a more stable lantency that help when you built a systemes as programeur can expect a given value.

[-peep-] french
 
<b>dwellman,</b> I understand what you're saying there. 1Ghz Itaniums are not really a good match for Opterons... but I said that <b>TODAY'S LAUNCH</b> has made a big difference. It's a completely redesigned core that can churn out (I'll just have to say it again) 1300 SPEC_INT and 2100 SPEC_FP. Its integer performance is <b>above the 1.8Ghz Opteron</b>, at 1200 SPEC_INT, and its <b>floating point performance is almost twice</b> the one of the 1.8Ghz Opteron at 1150 SPEC_FP. I don't mean the old McKinley at 1Ghz (btw, even the 1.0Ghz McKinley beats Opteron handily in floating point), I mean the new Madison at 1.5Ghz... which was also introduced in considerably cheaper 1.4 and 1.3Ghz versions. Got it?

So, reever2, Itanium not only competes in SPEC, but is devastating in SPEC <b>and makes short work of Opteron.</b>

And about 32-bit support: Why would that actually be THAT interesting in the server segment? I don't see that as much of an advantage for my possible uses of Itanium, given that large servers/high performance solutions mostly use code written and compiled specifically and open software. And I'm open to learning how to use an excellently-designed CPU in order to take advantage of it. That's the spirit behind everyone wanting to learn about Opteron, isn't it?
 
It is A MUCH better suited server CPU than Itanium.
Click to expand...
Eden, looks to me as if you're unhappy about Itanium's progress for no reason... A generation change has occured (madison is 3rd) and you're not pleased to see a +50% increase in processor performance? (not exclusively here... you said that in other threads) The damned thingie beats Opteron! It has set new SPEC records! What more do you want? What more do you need? I suppose you'd only be impressed if you saw an Itanium at 4Ghz and 1Ghz FSB that could score 5000/5000 in SPEC benchmarks? Loosen up! Itanium isn't as bad as you make it seem to be.

BTW, Itanium won even in the Int sector, at least in SPEC benchmarks. Who gets "bought more" is another issue. We'll just have to wait and see. The deal with Itanium is its price, of course.
 
Scaling under Spec benchmark

Madison 1.5GHZ 6MB

8P int 98.3 FPU 164
16 int 195 FPU 327
32 int 385 FPU 644
That pretty close to perfect scaling

Opteron 1.8GHZ

2P int 25 FPU 24.7
4P int 46.1 FPU 44.2
8P None as there no opteron that support this.

To what i see Madison have scale better and at a higher level and give more performance.

All score were take from SGI website on Numa flex mobo a HP box could as been much faster on INT.
If we double so a perfect scaling score from Opteron 4 way int score of 92.2 and FPU score 88.2.

Link

http://www.sgi.com/newsroom/press_releases/2003/june/altix_benchmarks.pdf

and aceshardware data base


[-peep-] french
 
Itanium @1.5 GHz may win as many SPEC benches as you want, that makes it only ideally suitable for an environment where the organization:
*) writes and compiles its own code
*) doesnt mind using intels compiler instead if GCC or others
*) doesnt mind learning a whole new ISA
*) does mind the huge costs associated with such a software migration
*) where they have plenty of cheap rackspace (Opteron wins just about any SPEC bench when you compare performance/rackspace)

Sure itanium has a market, and it those markets, it may be hard to beat. Opteron has its own market though, those organization where the above *doesnt* apply, and that want to run off the shelve (x86) software, and/or want to leverage existing investments in software/libraries, knowledge, or need dense processing power. Itanium can shine as much in SPEC as you want, it won't beat a Duron on x86 code. The cost of the hardware platforms or a few more or less CPU's pales in comparison to the investment most companies have in software and experience.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
 
No, I do see the impressive performance. I would never dare imagine how such a beast one day at 2GHZ, comparing to a 2GHZ AthlonXP would do.

What just didn't impress me (as you seem to have dragged my words from the other threads to here, so I shall express here) is the scaling. However I did state before that I understood why and that it had simply slipped my mind about the bandwidth scalability thing. So I'll stop here since I already explained and conceded.

To Juin: Thanks for the backup of facts. I am impressed by these near-perfect scalabilities. What I don't understand is, HOW? How and what platform does it run on to be near 100% scalable? Assuming HT was the apex of today's MP enthusiasm, and that it was much needed as MP had never been so impressive really (you barely double performance even if you add another same clocked CPU), this surprises me.
It's almost like saying for real that getting twice the CPUs means twice the performance, something no one would dare say for Dual processing systems. (except if you're a dumbass)

--
If I could see the Matrix, I'd tell you I am only seeing 0s inside your head! :tongue:
 
Eden: SPEC scaling =! real world scaling.
In SPEC you can achieve near 100% scaling by having each CPU its own thread, and they do not fight for cache or memory much. This is a far cry from most real world apps, don't be fooled by SPEC scores too much. They are interesting, but not more than that. No one buys a server to run SPEC all day :)

= The views stated herein are my personal views, and not necessarily the views of my wife. =
 
as you seem to have dragged my words from the other threads to here
Click to expand...
Sorry about that, I just wrote in haste, actually. And actually, I didn't care to explain myself as well as Juin... sorry 'bout that. :smile: I just took a look at all posts around and expressed my thoughts and didn't explain why I was thinking the way I was thinking...

Also, on this SPEC scalability benchmark, I think that it is reasonable to say that it represents an empirical maximum throughput. Of course there are very few applications that will be written exactly that way, but the platform is very scalable and doesn't suffer from any hardware scaling bottlenecks - and that can be assured by the benchmark, I believe. That's how I would interpret those results anyway.
 
Heheh, true.

In fact when one checks the SPEC ViewPerf results, we're tempted to think it performs THAT good in reality, but nope. Especially with how the AthlonXPs dominated a certain SPEC test when in reality P4s are much better now in graphics rendering.

--
If I could see the Matrix, I'd tell you I am only seeing 0s inside your head! :tongue:
 
OMG...

The confusion, and misinformation in these forums has reached a completely new level. Let me join in...

"Itanium is cool guys! Itanium 3 madison is completely redesigned, and it has a new compiler coming soon that will give it great 32 bit performance, plus it's a 64 bit cpu, so it's twice as fast as a 32 bit cpu!" :tongue:

If anyone's interested, Ars has some GREAT forums...

Anyways, I'm out. I just got a reminder of why I left these forums...

- - -
<font color=green>Ignorance is bliss...isn't it (especially if you're a fanboy)?

"... In the semiconductor industry, it's good to be paranoid ..." - [Andy Grove]</font color=green>
 
.... oh wait... I forgot to laugh

Please. Flaming me for silly reasons is not going to do anything. I'm not going to spend my energy arguing empty topics.

- - -
<font color=green>Ignorance is bliss...isn't it (especially if you're a fanboy)?

"... In the semiconductor industry, it's good to be paranoid ..." - [Andy Grove]</font color=green>
 
lol. a [-peep-] monkey is too good for these forums....

I'm with Dark...later..and enjoy your fanboy site...be sure you get your head stuck<b>all the way</b> up Tom's ass. I'm sure he likes the feeling anyways.
 
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom