Nehalem coverage from HKEPC

[epsilon84]

http://translate.google.com/translate?u=http%3A%2F%2Fwww.hkepc.com%2F%3Fid%3D568&langpair=zh|en&hl=en&ie=UTF8 (translated link)

Some bold performance claims there. 1.2x - 2.0x improvements in multithreaded performance would be amazing if true. 30% lower power consumption over Penryn is also impressive, given Penryn is already a very energy efficient process. The return of HT is also interesting, on P4s it was a positive feature for multithreaded applications but certain single threaded workloads took a performance hit. Let's see if they have managed to iron out the kinks of HT this time around.

 

[jimmysmitty]

They do state that it is a newer enhanced version of HT so I would imagine they would have gotten it to work well. I also wounder ho well it will work with the IMC and their "direct connect" style technology.

I can't wait for the octo core. 16 threads on one CPU. Imagine the responsiveness of that.

 

[Scotteq]

I can't wait for the octo core. 16 threads on one CPU. Imagine the responsiveness of that.

{Notes the lack of multithreaded consumer apps}

I guess it'll load Windows faster.... :lol:


Oh, nevermind... That's more limited by HDD anyhow...

 

[hairycat101]

when is this coming?

 

[wingless]

{Notes the lack of multithreaded consumer apps}

I guess it'll load Windows faster.... :lol:


Oh, nevermind... That's more limited by HDD anyhow...

I have to agree with you. Right now single threaded performance is the defining characteristic of a microprocessor. Lets take the Phenom for expample (gotta love it ). The Phenom was optimized for applications that can use all four cores. It does quite well in newer software titles that can use it's performance potential to the fullest like the new Unreal Engine (PROOF: http://www.lostcircuits.com/cpu/amd_phenom/20.shtml). However in anything that is single threaded oriented it just fails.

The ONLY relevant performance metric that you all care about is single threaded performance because that will matter most in your daily lives. Nehalem can have 8 cores and 16 threads but that is useless if single threaded performance fails. All I can say is it better live up to the 1.1x to 1.25x single threaded performance increase over Penryn. Its safe to say it probably will so you all shouldn't be too worried. My point tonight is only that talking about multiple threads is rather useless in today's software environment. Its fuggin sexy, but useless...

 

[hjjfffaa]

i agree wingless about how much of todays software is actually optimized for multicore processors maybe like 2% if that, i dont think we should worry about eight or so cores until all of our programs are able to run quickly and efficiently on our daul and quad cores

 

[wingless]

Since the software industry seems to adopt Intel's new tech faster than AMD's (3DNow!, 64-bit, Dual Cores/multi-threads), maybe we will finally see more multi-threaded applications. The ability to compute in 64-bit and 2+ threads has existed for a very long time now so I would love to finally see an abundance of programs that take advantage of even YESTERDAY's CPU tech (Athlon X2's).

I can't wait for Nehalem to come out though! Maybe, just maybe, we will finally be able to play Crysis on Very High settings then.

 

[jaguarskx]

when is this coming?

Last time I heard, Q4 2008.

Rumor has it ===> Q1/Q2 2009 due to lack of competition from AMD. But that's just a rumor.

 

[hewlettpucktard]

I wouldn't at all be surprised if Nehalem is delayed by AMD's implosion. That's a real shame as I would be one of the first adoptees. From what I've read of it, Nehalem is a significant leap forward in CPU performance and I would love to own one before 2009.

 

[OlSkoolChopper]

I'm trying to build a system this month that is designed for nothin more than to last til Nehalem. Im gonna be serously pissed off if it has to be much more than twleve months.