Exclusive: 'Nehalem' at 2.9GHz and X58

[apache_lives]

This all depends btw on wether the socket 1xxx is going to be mainstream or EE/Extreme only - this might just be a the ultra high end.

Overclocking all depends on how everything links up and wether intel will implement something to cap all our a$$es and might face the same limits as AMD when you raise the HT too high etc.

Whats so bad about the current mounting system? iv had no issues with the thousands of systems iv worked with/built, and besides the point it makes me alot of money when noobs cant get it right

 

[apache_lives]

wtf is with people thinking unmatched ram will work in the proper tri-channel setup? ALL STICKS HAVE TO BE MATCHING to get the full performance, that means 2+2+2gb = 6gb tri-channel, 2+2+4gb = SINGLE CHANNEL MODE (8gb) etc

3gb sticks? ever seen any sticks of ram that go up in binary etc? like 1.5gb sticks, or 768mb etc - hell no.

 

[dragunover]

Yeah,I thought the same...OC'ing may be limited.

I hope this becomes mainstream,then those companies hoarding DDR3 like the asses they are,will finally drop their prices.And,AND!Maybe make them perform higher than 10% better than my DDR2 800 mhz!

 

[Bullrun]

@apache_lives

http://www.computerbase.de/artikel/hardware/prozessoren/2008/bericht_intels_nehalem_architektur_ueberblick/

all Nehalems together,every sub-model on all sockets ...tells me that the mainstream socket is the L1160 with Lynnfield and Havendale (upper and lower ,mainstream) .L1366 ist the new "X" Series high end socket with Bloomfield (High-End) and Gainestown (EntryLevelServer?).Now i have suddenly another socket discovered . Never heard from it before , called L1567 and obviously equipped for (high-end?)-server-duty.

Im curious too what the OC cababilities of nehalem are.Looking OC was tried before , but somebody , somewhere, somehow has found a solution , everytime.... im sure its the same with nehalem.Maybe they didnt even try anything,clock speed isnt so important anymore.I would let the people clock what they want.Not clock speed but number of cores will make the difference in the future.The more cores you have , the more money you have to spend.A missing core is something nobody nowhere can with something replace. If you want the power you have to pay for it ... im sure thats gonna happen sooner or later
Mounting system is ok in my opinion,looks´strange at the beginning and i thought this not gonna last to long.... turns out i was completely wrong,S775 was far longer in action than any socket before ...despite special mounting tools i have like you a lot of work with badly damaged sockets. Its mechanicly not possible to install the cpu in any other way its was designed for ....a lot of people try this anywaySockets mostly too badly damaged to repair and can not be replaced .Entire board is down,no choice but to get another . L1366 will be far worse ,nearly same area and double amount of pins.Any damage on the pins are very hard to repair.Try to bend one back in position brings two more out of it...no chance

 

[Guest]

I think its funny that people are angry that there is ONLY a 30-40% Performance increase. Its relative. If you had a car that went 20 mph and then you were able to go 35 mph on the next car thats a nice 75% increase. now if your have a car that does 45 and now it does 60, thats still 15 mph but its down to a 33% increase. Quit your whining.

And AMD can go suck it btw.

 

[Guest]

I was hoping not to have to spend too much on memory (DDR2 is mainstream and very cheap) or on motherboards so I can put that money aside for a graphics card. But it seems Intel plans on pushing DDR3 (which I feel is overpiced) and a new socket that would mean I'd pay more for a motherboard.

I hope this new socket doesn't become mainstream and just stays for the real enthusiasts.

 

[Guest]

lobofanina: All I said was that we should wait for benchmarks before getting too excited about Nehalem, and my analysis of the article is spot on, no benchmarks, only pictures and stating "new" features that AMD already had. Is that not reasonable?

My last 3 computers: Phenom 9500, Core2Duo e6600, Athlon 64 3200+(socket 754). I can say that the AMD machines were great, and while the e6600 wasn't bad, the real-life performance of it was over-hyped to say the least, where as the performance of the Phenom was far greater than expected(my "real-life" performance is in the field of audio production)

 

[androticus]

Rain on the Parade:

1) Nehalem only goes *up* to DDR3 1333 -- but you can't even really buy 800, and newegg only has 2 pieces of 1066 -- there are more SKUs at 1600 than 1333. So the first gen parts have bottom-half performance, and will presumably be stuck there very hard.

2) DDR3 memory is still WAY more expensive than DDR2 -- still around 3x more -- would it really have been so hard for them to support DDR2???

 

[apache_lives]

Phenom 9500 vs Core2Duo e6600 vs Athlon 64 3200 - "real world" = general usage, which any cpu above the "3ghz" mark (amd 3000+ or P4 3ghz etc) is hard to tell apart, and comparing a dual to a quad isnt a fair comparison.

Intel doesnt want to do "an AMD" and get stuck with old memory tech and have to replace boards and sockets to update, so direct to ddr3 it is, and its a good idea at that, so long DDR2.

DDR2 is dirt cheap sure, as for DDR3 @ 3x the price, lest we forget how expensive ram was back in the day, and 3 x dirt cheap = still pretty cheap 😀

 

[psychicist]

Nehalem isn't about your standard pc anymore, but about finally taking Intel's x86 system architecture into the big league where HP Integrity (IA64), Sun/Fujitsu APL (SPARC64) and IBM's POWER6 servers rule. For single socket systems you'll see a modest performance increase but now Intel x86 solutions are finally viable for multi-socket systems as well. The old FSB technology prevented Intel systems from scaling up because of a.o. bus contention issues.

The lack of an integrated high-bandwidth memory controller also stopped their systems from excelling for I/O-bound workloads. Benchmarking the performance of games is not sufficient for that. Intel has also finally caught up with AMD's DCA (Direct Connect Architecture) thanks to QPI and possibly even passed them by thanks to technology that was standard or in development for RISC hardware 10 years ago.

 

[Guest]

"
Well, less than 10% until we really get into some nice multi-threaded programs. The gaming performance gap will surely be much bigger."

Games won't benefit from Nehalem any more than the apps that support upto 4 cores.And even then Nehalem will not be as compelling when compared to same clock Shanghai(AMD's 45nm chip).Shanghai will bring upto 3Ghz speeds in Q4 and will be roughly 10-15% on average faster than today's phenom,per clock.
So Nehalem better OC like a mofo or it will be noticeably bigger than Shaghai while performing marginally better in large number of cases and beating it by a good margin in a very few...

And just a side note on "Nehalem silicon" state:it is mosty(95%) finished product and performance is what you see.What is not finished is OCing functions and that is what will be done in the area of BIOS optimizations next couple of months.

 

[apache_lives]

Yeah_right82 you speak BS vs real performance figures, if that shiny new AMD ends up anywhere near 10-15% faster clock for clock they might finally have competition for a 1 year old cpu, named the Q6600 :lol:

Whats this less then 10% crap - check anandtech's review with a crippled chipset and look at the gains.

On the server front this will be interesting, Intels minimal lead might turn into an even greater lead then on the desktop front, depending on if they get it right...

 

[amberd1]

I do not think I will be in a hurry to get a X58 when they come out,even if it will be a true quadcore that will be in it.I have been surfing on the internet the wee hours of the morning,looking for info on the X58.I did not keep any book marks,but I seen a so called X58 that Micro Star International had and the cpu socket looked rectangular.Not like this articles square looking cpu socket.Although the DDR3 ram setup was the same as this articles one.Several other pages I read and I came across some article which made me think about this whole X58 problem.I think this LGA1366 will not make me interested for 2 to 3 years.I will wait for when Intels LGA1160 or LGA1060 comes out(I forgot the exact pin #).That is when intel and nvidia will supposedly get there act together and you may be able to buy an intel based chipset with nvidia's SLI chip features so you can have SLI in a Intel based mainboard.I have several computers that I tinker with and I am thinking of buying an Asus X48 or a XFX 790i ultra by nvidia for something new to tinker with.I will stick with intels duo core.I have another setup with quadcore.Quadcores are O.K. when you are downloading several movies and play games on the computer at the same time.Although I have done the same with duo core,but with less success.AMD's stuff has made me sad to buy in the past.When ever I bought one,intel came out with a faster superior cpu and AMD's cpu's I had trouble with anyway.If it is true,I will miss this old FSB routine.I like the idea of fiddling with the FSB.Anyway to cut a long story short,I am not interested in the upcoming X58 chipset mainboards at all.Please excuse the length of this post I have since this is my first....For those that will buy these X58 mainboard's,ENJOY.

 

[apache_lives]

uhh amberd1 you do realise what chipsets are, and in relation to cpus etc - chipsets cant be quad core, and we all saw what benifit a "true" quad core design had over a better designed dual die package setup had etc, and wtf is with your "duo core", i believe you mean a "Dual Core Core2 Duo"

The latter LGA design is for main-stream and will feature less memory channels and a supposed locked (un-overclockable) design whereas the larger socket will do tri-channel memory and will be unlocked or so it seems.

LOL MSI are a hopeless bunch they prolly do have a rectangular socket they implemented by accident.

 

[Guest]

This thing looks awssome...

I hope for AMD's sake that AM3 manages to match upto Nehalem. Else they are in for some serious thrashing.

But then again, I guess my upgrading can still wait... I am on my good old 2.66GHz P4, Intel D915GLVG with onboard GMA 900 and 256MB DDR RAM for what looks like ages now and its still offering moderately high amount of performance 😀

 

[Guest]

I'm the opposite, I was planning on 16GB in my next build, as 8 is only "a lot of RAM" and I wanted to go beyond that, but 16GB seems like quite a bit for personal computing for now. So 12GB would actually be rather nice 🙂

 

[eltoro]

Does anyone have any clue why THG is so persistent with using an outdated version of CPU-Z (v1.45.1), which doesn't identify fully any of the latest CPUs?
It happens often in THG articles.
Guys, if you need some assistant with downloading the latest CPU-Z release, I'll be happy to help

 

[apache_lives]

haha you can scope the net and see these benchmarks - every one i have seen only uses dual channel DDR3-1333 rather then the proper tri-channel setup, and there is still no mention of that hardware overclocking etc

what i want to see now is how 8 threads feel firing away with an almighty backbone - CSI, IMC etc feeding the cores PROPERLY

on a side note this is really more of a platform refresh like its the same cores as the conroe pretty much bar the IMC and FSB etc, the big questions is power consumption and heat etc now that its not using the fsb etc...