'Nehalem' 2.93 GHz Benches Revealed

whatlolvb0.jpg


Look at the quad core on the right :)
 
It doesn't matter what it is. The picture doesn't illustrate die size in anyway.
Doesnt matter if its single, dual or quad (which were all on LGA775).
That same LGA775 package could have a Pentium 4, Celeron, C2D, C2Q, etc.. all under the "hood".

Its about package size.

LGA1366 vs LGA775.

The big difference is the PACKAGING and the extra pins needed for tri-channel ram being directly connected to the CPU. And thats why Nehalem appears bigger, but I'm not sure if die size has been indicated. I'm pretty sure actual die space is smaller for Nehalem then 2 C2D's dies slapped together to make a Core 2 Quad
 
I don't there's much doubt that Nehalem is going to be kick ass just of matter what to get. I'm going TRY VERY HARD hold off building again till Windows 7 assuming a Q4 2009/Q1 2010 release. I think I've got enough power to easily get through the rest of the year, but I know I'm going to crave a Octo core when they come out! 16 threads with hyperthreading! The thing will be a beast for sure!
 
My thing is what desktop mainstream program could you possibly be using that would even take advantage of 8 cores let alone 16?

Grant it, even though you may not be able to find a program to stress 8 cores by itself, you can probably get a few programs together that will from time to time make it sweat. But 16 cores? Other than trying to break into the FBI, CIA, ATF, NASA, Pentagon, And Deparment of Defense servers, all at the same time, while folding of course ;),. what could you possibly be doing to be taking advantage of even half of that power in a non-server atmosphere?

Seriously though enlighten me, assuming you're not using 16 cores for some server app or something.
 
And that Tom's article was rather useless. Still thanks for the post. Not your fault that Tom's, whatever that article was supposed to be, sucked a hard lemon. I'm pretty sure I can get more facts about something from a horribly written unfactual article from the Inq than from that, whatever it was, from Tom's on Nehalem right there.
 
chrisf6969 said:
The big difference is the PACKAGING and the extra pins needed for tri-channel ram being directly connected to the CPU. And thats why Nehalem appears bigger, but I'm not sure if die size has been indicated. I'm pretty sure actual die space is smaller for Nehalem then 2 C2D's dies slapped together to make a Core 2 Quad
Click to expand...
Both yorkfield and bloomfield are built on a 45nm process.
Yorkfield is ~820M transistors with a die size of 2x107 mm^2.
Bloomfield is ~731M transistors. So basically they added HT, reworked some of the internals, AND added an IMC while losing some transistors (and half the cache, admittedly).

Id expect the die to drop a little bit, down to upper 100s or very low 200 based upon the same process and a little drop of transistors.
 
23% is pretty good but it's certainly not the 40-50% performance gain over Penryn some assumed. I'm happy with my Q6600 but Nehalem is going to kick some ass!
 
///M3 said:
23% is pretty good but it's certainly not the 40-50% performance gain over Penryn some assumed. I'm happy with my Q6600 but Nehalem is going to kick some ass!
Click to expand...
It's 23% over a dual-core 65nm CPU. If you read Tom's myriad of "updates" you'll see that it has 11% on a 65nm quad-core.

Weren't the 45nm quads supposed to have 10% on the quads? In any event, I trust Anand's numbers more than Tom's. Not to call him a liar, just that all he ran was 3DMark and PCMark. Anand ran, you know, real things like video encoders. Fancy that.
 
I, for one, am not waiting for this technology to be available. My system is a Socket A AMD 2800+. When I build my quad core Intel system, it will be plenty fast just because I'm used to this snail of a computer. LOL
 
http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3326

here's the anandtech preview of the nehalem from a month ago...i think it's still too early to take either of the reviews too seriously, but anand's article is a little more in-depth.

We've been told to expect a 20 - 30% overall advantage over Penryn and it looks like Intel is on track to delivering just that
Click to expand...

hopefully they do deliver that, and hopefully the overclocking is as delicious and robust as the penryns and conroes
 
Xaijin said:
Pretty sure I read that overclocking is going to be disabled.
Click to expand...

Yeah, current reports basically say that the FSB is locked at 133MHz with all of the different speeds being set by multipliers, which will probably be locked as well.
 
as of now only the cheaper socket 1160 cpus are reported as being fsb and multi locked. but it looks like you may have to go for 1k or more for an ee series cpu if you want to continue your hobby with intel. hello amd.
 
Yeah. s1160 is the locked one, all the s1366 is supposed to be overclockable.
Which is something that Id appreciate. With the multipliers being used, 22-24x? 4Mhz faster QPI and youve got another 100Mhz CPU. Going to 166QPI would take a 2.66 chip to 3.32. And thats the cheap one.

Well, s1366 and/or bloomfield and/or X58 is the unlocked thing.

Im wondering if P55 is going to be s1160 or s1366.

vanilla_guerilla said:
as of now only the cheaper socket 1160 cpus are reported as being fsb and multi locked. but it looks like you may have to go for 1k or more for an ee series cpu if you want to continue your hobby with intel. hello amd.
Click to expand...
And for the love of god Im tired of hearing this. Bloomfield is launching in 3 bins. Not all of them are EEs, and not all of them are $1k.
 
I heard that the S1366 frees up more clocking area because its OCable, but not that much. Is this true?
 
You must log in or register to reply here.
Back
Top