Intel Nehalem Working at 3.2GHz Pictured

[Cronos]

HT was somewhat useful for certain end-user desktop activity because there was only one core at that time.
It will be completely useless in native 4-core CPU like Nehalem. I can't see any real benefit from it. And as with its first implementation in P4, HT/SMT gonna hurt performance in all really optimized applications.

Massive improvement in bandwidth will definitely lead to massive improvement in certain number-crunching and server applications, way over 30%.
However, most desktop users and gamers should not hold any high expectations, there will be little to none improvement for their favorite applications.

[nonvtec]

Pure sex... I think I just wet myself!

[JumpingJack]

An amazing run through of the Nehalem Architecture at RWT:

http://www.realworldtech.com/page.cf...0208182719&p=1

Built from the ground up to support the SMT implementation, should be an order of magnitude better than the slapped on version of P4 in Willamette ... this will be something truly amazing to see if it works as intended.

It seems as though almost every part of Nehalem's pipeline has been tweaked, extended or somehow refined, except for the functional units. The bulk of the changes were made to the memory pipeline, to complement the changes in system architecture. However, the single biggest change and performance gain (made in the core) is simultaneous multi-threading which could improve server workloads anywhere in the range of 10-40%.

[shiznit93]

HT was somewhat useful for certain end-user desktop activity because there was only one core at that time.
It will be completely useless in native 4-core CPU like Nehalem. I can't see any real benefit from it. And as with its first implementation in P4, HT/SMT gonna hurt performance in all really optimized applications.

Massive improvement in bandwidth will definitely lead to massive improvement in certain number-crunching and server applications, way over 30%.
However, most desktop users and gamers should not hold any high expectations, there will be little to none improvement for their favorite applications.

For example, using arbitrary numbers, would you rather finish 4 workloads (such as a FAH unit) on 4 physical cores in 4 minutes, or 8 of those same workloads on 8 virtual cores in 6-7 minutes? That is why SMT is necessary.

For the average desktop I agree SMT is overkill but Nehalem was designed top down, they will disable features that don't make sense on certain platforms.

[NH|Delph1]

I don't think you understand. Hyperthreading is the marketing name for Simultaneous Multithreading. It is exactly the return of HT, but they didn't screw up the implementation this time.

Point taken, yet Gelsinger has said it out clear that Nehalem will not have any features previously known as Hyperthreading. It's misleading to keep calling it that since it was the marketing term for the implementation back then, and will make people believe it's the exact same thing, which we so far can't say for certain that it is.

//Andreas

[zakelwe]

Forgetting the pro's and cons of HT for a moment the reported speed of 3.2GHz is pretty good this far out. Unlike AMD Intel have that speed on the roadmap so it should debut at that speed. It being so fast so early might mean our newly found greed for 4GHz+ that was created by Penryn will also apply to Nehalem as well .

On the downside I'm going to guess that maybe SP 1M and perhaps even 3dmark2001 etc will not benefit much if at all. This is a pure guess though and could be entirely wrong.

I like the triple channel memory. It does not sound right at first but it shows somone has been thinking about what is needed and so it should do the trick nicely and remove a potential bottleneck. Bet Corsair and OCZ are rubbing their hands at the thought of all those 50% bigger packs they will be selling

Regards

Andy

[gOJDO]

IMO SuperPi will benefit from the 33% better instruction-level parallelism. The small L2 would be a drawback, but there is a L3, 2nd branch predictor and IMC that will compensate it.

I'm guessing around 15% shorter SPI times at same clock compared to Penryn.

[zakelwe]

IMO SuperPi will benefit from the 33% better instruction-level parallelism. The small L2 would be a drawback, but there is a L3, 2nd branch predictor and IMC that will compensate it.

I'm guessing around 15% shorter SPI times at same clock compared to Penryn.

Does anyone know if the L3 works at the same speed as the core and any indications what latency it works at? Both those would be good to know.

Thanks

Regards

Andy

[JumpingJack]

IMO SuperPi will benefit from the 33% better instruction-level parallelism. The small L2 would be a drawback, but there is a L3, 2nd branch predictor and IMC that will compensate it.

I'm guessing around 15% shorter SPI times at same clock compared to Penryn.

SuperPI will benefit from the additional work on the loop-branch predictor, SuperPI is inherently recursive -- one big loop ..

[gOJDO]

Does anyone know if the L3 works at the same speed as the core

It has separated PLL, so it operates at independent frequency.

and any indications what latency it works at?

30-40 depending of it's frequency and the the frequency of the cores.

http://www.realworldtech.com/page.cf...0208182719&p=7

[JumpingJack]

Does anyone know if the L3 works at the same speed as the core and any indications what latency it works at? Both those would be good to know.

Thanks

Regards

Andy

It's not entirely clear, but would appear so...

Nehalem’s 8MB and 16 way associative L3 cache is inclusive of all lower levels of the cache hierarchy and shared between all four cores. Although Intel has not discussed the physical design of Nehalem at all, it appears that the L3 cache sits on a separate power plane than the cores and operates at an independent frequency.

http://www.realworldtech.com/page.cf...0208182719&p=7

[Donnie27]

I know its the wrong thread but I was in a hurry. The one Euro isn't as big as a quarter but larger than a nickle. That die shot and it 731 million trannies is pretty amazing.

[Kaffebord]

I wonder what amount of ram will be the "standard" with this new plattform.

[clonez]

jeez i hope they dont make the same mistake as amd, demoing 3ghz+ systems and delivering >2.4 ghz

looking good though

[Donnie27]

jeez i hope they dont make the same mistake as amd, demoing 3ghz+ systems and delivering >2.4 ghz

looking good though

The real question is how much faster than 3.2GHz will Nehalem reach, not if it will reach 3.2GHz. I've talked to 3 8400 owners with easy 4 to 4.3GHz overclocks of Wolfdale without extreme or even water cooling.

[[XC] NetburstXE]

Source Hexus.net, DailyTech

I want one of these! Maybe in two years the price will be down to my level.

[Nedjo]

I wonder what amount of ram will be the "standard" with this new plattform.

well it's obvious that Intel wants to push size of 3 GB so it can obscure further advancement of x64bit OS's :p

[Jaco]

If you like this , how about a DP octo-core Nehalem system.
32 cores

[rusty]

how about this cpu ? http://www.sun.com/processors/UltraS...ex.xml?p=1&s=1

[clonez]

@ donnie27

shure, but that doesnt mean nehalem can reach that, too

i for instance still havent seen a peryn windows stable @ 8 ghz+(even on one core), which can (of course) be made by some pentium 4s

but it looks promising, should be a real monster if it can reach 4.5-5ghz @ LN2

[BiFfMaN]

so not till Jan-Mar time...oh well

[gOJDO]

how about this cpu ? http://www.sun.com/processors/UltraS...ex.xml?p=1&s=1

That CPU has nothing to do with x86 and general purpose computing.

[leomax]

http://www.theinquirer.net/gb/inquir...everywhere-idf

Inq sez 3.2Ghz Nehalem news is wrong

..The short story is that the boxes report as 2.53GHz but they are underclocked to 2.13....

...Reports of 3.2GHz boxes floating around are flat incorrect. The signs at the show were initially wrong, and people reporting that can't tell the difference between a Skulltrail or Harpertown box and a Nehalem rig by sight....

The proof they are offering is
http://www.hwupgrade.it/articoli/cpu...8-day-1_2.html

[Hornet331]

I wonder what amount of ram will be the "standard" with this new plattform.

still 2 or 4 gb.

mainstream platform "only" has dual channel memory.

[Shintai]

http://www.theinquirer.net/gb/inquir...everywhere-idf

Inq sez 3.2Ghz Nehalem news is wrong



The proof they are offering is
http://www.hwupgrade.it/articoli/cpu...8-day-1_2.html

They just saw a different system. DP system. Like the 3.2Ghz was the only one...Of you note their "edvidence" is a 16 logical CPU system, the other is an 8.

Theinq just showing their dumbass stupidity as usual in the material incarnation of the retard Charlie.