Pentium 4 EE 3,46 GHz lansert

[Boralis]

Tja , det blir øynene som ser, personelig synest jeg ikke den skuffer stort men er for dyr .

At du kun ser negative ting er dog ingen overraskelse .

[Knick Knack]

Foreløpig viser veikartene at alt utenom EE serien får 800FSB. Det er her OC potensialet ligger, om det var det du slet med å forstå, snorreh.

 

CPU med 800FSB + HK med 1066FSB. Ringer en bjelle?

Endret 2. november 2004 av Knick Knack

[snorreh]

Why are there so few 1066 prescott's planned?

So why is intel holding back on the 1066 bus prescotts?  You would think that intel would be excited about exploiting some of the advantages of DDR2, since they have already paid the price to move it into the market.

 

I don't think the reason is that intel is afraid of bothering the motherboard makers with certification issues.  And it doesn't seem that intel has problems making the 925xe chips, since they arn't really priced at a premium.

 

Or maybe it is tactical because they don't want AMD to come out with 466 or 500 mhz DDR1 althons.  Seems unlikely that intel is too worried about that though.

 

So could there be a technical problem with the prescotts?  What could it be?

I can't tell you how I know, but I do know it is not marketing, it is a technical problem. The yields on 1066 parts are miserable. 1066 mainstream chipsets were slated for last year, but they are not due before mid NEXT year now. The chipsets have been able to do 1066 for a long time, the Prescotts can't pull their weight.

[...]

Going from 800 to 1066 is a 33 % clock increase. Watching the very rare % clock increase we've seen in recent years from Intel, it might not that easy to raise FSB either.

 

As I said, maybe Prescott focus was too much on clock and not enough on FSB, since NW overclocks to 1066 quite often.

[...]

The initial Prescotts had trouble running at 800 fsb, remember the 533 ones ? Iirc it's something to do with voltage variations/drops at higher fsb.

[...]

I find it funny that even after moving to the LGA socket, which they said was done to enable FSB speeds up to 1600 MHz, they are having trouble making Prescotts that get up to 1066 MHz FSB reliably.

 

Pretty much invalidates LGA's usefulness.

[John Deere]

Syns selv det blir litt feil å sammenlikne FX 55 mot 3.46EE til Intel.

 

Intel burde hatt en 4.0ghz EE for at det skulle bli sammenlignbart..

Men det er vell det beste de har for øyeblikket så da er det ikke noen bedre å sammenligne med...

[Knick Knack]

Litt tidlig å påstå at det er store problemer med å nå 1066FSB. Revisjoner av chipset og CPU kan gjøre mye her og det er også så få OC tester å vise til. Å få en helt standard P4 6xx til å kjøre på 1066 i steden for 800FSB tar jeg ikke som et problem før jeg ser det i virkeligheten. Lite tyder foreløpig på at det blir et problem da produkter allerede eksisterer på disse spesifikasjonene og det vil komme ny revisjon av CPU innen det er aktuelt med denne overklokkingen.

 

Skylapper snorreh... slutt å beskylde andre for å ha det på før du har fått på motorsykkelhjelmen rett vei!

[snorreh]

Syns selv det blir litt feil å sammenlikne FX 55 mot 3.46EE til Intel.

 

Intel burde hatt en 4.0ghz EE for at det skulle bli sammenlignbart..

Men det er vell det beste de har for øyeblikket så da er det ikke noen bedre å sammenligne med...

Intel har kansellert P4 4.0GHz, men de kommer med P4 EE 3.73GHz i første kvartal neste år. Innen den tid har sikkert AMD kommet ut med Athlon 64 FX-57, så da blir det den man må sammenligne med...

[b0nna]

"Why are there so few 1066 prescott's planned?

I can't tell you how I know, but I do know it is not marketing, it is a technical problem. The yields on 1066 parts are miserable. 1066 mainstream chipsets were slated for last year, but they are not due before mid NEXT year now. The chipsets have been able to do 1066 for a long time, Intel makes damn good chipsets. The Prescotts can't pull their weight.

 

Now, this does raise a good question, was Northwood any better at OCing the FSB?"

 

http://www.aceshardware.com/forum?read=115109560

[Ueland]

Krangle om andre forumbrukere samt å diskutere andre forumbrukere kan man gjøre et annet sted....

 

da er dere advart

[snorreh]

Hvis disse testene skal brukes som referanser må de i det minste være semi-vitenskapelige, dvs. bl.a. foregå i ordnede former og gi reproduserbare resultater.

 

Selv med dagens "vitenskapelige" tester beskyldes diverse steder for å la sine egne meninger skinne gjennom. I en slik "reell opplevelse"-test som du snakker om er jeg redd at disse anklagene bare blir forsterket, og gjerne med god grunn også. Benchmarking er et evig kompromiss mellom kontrollerbare testmetodikker og testenes relevans i "real life".

Enig i det

 

I denne sammenheng er kanskje denne saken litt artig:

Intel wants to create new gaming benchmarks

HAVING been whipped by competitor AMD in video game-oriented benchmarks, Intel is proposing that real-world tests be conducted to determine the actual performance levels of today’s microprocessors.

 

Intel is pushing its Pentium 4 Extreme Edition 3.4Ghz chip for the gaming market while rival AMD has the Athlon 64 FX-53 at 2.4Ghz. These two companies have a history of being at loggerheads when it comes to benchmark tests, and most of the time they don’t agree with independent test results conducted on their products.

 

The latest benchmark results for Doom3, for example, show the Athlon 64 FX-53 leading the P4 Extreme Edition. The AMD chip registered a high of 104.4 frames per second (FPS) compared to Intel’s 100.7. Doom3 is a popular PC game developed by id Software and published by Activision, and the developers conducted their own benchmarks for the two chips using what is called a “time demo.”

 

“We’re proposing a real-world gaming test so that readers won’t be misled with time demo results,” says Matt Dunford, worldwide client benchmark manager for Intel Corp. “With a real-world benchmark, you have a human running the game instead of a prerecorded script. Realistic gameplay performance cannot be adequately measured using time demo benchmarks.”

 

In a time demo benchmark, a certain portion of a game is recorded for a certain amount of time. This is then run on the PC to measure the number of FPS the system is able to handle.

 

Dunford believes such a benchmark does not accurately measure chip performance.

 

“Time demos measure more of the graphics than CPU performance,” he says. “There are no physics, sound or artificial intelligence calculations, and these are the ones which put stress on the CPU. Time demos are more applicable when you want to test the video card.”

 

Time demos are recorded at 30Hz or 30 FPS. However, Doom3 runs at a maximum of 60 FPS. So even if a CPU achieves test results of more than 100 FPS, the game itself will only play at 60 FPS.

 

“The highest average frame rate alone does not equate to best gameplay,” Dunford says. “Gamers want a smooth, consistent gaming experience for the duration of the game. The problem is, frame rates can fluctuate, causing hiccups or slowdown during gameplay.”

 

In obvious reference to the higher mark achieved by AMD on the Doom3 time demo, Dunford stresses that even a high frame rate achieved during benchmarks can come with a low frame rate in some instances, making for an inconsistent gaming experience.

 

Dunford adds that even popular benchmarks such as 3DMark 2003 by Futuremark do not accurately measure CPU performance, but instead measures the performance of the graphics card.

 

Again, benchmarks by Digital Daily and other websites place the Athlon 64 FX-53 ahead of the P4 Extreme Edition using 3DMark.

 

As expected, Dunford also discredits the 3DMark results.

 

“It does not make sense to use 3DMark game tests to compare CPU performance,” he says. “It primarily measures graphics performance of the video card. Benchmarks should be up-to-date because hardware improvements take place rapidly and tests need to keep up with the changes. Benchmark tests need to account for changes in usage models.”

 

With existing benchmarks not going in their favor, Intel now wants AMD and game developers to play in Intel’s playground. This would include running CPU-related game tasks such as sound, artificial intelligence and physics on another “thread” or computational task. The P4 Extreme Edition features hyperthreading which performs tasks on another thread as if the system were running on two CPUs. AMD has no such feature in the Athlon.

 

Intel is proposing that benchmarks measure real-world gameplay where a gamer actually plays the game in real time instead of using a prerecorded script like what is done with time demos. The company is developing a “quality of service” metric which defines the ideal level of gaming experience as determined by a set of game quality attributes such as smoothness and responsiveness of the game.

 

“We’re in the process of collecting data from gamers and then we’ll analyze that data to identify trends,” Dunford says. “Then we’ll release the results of this analysis at the Intel Developers’ Forum and go back to the game developers and talk to them about coming up with real-world metrics. Then we’ll share these results with the press.”

 

Intel is essentially pushing for new benchmark tests specifically for gaming.

 

“My request is that people should be cautious on the use of time demos, be selective with the number of time demos being run and if you use 3DMark, use the CPU test in 3DMark 2005 and not the overall test, and then explain what exactly it is you are measuring, whether it’s the graphics or the CPU.”

[Malevolence]

hva er galt med timedemoer?

 

for øyeblikket er amd en soleklar leder, som intel var lenge før igjen.

 

men hvis intel får uit dualcore tidsnok, kan vi nok se en skikkelig ytelsesøkning...de 2 mb med cache tar vel omtrent like mye plass som en helt ny cpu

[sub1343]

Syns selv det blir litt feil å sammenlikne FX 55 mot 3.46EE til Intel.

 

Intel burde hatt en 4.0ghz EE for at det skulle bli sammenlignbart..

Men det er vell det beste de har for øyeblikket så da er det ikke noen bedre å sammenligne med...

det er en sammenligning av de raskeste fra hvert merke og viser hva det råeste du får i dag er.

[Knick Knack]

hva er galt med timedemoer?

 

for øyeblikket er amd en soleklar leder, som intel var lenge før igjen.

 

men hvis intel får uit dualcore tidsnok, kan vi nok se en skikkelig ytelsesøkning...de 2 mb med cache tar vel omtrent like mye plass som en helt ny cpu

Det er vel ikke hvem som er best eller at noen er bedre som i prinsippet avgjør om en testmetode er gal eller ikke. Problemet er at tester gjort under vanlig spilling har vist at timedemoer belaster systemet så forskjellig fra vanlig spilling at testresultatet ikke gjenspeiler det leserne leser inn i testen. mao. vinner du en timedemo så er det det maskina er god på ikke nødvendig vis det du hadde tenkt å gjøre. (med mindre en har som hobby å se på timedemoer...)

 

Det er også et problem at avg. FPS oppgis siden dette ofte ikke er av interesse i dag. om du får 70 eller 90 avg. FPS er irrelevant. Det som er interessant er hvor ofte en dropper under ca 60. Det viser seg også at høyere avg. FPS ikke nødvendigvis betyr at en dropper ferre ganger under 60. Årsaken er at det er forskjellig hvordan variasjon i FPS oppstår på forskjellige systemer.