ja sam razocaran sto niti jedan od ova dva "renomirana" sajta nije dao rezultate testiranja pod VMWareom... No, verovatno nisu imali vremena da se bave instalacijom virtuelnih masina.
Verujem da nece mnogo proci dok neko ne izbaci VMWare testove. Tu ce da se vidi prava snaga Barcelone. A verujte Virtuelizacija ce "fora" koja ce prodavati servere sa Barcelonama i dolazecim Xeonima. Ma Virtuelizacija je buducnost serverskog trzista.
Sto se tice gaming testova i "simuliranja" Phenoma, ono sto je Anand dao je totalno neupotrebljivo. Em je koristio 1 CPU vs. 2 K8, koji imaju benefite od NUMA-e, em ce Phenom raditi sa DDR2-10667 memarom, sto ce itekako znaciti za performanse. Narocito ce biti zanimljivo videti Penom pod MultyGPU sistemom...
Ono sto je sigurno, jeste da kljuc performansi lezi u optimizaciji. AMD je to konacno shvatio pa je zato dve godine u napred izbacio specifikaciju za SSE5. Slaba vajda od mocnog SSE podsistem u K10, ako on ne bude adekvatno exponiran u softveru...
Virtualizacija nije samo buducnost - virtualizacija je realnost u tom segmentu trzista. Danas, kada iznajmis server, niko ti vise ne iznajmljuje fizicki server, vec virtualnu masinu koja radi u svom softverskom okruzenju (biras OS i sta vec). Jedan realan server moze da pokrene desetine virtualnih masina na kojima ce se vrteti web serveri. Sve to se poveze na storage uredjaj i imas neogranicen prostor za skladistenje podataka i neogranicen broj virtualnih servera koje mozes da iznajmljues zainteresovanim licima.
Naravno, dobra je stvar, ako Barsa, radi dobro virtualizaciju, ali razlog zasto nije detaljno istetstirana, ako je suditi po TG Daily-ju moze biti i ovaj...
Many regular review websites were not even given Barcelona samples. Many of those that were did not have adequate prep time to perform a wide range of benchmarks. Some websites even had to go outside of normal AMD channels to obtain samples for their own benchmarks. From all outward appearances, this product launch seems to be deliberately hampered by AMD's own hand more than anything else. For those websites which were able to get Barcelonas and have now published benchmarks the reason for the delay is apparent. On most benchmarks, existing Intel products (65 nm products) are able to exceed Barcelona's performance with very few benchmarks showing up where Barcelona wins.
Zanimljivo je i ostalo...
Power consumption is actually the only real area where we've seen notable wins. Or if you value price over performance then AMD's offering could be seen as a win as well.
Instead of providing power numbers based on the generally used "thermal design power" (TDP) rating, the company provides power consumption performance specifications for Barcelona based on ACP (Average CPU Power), a rating that AMD also introduced today. Rather than specifying the worst case scenario, the company told us that it is trying to provide a real world indication, how much power these CPUs will consume. While the HE versions of Barcelona have a TDP of 65W, AMD says that these processors will not consume more than 55W (ACP); the standard versions carry a TDP of 95W, but should not use more than 75W in real-world applications, AMD says.
AMD's website shows some cases where Barcelona is able to trounce its competition. We'll be putting together an article today which shows a real-world analysis compared to AMD's in-house benchmarks. Check back throughout the day for that article.
Simptomaticno mi je da npr. Xbit Labs nije objavio nikakav review - ocigledno na nisu hteli da budu prvi po svaku cenu. Mada, ko kupuje server kupovace ga zbog dodatne vrednosti koju platforma nudi (power consumption, eto ta virtualizacija itd...), a cini mi se da performanse i nesu toliko izrazene. Naravno, kljucna stvar je namena servera - u nekim poljima, jednostavno, performanse su alfa i omega i tu nema sporenja.
Takodje, vezano za K10, ali ne za samu Barsu, jeste da je i desktop segment upravo polje gde su performanse najbitnije - ljude boli uvo za potrosnju (voze jedan CPU i to ne 24/7), a virtualizacija im je bitna malo vise od lanjskog snega.
Anand, kao i AMD svojevremeno, bolje da nije ni radio onu nesretnu simulaciju, jer je besmislena, osim da pokaze kako eto mi smo Anand mozemo da experimentisemo. Naime, em je Core 2 radio sa FSB na 800, em je ”Phenom” radio sa ECC memorijom. Pored toga, zbog prirode arhitekture Core 2 CPU koji i dalje koristi FSB je vise osakacen suzavanjem magistrale sa 1333 MHz na 800 MHz nego ”Barsa-Phenom” izborom sporije memorije. Ipak, verovatno su hteli da osakate obe platforme.
