"Opteron-killer" fra Intel

[b0nna]

Ansys?

http://translate.google.com/translate?hl=e...DUTF-8%26sa%3DG

 

Ser ikke så bra ut for itanium.

[Falcon_]

Jeg har alltid sammenlignet Itanium og ...

[The.Arisen.One]

"Opteron-killer"?

 

Si heller "Opteron-wannabe".

 

[Eplekakespising]

Jeg tror problemet med denne prosessoren er som ved nesten alle(om ikke alle) andre pentium prosessorer, nemlig prisen

[Caveman]

Skjønner ikke at folk kan være så naive. Hver gang detajer rundt nye prosessorer eller skjermkort kommer blir killer stempelet alltid satt. Men det ender om det samme hver eneste gang. Enten litt bedre eller litt dårligere. Det er aldri noe killer. Lærer dere aldri?

[Knick Knack]

Ansys?

http://translate.google.com/translate?hl=e...DUTF-8%26sa%3DG

 

Ser ikke så bra ut for itanium.

Kommer an på hvordan du leser testen. 1.3GHz Itanium 2 slår her 2.2GHz Opteron (2-way systemer). At Itanium ikke kan konkurrere på pris/ytelse i dette markedet vet alle. Det har med volum og ikke minst chipset/HK løsninger å gjøre. Med tanke på ytelse ser det imidlertid svært bra ut. 1.6GHz Itanium 2 blir tilgjengelig i disse dager.

[snorreh]

Ansys?

http://translate.google.com/translate?hl=e...DUTF-8%26sa%3DG

 

Ser ikke så bra ut for itanium.

Hehe

 

Den samme historien gjentar seg for hver eneste uavhengige test jeg ser av Itanium, og det bekrefter egentlig bare mine egne erfaringer med Itanium som gav veldig variabel ytelse sammenlignet med Opteron.

 

ANSYS var faktisk et av de første programvarehusene som portet programmene sine over til Itanium, og har brukt masse tid og ressurser på optimaliseringer til liten nytte (kundene har uteblitt). Etter stort press fra kundene deres så har ANSYS endelig tatt til fornuften og porter nå programmene sine også til AMD64 som gir vesentlig bedre pris/ytelse:

ANSYS to Deliver Cost-Effective 64-Bit Computing Software to Customers

Porting ANSYS 9.0 to the AMD Opteron processor-based platform dramatically improves speed and reliability, enables greater flexibility for high-performance computing

 

SOUTHPOINTE, Pa., Jun 22, 2004 /PRNewswire-FirstCall via COMTEX/ -- ANSYS, Inc., (Nasdaq: ANSS), a global innovator of simulation software and technologies designed to optimize product development processes, today announced that the upcoming ANSYS 9.0 release will be ported to the AMD (NYSE: AMD) Opteron processor running the SuSE version of the Linux operating system.

 

The AMD Opteron processor allows companies to preserve their investments in 32-bit applications, while providing a transition path to 64-bit computing for those projects demanding high-performance capabilities.

 

"With design cycles shortening, users want the highest-performing processors to keep up with their increasingly intense software needs," said Ben Williams, vice president of AMD's Enterprise and Server/Workstation Business Segment. "By porting ANSYS software to the AMD64 platform, ANSYS users will have the power and speed to quickly create their simulation runs and be unhindered by processor limitations."

 

In addition to being extremely fast, future ANSYS 9.0 customers who purchase systems based on the AMD Opteron processor will find it very cost effective. Plus, with the AMD Opteron processor price performance, customers using ANSYS products as their distributed domain solver or probabilistic design system may want to consider clustering their machines to run distributed processing applications.

 

Currently, the newly released ANSYS ICEM CFD and AI*Environment 5.0 is ported to the AMD Opteron processor platform enabling customer to access more than 2GB of memory, and thus, tackle problems of almost unlimited size. Additionally, ANSYS CFX-5.7 has been tested on AMD Opteron processor-based systems under both Windows and Linux, and has shown excellent stability and performance. Together, CFX-5 and the AMD Opteron processor have shown linear parallel scalability in Windows-based x86-class workstations. The next major release of CFX-5 will support the large memory access provided by the 64-bit capability of the AMD Opteron processor.

 

"Customers who try ANSYS on the AMD64 platform should be pleased with both the price and the speed at which they can run their design calculations on their systems," said Mike Wheeler, vice president and general manager, Mechanical Business Unit at ANSYS, Inc. "We realize the importance of having optimal performing systems to maximize the capabilities of our software, and we're committed to helping facilitate that."

Endret 27. juni 2004 av snorreh

[snorreh]

Hvor i huleste blir det av testene av denne såkalte "Opteron-killeren" ?

 

Eller er det nok en papirtiger fra Intel?

[snorreh]

Grunnen til mangelen på tester skyldes høyst sannsynlig at Nocona ikke er kompatibel med 64-bits Windows build 1069 som er fritt tilgjengelig:

http://www.infoworld.com/article/04/07/06/...wsnocona_1.html

The public beta for 64-bit Extended Systems was released in September 2003, before Intel had announced its plans to release the Nocona core.It has not been updated since then, and Microsoft's engineers wanted to ensure the beta operating system provided the same customer experience on Intel's chip as it does for AMD's chip before certifying it for Intel's chips.

 

There are some small differences between the instruction sets used in the two chips. Intel, for example, does not support AMD's 3DNow graphics instructions. Intel uses hyperthreading technology in its Xeon processors, but AMD does not.

 

AMD also uses two instructions designed to improve Opteron's ability to quickly switch back and forth between applications. The additional instructions were added after AMD published its design papers that Intel used to create the architecture for the Nocona chip. The instructions don't increase performance to any significant degree that most users would notice, according to analysts.

 

Aside from a few discrepancies, the chips are largely compatible. But even if they use the same instruction set, the chips don't have to use the same method to tell the operating system what instructions are available, said Kevin Krewell, editor in chief of the Microprocessor Report in San Jose, California.

 

For example, AMD's and Intel's 32-bit chips are compatible, but don't necessarily use the same code to let the operating system know what types of instructions are available, Krewell said. Right now, the beta version of the operating system is probably tuned to recognize only AMD's method of identifying its instructions, but it will be relatively easy for Microsoft to add support for Intel's code, he said.

 

Microsoft is working on a new public release of the 64-bit beta operating system, the company spokesman said, but customers who have access to Microsoft's Tech Beta program will be able to preview the 64-bit operating system on Intel-based workstations.

Build 1184 som jeg har brukt i noen måneder allerede er altså kompatibel med Nocona, men Microsoft har altså enda ikke oppdatert den versjonen som er offentlig tilgjengelig...

[b0nna]

Benches=> http://www.gamepc.com/labs/view_content.as...aopteron&page=1

[snorreh]

Benches=> http://www.gamepc.com/labs/view_content.as...aopteron&page=1

Interessant, men det var bare 32-bits tester dessverre og ytelsesforskjellene var som forventet (ref. Prescott vs. Athlon 64). Det blir mer interessant når vi får se 64-bits tester:

As for 64-bit performance, we can’t really tell anything at this time. While we were able to load our latest build of Windows XP 64-bit Edition on both systems and were able to run some quick 64-bit SiSoft Sandra benchmarks, these items combined are not enough to get a good picture of the 64-bit performance of these chips. We’re planning on re-visiting these processors in a future lab report with 64-bit Windows XP, Server 2003, and Linux, for a deeper look at performance and compatibility between these two processors. Stay tuned.

[snorreh]

Det er nå helt klart at Intel har gjort en særdeles dårlig jobb med å kopiere AMD64:

http://www.siliconinvestor.com/stocktalk/m...?msgid=20289129

One of the errata for Nocona shows that Intel did not implement 40-bit physical addressing in the CPU itself, irrespective of the chipset, but only 36 bit addressing.

 

23. Incorrect physical address size returned by CPUID instruction

Problem: The CPUID instruction Function 80000008H (Extended Address Sizes Function) returns the address sizes supported by the processor in the EAX register. This Function returns an incorrect physical address size value of 40 bits. The correct physical address size is 36 bits.

 

I looked through the description of memory management in the EM64T manual volume 1 and found dozens of places that incorrectly refer to a 40-bit physical address limit. These are flat out errors in this manual that should be corrected.

 

For example, the following paragraph is nearly a word-for-word copy from the AMD System Programming Manual for AMD64 (volume 2):

 

from ftp://download.intel.com/technology/64bit...ns/30083401.pdf

1-21 64-Bit Extension Technology Software Developer’s Guide, Volume 1 of ...

1.6.5.2. PAGING DATA STRUCTURES

The 64-bit extensions architecture expands physical address extension (PAE) paging structures to potentially support mapping a 64-bit linear address into a 52-bit physical address. In the first implementation of the 64-bit extension technology, PAE paging structures are extended to support translation of a 48-bit linear address into a 40-bit physical address.

 

This is totally incorrect for Nocona, and illustrates how Intel just blindly copied AMD documentation.

The document is rife with references to the supposed 40-bit physical address space, which doesn't exist in Nocona.

 

The copying also extended to the hardware, so that they had the CPUID instruction return the Hammer's physical address size, rather than the Nocona's physical address size.

I praksis betyr det at EM64T bare støtter maks 64GB fysisk minne, sammenlignet med opptil 1TB fysisk minne for AMD64...

Endret 13. juli 2004 av snorreh

[Kimble]

Håper den tar knerten på Amd.....

Håper det var ironisk ment Hadde det ikke vært for amd hadde intel enda pusshet ny prosessorer med 11MHz økning i frekvens et par ganger i året for å melke ut hver eneste krone av en gammel teknologi! Det samme hadde sikkert vært tilfelle om amd hadde vært alene. Er jo ikke attraktivt å bruke mange mrd. på forskning og nye fabrikker om en kan fortette å selge de gamle prosessorene sine til høye priser.

[mad_arab]

I praksis betyr det at EM64T bare støtter maks 64GB fysisk minne, sammenlignet med opptil 1TB fysisk minne for AMD64...

 

Å nei! Da er vi fortapt! Ingen vil noen gang kjøpe Intel mer, spesielt ikke desktop brukere, som vil bli hardt rammet! Og det kan jo aldri tenkes at Intel for forandre dette når det blir nødvendig.

 

Forlat skuta!

 

Her tøyer du strikken enda mer enn vanlig, snorreh. Blir du aldri lei av å plukke frem slike ting for å diskreditere Intel?

Endret 14. juli 2004 av mad_arab

[hanigjen]

Huff, husker PII266 kosta over 10k ved launch. Med AMD i markedet kan ikke Intel tillate seg å ta slike priser lenger.

 

Jeg kjøper gjerne Intel om de har best pris/ytelse, men de siste årene har det blitt AMD på meg. Den neste maskinen blir en vannkjølt FX på 939 for min del

 

Glemte nesten at jeg har en Dell P4@2Ghz slepbar

[Falcon_]

Her tøyer du strikken enda mer enn vanlig, snorreh. Blir du aldri lei av å plukke frem slike ting for å diskreditere Intel?

Rart du ikke skriver slikt mot en annen person her på forumet(en som hakker ned på AMD hele tiden og det med enda dårligere grunner enn snorreh) Du liker vel ikke intel bedre enn AMD

[mad_arab]

Falcon_:

 

Hvis det er personer du har problemer med, hvorfor ikke ta det med dem direkte, og hvorfor ikke si navnet på vedkommende du snakker om? Jeg skjønner ikke hva ditt problem med en annen bruker har med meg å gjøre, og ikke vil noen andre skjønne det før du forteller oss hva det dreier seg om.

 

Jeg reagerer på ensidig, gjentatt negativitet slik jeg påpekte her. Jeg har ikke sett tilsvarende fra noen andre, dersom man ser bort ifra enkelte fjortisser som popper opp i ny og ne, men aldri orker å holde på lenge. Fans irriterer meg ikke like mye som de som i tillegg konstant skal hakke på motparten, uansett hvor lite grunnlag de har.