IS THE OPTERON GONNA SUCK?

[AmdMELTDOWN]

my opterinion is that the opteron is gonna do poorly when it's finally available. that's just my opterinion.

what makes you so sure it will do well?

"<b>AMD/VIA!</b>...you are <i>still</i> the weakest link, good bye!"

 

[IIB]

Personally I think that the best feature of the Opteron system would be its scalling on MP systems.
all CPUs today in an MP configration are limted to the bus bandwith they share to get data from and to comunicate with the other processor/s to maintane cach coherncy and do Bus sniffs.
this is pretty much limting CPU scaling above 4 processors on a mobo (actully the preformacne benifit for 2 to 4 processor is also very limtied). this is part of the reason high end processors have enourmce amounts of costly Cache.

in an Opteron system the System bandwith actully rises with each processo added - each processor has its on HT link to memory and HT link to two other CPUs in the same system - this will elminate the shared bus burden in MP. a very high bandwith (and low latncy) per processor in an MP configration will enable the Hammer to scale much higher then current processors in MP config. and will allow to elemnate some of the need for extreme amountrs of on die cahce. delivering a very cheap very high preforming server/workstation system.


This post is best viewed with common sense enabled<P ID="edit"><FONT SIZE=-1><EM>Edited by iib on 04/29/02 01:35 AM.</EM></FONT></P>

 

[eden]

Wow a non-anti-AMD question AND post AND topic from Melty...ow maybe Opteron WILL suck!

No seriously, on top of ClawHammer's powerful improvements over K7, Opteron will have 1MB cache, versus Prescott's, will have Dual Channel DDR and much much more. As well as its 64-bit performance 15% better, the overall performance boost over CH is about 15%, which carries a total of 40% better per clock than any Palomino, and that is more than enough to compete against a P4 2.6GHZ 533MHZ FSB, if you took in a 1.8GHZ Hammer.

--
Thunderbirds in wintertime, Northwoods in summertime! :lol:

 

[lhgpoobaa]

possibly. alot of new stuff to implement...
and we can always rely on Via to 'Via 'things up for us .

not sure if i like the idea of the type of ram & cpu being linked either.

time will tell.

P.S.
Yes i think the name Opteron kinda sucks...
Octeron coulda been better, or something meaning "8" for 8th gen core instead of wierd latin.

Despite appearances im not Phsysic. I may need your system specifications to solve your problem!<P ID="edit"><FONT SIZE=-1><EM>Edited by lhgpoobaa on 04/28/02 09:43 PM.</EM></FONT></P>

 

[eden]

I wonder what we'll nickname it! Like Celery...
Opera maybe? Or dare I say...Oprah? :lol:

--
Thunderbirds in wintertime, Northwoods in summertime! :lol:

 

[lhgpoobaa]

oprah????
OH GOD

funny though

i personally prefer octeron.
matches the core generation, and fits better with whats gone before
duron, athlon, octeron.

do i get any other votes?

Despite appearances im not Phsysic. I may need your system specifications to solve your problem!

 

[castle]

It looks to me you are comparing Opteron with P4 2.6G, right? A server CPU with a desktop one? It is not appropriate. Moreover, even by the time CH (1.8G as you said) is released, frequency of P4 will not be only 2.6G.

 

[rcf84]

Actually the OPTERON has to face the Itanium 2 :-|

You GeForce Ti4xxx is faster then my R8500 but my R8500 is the king of Aniso baby

 

[74merc]

IS AMDMELTDOWN EVER GONNA FIX HIS STUCK CAPS LOCK KEY?!?!


<i>Better to be silent and thought an idiot than to speak and remove all doubt.</i>

 

[eden]

There's a chance it comes out on single CPU too, so if someone dared to use, it may. I am looking at overall performance not any structural and costs.

--
Thunderbirds in wintertime, Northwoods in summertime! :lol:

 

[Phelk]

~ There's a chance it comes out on single CPU too ~

If you review the design documentation in the AMD presentations it becomes pretty clear no one would design a single processor Opteron system. The design will be different to a standard PC mainboard like the ClawHammer will use.

Instead the CPU's will come in plug-in modules with the CPU and memory together on one board which more than likely will have two CPU's with 4 memory slots associated with each CPU.

When it comes down to it an Opteron based server will be constructed very similar to the <A HREF="http://www.sun.com/servers/entry/880/" target="_new"> Sun V880</A> where you can have 2, 4, 6 or 8 cpus. Both systems are designed for 64bit NUMA.

<font color=blue> Smoke me a Chip'er ... I'll be back in the Morgan </font color=blue>

 

[Phelk]

~ Actually the OPTERON has to face the Itanium 2 ~

Don't forget Sun's UltraSPARCIII and IBM's Power4


<font color=blue> The Opteration was a success... I'm now a full-wit</font color=blue>

 

[IIB]

Eden, I hate to tell you this - but you speculate way to much....
its hard enough for chip-architects to forsee a processor preformance and even they dont get it right all the time.
I dont know how you get an exect number like 15% form added cache and dual chanell on di memory controler - I am safe to assume you (as me) dont know that much about cache architecture and memory lathncys and its effect on preformance on X86 CPUs. its a multi varible equasion which ppl study for PHds in order to solve.


This post is best viewed with common sense enabled

 

[eden]

I know I speculate too much, but given the average increases we see in other test examples like the P4 and 512K L2 (although not as effective for Athlon) I like to assume a little. I think it is also logical for a more advanced server version of a ClawHammer, that it has so much more performance. I will follow Anandtech's speculations for now, but still who knows if I am wrong, or right in the end?

--
Thunderbirds in wintertime, Northwoods in summertime! :lol:

 

[juin]

do you really think a X87-64 with a execution core from a K7 can beat a itanium even Merced.

SRI (systemes request interface)is very good but does AMD can make good server chipset.Even if AMD sledehammer can be faster that Xeon MP SATA infinidband PCI express will help Intel in real time.

Maybe server work can do a little chipset for AMD.They allready got a reputation.

cheap, cheap. Think cheap, and you'll always be cheap.AMD version of semi conducteur industrie

 

[FatBurger]

I'd have to agree with Juin, I don't think that Opteron (ugh, we need to come up with a good nickname) will be able to beat McKinley. And when you're buying a server from an OEM with several SCSI hard drives and 2GB of RAM, the price of the CPU is suddenly not nearly as important.

<font color=blue>If you don't buy Windows, then the terrorists have already won!</font color=blue> - Microsoft

 

[rcf84]

Well who long do you think it will take IA64 to get in Desktop Rig's. My guess is 10 to 15 years.

You GeForce Ti4xxx is faster then my R8500 but my R8500 is the king of Aniso baby

 

[rcf84]

Here's a good nick name "OPTy"

You GeForce Ti4xxx is faster then my R8500 but my R8500 is the king of Aniso baby

 

[mr_gobbledegook]

What I dont understand is AMD's performance claims of the Opteron processor.

In the Opteron PDF that AMD have released it states that the Opteron offers 25% more performance than the Athlon.

Lets be clear here...the Opteron (Sledgehammer) is the server version of the Hammer.

Since the <i>Athlon XP 64</i>(Clawhammer) will lack the dual channel memory controller and 1MB L2 cache (only 512kb) will this mean it will perform < 25% compared to a normal Athlon ??!

Also what core of Athlon is AMD refering to ? the Throughbred or Barton ? Barton is supposed to have 512KB of cache..so is Opteron supposed to be 25% faster than a Barton or Throughbred ?

AMD is obviously not making it clear about the real performance benifit we will see from <i>Athlon XP 64</i>(Clawhammer) and Opteron (Sledgehammer).

Are AMD sandbagging or are they trying to cover up the <i>hyped</i> performance claims of Hammer ?

I really dont know what to think...

<font color=purple>Ladies and Gentlemen, its...Hammer Time !</font color=purple>

 

[IIB]

I dont know what the hell are you talking about - show me one link where AMD specifcly indicates the IPC improvment of Clawhammer or sledgehammer compareed to the Athlon.

all numbers given are rumers/speculations. not AMD info.



This post is best viewed with common sense enabled

 

[mr_gobbledegook]

<A HREF="http://www.amd.com/us-en/assets/content_type/DownloadableAssets/Opteron_Press_Preso2.pdf" target="_new">OFFICIAL AMD OPTERON PDF DOCUMENT</A>

Turn to page 13 and you will find the following quote:

<i><font color=blue>"Note: The AMD <b>Opteron</b> processor is planned to provide <b>20-25%</b> better performance than the AMD Athlon processor -- 20% from on-chip low latency memory controller and 5% from improvements on processor core"

<font color=purple>Ladies and Gentlemen, its...Hammer Time !</font color=purple><P ID="edit"><FONT SIZE=-1><EM>Edited by mr_gobbledegook on 04/29/02 04:32 PM.</EM></FONT></P>

 

[IIB]

do you really think a X87-64 with a execution core from a K7 can beat a itanium even Merced.

it will be undoubtably faster then Merced on FPU preformance. it will be undoubtably slower then Itanium 2 on FPU. but Faster on Integer prefomance. this is widly known in the microarchitecture comunity.

infinidband PCI express will help Intel in real time.

Your talking totaly diffrente issues here...
none of those technlogys are aimed at elemnating the shared bus bottleneck... and thats the Main bootleneck in MP systems. esspacily that now even desktop chip-sets have very high bandwith between the south and north bridges.

there is no pont argueing - even in Intels own demonstation of the Itanium2 at CeBit - there was a very very limted preformance jump between 2way Itanium2 and 4way Itanium2. and this is on intells own, hand-picked, benchmarks.

This post is best viewed with common sense enabled

 

[IIB]

I downloaded the PDF - its the Hammer presentation slides from MPF2001. all I see in page 13 is hammers execution core block digram...


This post is best viewed with common sense enabled

 

[eden]

Wow either I've been wrong and ClawHammer has even worse performance than 20% increase, or AMD is getting numbers mixed. I mean if a mem controller WAS expected to give 20% base, then how in the world is Opteron with Dual Channel giving same thing?

--
Thunderbirds in wintertime, Northwoods in summertime! :lol:

 

[IIB]

hi eden, Have downloded the PDF? can you see the quote? coz I havn't found it...


This post is best viewed with common sense enabled